title
stringlengths 1
827
⌀ | uuid
stringlengths 36
36
| pmc_id
stringlengths 5
8
| search_term
stringclasses 44
values | text
stringlengths 0
8.58M
|
|---|---|---|---|---|
BRCA1 gene polymorphism and finger dermatoglyphic patterns in Ghanaian breast cancer patients: a quantitative cross-sectional approach | 3b8dd4f7-8a93-4e9a-8018-8fdefa8fc134 | 10024563 | Forensic Medicine[mh] | Cancer of the breast is a malignant disease that results from abnormal proliferation of cells within the breast tissue, which in advanced cases metastasize to other organs including bony skeleton, brain, lung and the liver . Globally in 2020 alone, breast cancer affected 2.3 million women with 685,000 deaths. This contributed to the estimated 7.8 million women alive who had been diagnosed with breast cancer in the preceding 5 years (2015-2020), making the disease the world´s commonest malignancy . In Africa, there is an increasing incidence of the disease with relatively high mortality due to late diagnosis and treatment. Breast cancer incidence in Ghana in 2020 was 4,482 out of which 2,055 died, a mortality rate of nearly 46% . Breast cancer, a heterogeneous disease, is representative of numerous subcategories of several cellular compositions, molecular alterations and clinical behaviour. The molecular subtypes of breast cancer are specified based on statuses of the oestrogen and progesterone receptors, human epidermal growth factor receptor 2 (HER2) and Ki-67 within the luminal and myoepithelial cells. They include; luminal A, luminal B, triple-negative, HER 2 enriched and normal-like breast cancers . It has been established that breast cancer development has genetic undertones with breast cancer susceptibility genes 1 and 2 (BRCA1 and BRCA2) as the most famous . Breast cancer type 1 gene is a tumour suppressor gene located on chromosome 17q21 and has DNA repair ability . Mutations in the gene has been associated with breast cancer development with reports suggesting that about 72% of women inheriting a mutated BRCA1 gene develop breast cancer by the age of 80 years . A common source of genetic mutation is single nucleotide polymorphisms (SNPs). Single nucleotide polymorphisms are genetic variations that occur at a specific locus within genomic sequence where a single nucleotide adenine (A), thymine (T), cytosine (C) or guanine (G) is altered . Single nucleotide polymorphisms of BRCA1 gene, therefore, result from a change of a single nucleotide within the genomic sequence of BRCA1 gene in a human population and presents as a homozygous wild, heterozygous carrier or a homozygous mutant. The mutant form of the SNPs of BRCA1 gene has been reported to be associated with the risk of breast cancer development . Finger dermatoglyphics is the scientific study of the skin ridge patterns on the palmar surface of the distal end of the fingers . Three (3) broad types of fingerprints patterns have generally been described: loop, whorl and arch . Fingerprint pattern formation is intrauterine and genetically influenced, and remains stable and unique for an individual throughout life. As a result, finger dermatoglyphics have gained tremendous forensic utility and for many decades have been explored for their usefulness in screening for genetic diseases, including familial breast cancers . While Chintamani et al . reported that six or more whorls were significantly higher among females with breast cancer compared to females without breast cancer, Natekar and colleagues observed six or more loops to be significantly higher among breast cancer females compared to healthy females. These reports among many similar others, give inspiration to consider finger dermatoglyphics as potential non-invasive anatomical marker for screening diseases including the assessment of risk of breast cancer development. The present study aimed at and reports on the relationship between the single nucleotide polymorphism of BRCA1 gene and finger dermatoglyphic patterns in breast cancer patients. The specific objectives included: (1) to determine the predominant location of single nucleotide polymorphism and the resulting nucleotide variants of BRCA1 gene (exon 11 and surrounding introns) among study participants; (2) determine the predominant finger dermatoglyphic patterns in individuals with breast cancer and individuals without breast cancer; (3) to ascertain tumour characteristics in individuals with breast cancer in relation to fingerprint patterns and; (4) to assess the predominant variants of BRCA1 gene in relation to the predominant finger dermatoglyphic pattern among individuals with breast cancer and individuals without breast cancer. We hypothesized that there was no relationship between finger dermatoglyphic patterns and BRCA1 gene polymorphism among individuals with breast cancer.
Study site, study design, and participant recruitment: a quantitative cross-sectional design was adopted for the study. Seventy (70) females clinically diagnosed with breast cancer and undergoing chemotherapy at the Oncology Unit of the Department of Surgery, Korle-Bu Teaching Hospital were recruited through simple random sampling. Additionally, seventy (70) age-matched apparently healthy females within the immediate environment of the Korle-Bu Teaching Hospital who met the inclusion criteria were recruited through simple random sampling. Control patients provided responses through interviews to exclude those with self-reported clinical conditions such as hypertension, diabetes, and cancers. Clinical and histopathological information including molecular sub-types on breast cancer patients were obtained from the hospital folders of each consented breast cancer participant. The sample size was estimated using the formula N = 2 ( Z α + Z β ) 2 σ Δ 2 α predicts the probability of making a Type I error and β predicts the probability of making a Type II error, where α is the significance level, α for this study = 0.05. Z α = A standardized normal deviation value that corresponds to a level of statistical significance (0.05) equals 1.96, Z α =1.96. β = power, probability of detecting a significant result (typically 80%, 90%). β for this study was 80%=0.20. Z β = probability of detection or power (80%), Z β =0.84. β is the standard deviation for the data (in similar studies, σ = 1). Δ = the effect size. Effect sizes as small, moderate, and large (0.2, 0.5, and 0.8 for two-group comparisons. Solving for N resulted in N= 63. The total number of breast cancer participants was rounded to seventy (70) with an equal number of apparently healthy participants. Fingerprint pattern acquisition: finger dermatoglyphic patterns were obtained using the ink method. Each participant was directed appropriately to dip the palmar surface of the distal part of the fingers of each of the ten (10) fingers in dry ink. The ten (10) inked fingers of each participant were then pressed on a piece of white paper, one finger at a time. The resulting fingerprints were identified and recorded. Blood sampling and preparation: by venipuncture using the median cubital vein, a total of 5 milliliters of peripheral blood was collected from each study participant into Ethylenediaminetetraacetic acid (EDTA) coated test tube. Samples were centrifuged at 3000rpm for 5 minutes to separate the blood components. The plasma was gently aspirated into an Eppendorf tube to make the buffy coat accessible. The buffy coat (containing the white blood cells) was then aspirated into fresh labeled Eppendorf tubes using a Pasteur pipette and placed in a cryo-box before storage at -20°C until use. Deoxyribonucleic acid (DNA) extraction: DNA was extracted from the white blood cells within the buffy coat using the Zymo Research Kit [Zymo Research, USA]. Five hundred (500) microliters of genomic lysis buffer [Inqaba Biotechnical Industries (PTY) Limited, South Africa] was added to 150 μl of thawed buffy coat. The mixture was vortexed for 5 seconds and then let to stand for 10 minutes at room temperature. The mixture was transferred to a zymo-spin column placed in a collection tube and centrifuged at 10,000 xg for 1 minute. The collection tube together with the flow-through was discarded. The zymo-spin column was transferred to a new collection tube. Subsequently, 200 μl of DNA pre-wash buffer was added to the spin column and centrifuged at 10,000 xg for one minute. Five hundred (500) microliters of gDNA wash buffer were then added to the spin column and centrifuged at 10,000 xg for 1 minute. The spin column was then transferred to a clean microcentrifuge tube. 40 μl of DNA elution buffer was added to the spin column and incubated for 5 minutes and then centrifuged at 10,000 xg for 30 seconds to elute the DNA. The purity and concentration of the eluted DNA were determined using a NanoDrop Spectrophotometer. The purity of 1.8-2.0 (A260/A280) and concentration of ≥20 ng/μl were required to qualify the eluted DNA as good. The extracted DNA served as a template DNA for the polymerase chain reaction. Polymerase chain reaction (PCR) procedure and conditions: polymerase chain reaction (PCR) was used for amplifying the DNA extracted using appropriate primers. A total reaction mixture of 25 μl was constituted for cycling in each reaction tube. It was composed of master mix (12.5 μl), nuclease-free water (10.5 μl), template DNA (1 μl), BRCA1 forward primer (0.5 μl) and BRCA1 reverse primer (0.5 μl). The reaction mixture was placed in a thermocycler (Prime thermal cycler, Bibby Scientific Limited) and ran for 29 cycles with the following reaction conditions: initial denaturation at 94°C for 5 minutes; denaturation at 94°C for 1 minute, annealing at 62.5°C for 3 minutes, an extension of DNA strand at 72°C for 3 minutes and final extension at 72°C for 3 minutes. The PCR products were subsequently subjected to agarose gel electrophoresis to visualize and confirm correct amplification. Agarose gel electrophoresis: agarose gel (1.5%) was prepared by weighing 1.5 g of agarose powder [Bio-Rad, USA] and mixed with 100 ml of Tris-acetate-EDTA (TAE) buffer. The 1.5% Agarose mixture was heated till boiling on a hot plate. The beaker was covered with aluminum foil to prevent evaporation of the mixture during heating. The mixture was then removed from the hot plate and allowed to cool after which 2 μl of ethidium bromide [New England Biolabs, USA] was added and then poured into a casting tray (with a comb fitted) and allowed to solidify. The electrophoresis tank was disinfected with 70% ethanol before the running buffer (1XTAE) was introduced. The solid gel which was within the casting tray was placed in the electrophoresis tank. Approximately 8 μl of PCR product from each sample was loaded into separate wells in the gel. A 100bp DNA ladder [Inqaba Biotechnical Industries (PTY) Limited, South Africa] was used as a molecular marker. Gel electrophoresis was run for 45 minutes at 100V. After gel electrophoresis, the gel with the amplicons was visualized using an Amersham imager (General Electric Healthcare Manufacturing Company, USA). Breast cancer type 1 gene sequencing: having confirmed the successful amplification of the target gene, the aliquots of PCR products of samples were shipped to Inqaba Biotechnical Industries (PTY) Limited, South Africa for BRCA1 gene sequencing. The forward strands of the amplified DNA (BRCA1 gene) of samples specifically exon 11, with upstream and downstream introns were sequenced by Sanger Sequencing. The BRCA1 forward and reverse primers were, respectively, 5’CACACAGCTAGGACGTCATC-3’ and 5’TCCATCAAGGTGCTTACAGTC-3’. Single nucleotide polymorphisms determination: when the sequencing results were received, Molecular Evolutionary Genetics Analysis software (MEGA X) was used to retrieve forward sequences of BRCA1 gene exon 11 and surrounding introns. Benchling® software was used to align the forward sequence of the BRCA1 gene (template or wild type) and the forward sequences of the amplified BRCA1 gene from study participants. After alignment of the forward sequences, nucleotides present on the amplified BRCA1 gene which mismatched the template sequence at specific locations were noted as SNPs. Statistical analysis: data was entered and cleaned using IBM SPSS Statistics 20. Data on molecular subtypes among breast cancer and fingerprint patterns were summarized using tables and analyzed with Chi-square. The mean frequency of fingerprint patterns among study participants was analyzed with independent samples student t-test using Graph pad prism version 8 software. Six or more whorls and six or more loops were analyzed with Chi-square. The normality test confirms the normal distribution of data (p > 0.05) before conducting the independent samples t-test. Differences in data sets with p < 0.05 were considered statistically significant. Ethical considerations: ethical approval for the study was obtained from the Ethics and Protocol Review Committee of the College of Health Sciences, the University of Ghana with Protocol Identification Number CHS-Et/M2-5.6/2019-2020. Participation in the study was voluntary and informed consent was sought before recruitment. Participants were given the liberty to withdraw from the study at any time with no consequences. Funding: the Department of Anatomy of the University of Ghana Medical School, partly funded the research through its postgraduate research support fund.
Age distribution among study participants: breast cancer females were age-matched with apparently healthy females; therefore, the age distribution was similar for both groups. More than 50% of the participants were pre-menopausal. The age groups of 20-30, 31-40, 41-50, 51-60, 61-70, 71-80 and 81-90 years recorded, respectively, 1, 18, 19, 19, 8, 4 and 1. The minimum and maximum ages of study participants (breast cancer females and apparently healthy females) were 29 and 85 years, respectively. The modal age was 47 years. Generally, individuals 40 years and below were classified as “younger women” while those 41 years and above were classified as “older women”. Of the 70 breast cancer patients, 19 (27%) were younger and 51 (73%) older. Clinico-pathological characteristics of breast cancer participants: in 50% of the breast cancer participants, the pathology affected the left breast, while in 47% of the cases, it was the right breast. In about 3% of the cases, the pathology was bilateral. In 11 of the 19 “younger” cancer patients, the left breast was affected while in 8 it was the right breast. The “older” group of 51, 24 and 25 individuals, respectively, had the left and right breasts affected and bilateral in 2 cases. Of the 70 breast cancer participants, 49 (70%) had been diagnosed with invasive carcinoma of no special type, 18.6% with invasive ductal carcinoma, and 4.3% with mixed carcinoma (mucinous carcinoma and invasive carcinoma of no special type). Other less frequent histopathological subtypes are shown in . Of the breast cancer participants, 48.6% were diagnosed with Grade II, 35.7% with Grade III and 12.9% with Grade I ( ). Stage IIIB was the commonest (34.3%) followed by stage IIIA (21.4%) and stage IIB (14.3%). Six (8.6%) of the participants were at stage IV ( ). Concerning molecular subtypes, luminal B was predominant among the patients ( ). Twenty-five (25) out of the 70 breast cancer patients representing 35.7% were diagnosed with luminal B. The second most frequent subtype was the triple-negative subtype (25.7%) followed by HER2 enriched subtype (21.4%) and then Normal-like breast cancer 9 (12.9%). The least frequently diagnosed molecular subtype was luminal A. For two of the cases, information on the molecular subtype was not available. Fingerprint patterns among study participants: generally, the loop was the predominant fingerprint pattern on all right fingers of breast cancer participants. In apparently healthy participants, the loop fingerprint pattern had the highest frequency on the right index finger, right middle finger, right ring finger and right little finger. However, there were more whorls than loops on the right thumb ( Annex 1 ). The right ring fingers of breast cancer participants did not have an arch fingerprint pattern. On the left hand, the loop was the predominant fingerprint pattern among breast cancer participants. In apparently healthy participants, the loop fingerprint pattern had the highest frequency of occurrence on the left index fingers, left middle fingers, left ring fingers and the left little fingers, however, there were more whorls than loops on the left thumb ( Annex 1 ). Among breast cancer participants, the mean frequencies of fingerprint patterns for loop, whorl and arch were 48.6, 14.9 and 6.5, respectively; and for the control participants, 42.2, 19.6 and 8.2 in the same order. There was, however, no significant difference between the two groups. Six or more loops and six or more whorls among study participants: comparing the presence of six or more loops or whorls among breast cancer participants and controls, no difference was observed. Although six or more loops were more common among breast cancer participants, compared to apparently healthy participants, but the difference was not statistically significant (p = 0.61). Also, six or more whorls had higher frequency among apparently healthy participants compared with breast cancer participants, but the difference was not statistically significant, (p = 0.30). summarizes the presence of six or more loops or whorls in relation to tumour characteristics in breast cancer patients. Out of 70 cancer participants, 51 recorded six or more loops. Of the 51, 19 (37.3%) diagnosed with luminal B had six or more loops. Regarding the tumour staging and grading, stage III and grade II individuals recorded the highest frequencies [32 (62.7%)] and [26 (51.0%)] respectively, of six or more loops. The association between the presence of six or more loops with tumour characteristics determined by Cramer´s V was very low. The highest Cramer´s V of 0.124 (p = 0.793) was observed between six or more loops and the breast cancer stage. Single nucleotide polymorphism of BRCA1 gene: shows the average size of the bands representing the amplified BRCA1 gene as approximately 450 base pairs. After alignment of the forward sequence of the amplified BRCA1 gene to the forward sequence of the wild type of BRCA1 gene, a heterozygous allele, c.34311A>C had the highest frequency of occurrence in both breast cancer participants and apparently healthy participants ( Annex 2 ). Variant location c.34320 was the second predominant location of nucleotide mismatch having two heterozygous alleles c.34320A>C (major) and c.34320A>T (minor) among study participants. The third predominant variant location was c.34321 and the resulting variant was c.34321A>T ( Annex 2 ). These variants c.34311A>C, c.34320A>C (major) and c.34320A>T (minor), c.34321A>T are introns located upstream of exon 11. Molecular subtypes of breast cancer and variant location of BRCA1 gene: out of the seventy (70) samples from breast cancer participants, 21 amplified DNA samples from breast cancer participants were selected through simple random sampling and sequenced. Out of the 21 breast cancer participants, 5 were diagnosed with luminal B breast cancer, 5 had triple-negative breast cancer, 5 had HER2 enriched breast cancer, 5 had normal-like breast cancer and one was diagnosed with luminal A breast cancer. From Annex 3 , all 5 participants with HER2-enriched breast cancer had c.34311 as the predominant variant location, 4 participants with triple-negative breast cancer had c.34311 as the predominant variant location, 3 participants with normal-like breast cancer had c.34311 as the predominant variant location and 2 participants with luminal B breast cancer had c.34311 as the predominant variant location. Of the seventy (70) breast cancer participants only one was diagnosed with Luminal A and had the variant locations c.34311, c.34320 and c.34321. Fingerprint patterns (six or more loops) and predominant BRCA1 gene variants: forty-eight per cent (48%) of breast cancer participants had six or more loops and c.34311A>C ( ). Also, 60% of apparently healthy participants had six or more loops and c.34311A>C. However, the difference in the frequency of occurrence of six or more loops and c.34311A>C among study participants was not statistically significant.
This study evaluated single nucleotide polymorphism of the BRCA1 gene and finger dermatoglyphic patterns in relation to tumour characteristics in breast cancer participants. BRCA1 mutations are responsible for approximately 40% of inherited breast cancers and more than 80% of families with both inherited breast and ovarian cancers. However, there has been limited study and therefore poor understanding of the role of BRCA1 in sporadic breast cancer. Several BRCA1 SNPs have been identified with some resulting in amino acid changes [ - ]. We report for the first time the existence of BRCA1 SNPs in Ghanaian females with breast cancer. The predominant variant locations observed are c.34311, c.34320, and c.34321 all resulting in introns located upstream of exon 11. Exons are the coding sequences of a genome whereas introns are usually the non-coding sequences and are both involved in gene expression . Single nucleotide polymorphism may occur in both exons and introns but alterations in the intronic sequences are more common than in exons . Though the functional significance of the polymorphic intronic sequences is unknown, it is predicted that single nucleotide variants occurring within intron branch point sites, especially at the position with adenine (A), would presumably affect splicing . Promoter regions are sites for the initiation of transcription of RNA and they are located upstream of exons and in eukaryotes usually contain TATA sequences which are bound by TATA-binding proteins to initiate transcription [ - ]. The three predominant variants c.34311A>C, c.34320A>C and c.34321A>T had adenine (A) replaced with cytosine (C), cytosine (C) and thymine (T), respectively. This would potentially disrupt BRCA1 transcription if any of the predominant variant locations were part of the promoter sites. Introns play a role in the regulation of gene expression and alternative splicing. Variations in intronic sequences may cause downregulation of gene expression and eventually lead to neoplasms . However, c.34311A>C which was predominant in both breast cancer females and apparently healthy females indicates that the variant may be naturally occurring. Though we acknowledge the small sample size for this study as a limitation, this pilot study serves as the basis for further work to provide relevant information on the variations of introns and their association with neoplasms among Ghanaians. In a participant with HER2-enriched breast cancer, however, a variant c.34474G>A was observed in exon 11 at c.34474 which resulted in an amino acid alteration; p. Arg1373Lys. The resulting amino acid, lysine is a basic amino acid just like arginine (the default amino acid) as such protein folding would not be adversely affected since lysine and arginine have similar qualities . Exon 11 of the BRCA1 gene is the site for frequent mutations in breast cancer individuals and has been reported among breast cancer patients in Africa including populations from Nigeria, Egypt, Tunisia, Morocco, Algeria and South Africa [ , , ]. Several studies have reported on the SNPs or mutations of the BRCA1 gene with regard to individuals with breast cancer in many countries [ , - ]. Such information is unavailable in Ghana and to the best of our knowledge, this is the first report from Ghana setting the tone for further studies. Of the different molecular breast cancer subtypes, triple-negative breast cancers reportedly common among younger women and African-Americans, are frequently characterized by BRCA1 mutations . In this study, however, HER2-enriched breast cancer recorded the highest total frequency of BRCA1 mutations of 10 ( Annex 3 ) followed by triple-negative. A subtype of HER2-enriched and luminal B breast cancers are noted for their fast growth and poorer prognosis and their high prevalence among the study population may provide a window of explanation to why breast cancer among Africans and Ghanaians are thought to be generally more aggressive . From Annex 3 , the frequency of occurrence of the types of fingerprint patterns in reducing order was a loop, whorl and arch, an observation consistent with the literature . Though fingerprint patterns are formed before the onset of diseases, they have over many decades been explored for their usefulness in diagnosis or predicting the risk of contracting certain diseases. We adopted and analyzed parameters such as six or more loops from Natekar et al . and six or more whorls from Chintamani et al . . Though six or more loops were higher among breast cancer participants compared to apparently healthy participants, the difference was not statistically significant since p > 0.05 ( Annex 1 ). Also, six or more whorls had higher frequency among apparently healthy participants compared to breast cancer participants however, the difference was not statistically significant, since p > 0.05 ( ). Breast cancer participants besides having no arch on the right ring fingers had a higher frequency of six or more loops. Apparently healthy participants had a higher percentage frequency of six or more loops in relation to c.34311A>C compared to breast cancer participants ( ) though the difference was statistically insignificant. Also, the higher percentage frequencies of six or more loops in relation to c.34320A>T and c.34321A>T among apparently healthy participants compared to breast cancer participants was noticed. We again emphasize that these observations may be affected by the small sample size of sequenced samples and thus, a larger sample size will be needed to validate our findings. Therefore, this information is insufficient to be used to explore the risk and prognosis of breast cancer, because six or more loops and the predominant BRCA1 gene variant, c.34311A>C were present in both breast cancer participants and apparently healthy participants and the difference in frequency of occurrence among study participants was not statistically significant. In summary, we explored the molecular subtypes of breast cancer in relation to BRCA1 gene variants and fingerprint patterns. Key findings include: (1) there was the absence of arch fingerprint pattern on the right ring finger of breast cancer participants; (2) six or more loops had higher frequency among females with breast cancer compared to apparently healthy females; (3) c.34311, c.34320 and c.34321 are the predominant BRCA1 gene variant locations in the study population with c.34311A>C being the predominant variant; and (4) a higher percentage frequency of six or more loops in relation to c.34311A>C was observed in apparently healthy females compared to breast cancer females. The findings of this study, however, are preliminary and required extensive study to improve the generalizability. Limitations in the study include: (1) the small sample size of the sequenced BRCA1 gene from study participants potentially missing other relevant information and limiting the generalizability of study findings; and (2) inability to study details of fingerprint patterns such as subtypes of loops, whorls and arches as well as ridge counts, which could provide further information; and (3) inability to determine the duration of the disease prior to clinical diagnosis. This is solely because the clinical diagnosis was contingent on when patients reported to the hospital.
This study sought to analyze the relationship between the single nucleotide polymorphism of BRCA1 gene and finger dermatoglyphic patterns in breast cancer patients and the potential usefulness of that information. We report that the predominant location in the studied population are c.34311 (c.34311A>C), c.34320 [c.34320A>C (major) and c.34320A>T (minor)] and c.34321 (c.34321A>T). The loop pattern was the commonest fingerprint pattern in both right and left hands in cancer patients similar to that exhibited by the controls. The study could not establish any tumour characteristic(s) that relate(s) to fingerprint patterns. The frequency of occurrence of six or more loops was common in c.34311A>C SNPs in both cases and controls. Besides the BRCA1 variants reported, the results are very preliminary and inconclusive and create an avenue for extended studies. What is known about this topic Breast cancer development is linked to mutant single nucleotide polymorphism of BRCA1 gene usually harboured within exon 11; Breast cancer has been linked to finger dermatoglyphics where certain patterns have been associated with breast cancer. What this study adds c.34311, c.34320 and c.34321 are the predominant BRCA1 gene variant locations in the study population with c.34311A>C being the predominant variant; Six or more loops had higher frequency among females with breast cancer compared to apparently healthy females.
Breast cancer development is linked to mutant single nucleotide polymorphism of BRCA1 gene usually harboured within exon 11; Breast cancer has been linked to finger dermatoglyphics where certain patterns have been associated with breast cancer.
c.34311, c.34320 and c.34321 are the predominant BRCA1 gene variant locations in the study population with c.34311A>C being the predominant variant; Six or more loops had higher frequency among females with breast cancer compared to apparently healthy females.
|
qPCR detection of viable | 5330c940-37a7-40a5-9053-baa8e367f5e3 | 10024758 | Microbiology[mh] | Although cosmetics do not need to be sterile, to protect the safety of people who use cosmetics, bacterial contamination above 1000 colony forming units (CFU/g or ml) is grounds for regulatory action in the United States (USA) and removal of contaminated products from the market; for products used around the eye area, this limit is 100 CFU/g or ml . As these grounds for action are culture-based, testing to assess cosmetics for contamination has also remained culture-based. In the United States, detection of B. cereus in cosmetics is typically performed using the culture-based methods described in the Bacteriological Analytical Manual (BAM) of the U.S. Food and Drug Administration (Tallent, Knollhoff et al. 2021). The reference techniques for cosmetics, described in Chapter 23 of the BAM, use selective and chromogenic media, biochemical confirmation, and phenotypic characterizations (e.g., cell and spore morphology, rhizoid growth, and the presence of specific proteins) for bacterial identification, which can take up to 7 days, including steps to pre-enrich cultures to support detection of low-level microorganisms . Although molecular methods are faster, the sensitivity of molecular methods means these methods may detect the presence of bacterial DNA whether or not the organisms from which that DNA comes are still alive . Specifically, preservatives in cosmetics products are expected to kill or damage microbial cells that may be present during manufacture or introduced during use by the consumer. The process of sample enrichment during testing of potentially contaminated products revives and amplifies damaged cells. DNA of these cells, along with that from any dead microorganisms, can be amplified by qPCR. This can result in overestimating the amount of live microorganisms’ present . This problem potentially prevents the accurate comparison to regulatory standards based on CFUs. Therefore, if rapid methods are to become accepted for cosmetics testing, some way must be found to preferentially identify live bacteria in cosmetics. One method for preferentially improving the detection of viable cells uses intercalating nucleic acid dyes that penetrate only into dead and compromised membrane cells. Once inside the cell, these dyes intercalate covalently into the DNA after photoactivation with light and thereby interfere with the amplification of that DNA. These dyes include ethidium monoazide , propidium monoazide , and propidium monoazide (PMAxx, Botium 2016), which can be used to pretreat samples prior to performing qPCR. A wide variety of microorganisms have been assessed using these dyes, including: viruses , fungi , protozoa , and foodborne pathogens such as Escherichia coli O157:H7 , Listeria monocytogenes and Salmonella ; Campylobacter jejuni , Staphylococcus aureus and B. cereus . About B. cereus. The B. cereus group species complex, also known as B. cereus sensu lato (s.l.), is a subgroup of closely related species belonging to the genus Bacillus . Group members are Gram-positive, spore-forming, and widely distributed throughout the environment . This B. cereus group comprises at least twelve closely related species: B. anthracis , B. cereus , B. thuringiensis , B. mycoides , B. pseudomycoides , B. weihenstephanensis , B. cytotoxicus , B. wiedmanni , B. toyonensis , and the recently identified B. paranthracis, B. pacificus, B. tropicus, B. albus, B. mobilis, B. Luti, B. proteolyticus, B. nitratireducens, B. paramycoides, B. gaemokensis, B. manliponensis, B. bingmayongensis , and B. fungorum . The best-known members of this group are B. anthracis , B. cereus and B. thuringiensis , each of which can have a significant impact on human health, agriculture, and the food industry . Rapid detection of B. cereus and its close relatives is important because B. cereus group members can produce a variety of hemolysins, phospholipase C (PLC), emetic toxins, enterotoxins, metalloproteases, collagenases, and beta-lactamases , that cause gastrointestinal diseases and localized infections, such as wound and eye infections . While B. cereus has not been a major concern for cosmetics, its presence is still undesirable, and developing rapid detection methods applicable to cosmetic products would be advantageous. Detection methods using PCR and qPCR have already been established for identifying B. cereus in foods, for example: targeting the 16S rRNA gene in naturally contaminated food gelatin , infant food , , and in food contact surfaces, such as cardboard and paper . The phosphatidylcholine-specific phospholipase C (PLC) gene has been detected in artificially contaminated liquid eggs and reconstituted infant formula . PLC gene was chosen in this study because of its prevalence in B. cereus and contribution in the pathogenicity of eye infection . To our knowledge, qPCR targeting the 16S rRNA and PLC genes have not yet been used formally to detect B. cereus in cosmetics. Therefore, here we present the development and use of two qPCR assays (simgleplex and multiplex) for the detection of B. cereus in powdered and liquid cosmetics, which may help bridge the gap between culture-based and rapid molecular methods of detection. To ensure our qPCR is fit for purpose to assess bacterial contamination in cosmetics, we first established the limit of detection (LOD) using a known B. cereus isolate from eye shadow cosmetic product (“3A”), then we performed inclusivity tests using a panel of 143 B. cereus group members to determine that our method can detect a broad range of B. cereus strains. Then we performed exclusivity tests using 38 non- B. cereus and 31 non- Bacillus to establish that the method did not amplify non-target strains or species. Finally, we assessed how well these qPCR assays (with or without PMAxx, and in both singleplex and multiplex formats) could detect the 16S rRNA and PLC genes of B. cereus in naturally or artificially contaminated cosmetics (powder and liquid types) and we directly compared this performance with outcomes using the standard culture method described in BAM Chapter 23 .
Bacterial strains and preparation The 212 bacterial strains used in this study were obtained from cosmetics, food, environmental sources, as well as from the American Type Culture Collection (Table ). All 212 strains used in this study were maintained at − 80ºC in broth supplemented with 20% glycerol. Each strain was aseptically streaked onto Tryptic soy agar (TSA) (Difco™, Franklin Lakes, NJ) and incubated for 24 h at 30ºC, from which an isolated colony was sub-cultured in Nutrient Broth (NB, pH 7.2) (Difco™), then incubated at 30ºC for 24 h. Inclusivity testing was performed using 143 members of the B. cereus group. One of these strains, B. cereus “3A”, previously obtained from eye shadow (Yossa and Jo Huang, personal communication), was used to establish the limit of detection (LOD) for our assays. Exclusivity testing was performed using a panel of 69 strains: 38 non-cereus strains of Bacillus , and 31 strains of non- Bacillus bacteria. Cosmetic products There were two types of cosmetics used in this study: liquid and powder. To assess how well our assays could detect the presence of B. cereus 3A in artificially contaminated samples of liquid-type cosmetics, we purchased facial toner products [N = 3], all of which were labeled “alcohol free”. These facial toners were primarily composed of water and plant-derived compounds and preserved with phenoxyethanol, an antibacterial agent. To assess the performance of our assays on powder-type cosmetics, we selected 8 cosmetic products: Green Clay (GC), Pink Clay (C1–4), Rice Powder (RP), and Tattoo Powder (O1–2), which had been purchased from a retail establishment online. The two clay products contained no preservatives, the rice powder product contained phenoxyethanol, and no information was available about any preservative in the tattoo product. Prior research had determined 6 powder-based products to be contaminated with B. cereus (Yossa and Jo Huang, personal communication); therefore, we classified these as “naturally-contaminated” and did not add any additional bacteria to these powders. DNA extractions Inclusivity, exclusivity, limits of detection We isolated DNA from overnight cultures of each pure bacterial strain using the MagMAX™ Express 96 Magnetic Particle Processor (ThermoFisher) with PrepSEQ Nucleic Acid Extraction Kit for Food and Environment (ThermoFisher, P/N 4,428,176) using protocol PrepSeq_ResDNA_20011 (Life Technologies, Carlsbad, CA). Total DNA extracted from these pure cultures were used to evaluate the inclusivity (n = 143) and exclusivity (n = 69) of the qPCR assays. To determine the limit of detection of our qPCR assays, we used DNA extracted from the B. cereus 3A strain, which was grown overnight in NB and adjusted to a density of 0.5 ± 0.05 MacFarland (McF; ~ 6.7 log CFU/ml). DNA extractions from cosmetics Two different extraction procedures were used to perform our qPCR assays on the sets of contaminated cosmetics, due to differences in the matrices. The MagMAX™ and PrepSeq kits were only appropriate for liquid cosmetics. Effective DNA extraction from the powder products required using the DNeasy® PowerPlant® Pro kit (Qiagen, Catalog Number 69204) instead, because powder products become thick and dense in the presence of the proteinase/proteinase DNA extraction buffer used in the MagMAX and PrepSeq kits . Primers/probes Table shows the primers we used and the associated publications documenting their first use. These were purchased from Life Technologies (ThermoFisher). The 16S rRNA sequences (P/N 4,331,348, Life Technologies) used here had previously been determined by De Clerck, et al.; their team targeted the 5’ hypervariable fragment, which can be amplified using a universal forward and a B. cereus -specific reverse primer . Our TaqMan probe was labeled with 6 -carboxyfluorescein (FAM) reporter dye at the 5’ end and labeled at the 3’-end with a Black Hole quencher, BHQ1, to reduce background fluorescence . We ordered the PLC primer probe (CCU001SNR, Life Technologies) labeled with a JUN reporter dye at the 5’-end and a non-fluorescent quencher at the 3’-end, to facilitate multiplex qPCR using 4 dyes. QSY PLC- JUN probe is a custom probe, compatible with the TaqMan Multiplex Master Mix (P/N 446,188; Applied Biosystems, Life Technologies) and ordered through Life Technologies. The TaqMan Internal Positive Control (IPC) from Life Technologies (P/N 4,308,323) was included to monitor the PCR progress and ensure that a negative result is not caused by failed PCR in the sample and used with the TaqMan Multiplex Master Mix to amplify both the single target and the multiple target reactions. qPCR reactions of non-treated and PMAxx -treated samples PMAxx -PCR is an innovative technology that allows differentiation between live and dead microorganisms, based on the loss of cell membrane integrity in dead cells . This system uses a DNA-intercalating dye, PMAxx, that disrupts DNA transcription only in dead cells, as their damaged cell membranes permit entry of the dye. After photoactivation with a defined wavelength, PMAxx intercalates and binds covalently to DNA. Subsequent amplification of that modified DNA is inhibited, thereby reducing the amplification signal from dead/damaged cells in comparison to that from live cells. For each qPCR reaction, a 20 μl volume consisting of 10 μl Multiplex Master Mix, 2 μl 10X Exo IPC Mix, 0.4 μl 50X EXO IPC DNA, 1 μl of primer assay, 2 μl of sample/template DNA, and 4.6 μl of sterile deionized water was placed into a 96 well fast plate. Both primers for the multiplex PCR were combined to reach a final concentration representing 5% of the total reaction volume. These qPCR runs were performed on a 7500 Fast qPCR System (Applied Biosystems) under the following conditions: 2 min at 50 °C, then 10 min at 95 °C, followed by 40 cycles of 15 s at 95 °C and 1 min at 60 °C. B. cereus 3A testing in artificially contaminated liquid products Here we used 3 different cosmetic toner products, which we artificially contaminated with B. cereus 3A. For each trial (N = 3), 3 sterile Wheaton bottles (125 ml) were filled with 30 ml of facial toner and were repeatedly spiked with 300 μl of microbial suspension, containing either a High (~ 4 or 3 log CFU/ml) or Low (~ 3 or 2 log CFU/ml) level of B. cereus 3A vegetative cells (optical density of 0.5 ± 0.05 McF), over three consecutive days, and the 3rd bottle of sample remained un-spiked for the negative controls. These repeated inoculations were necessary to enable recovery of B. cereus from products containing preservatives . To mimic a product contamination event more closely, these inoculated toner samples, along with the un-spiked negative control, were aged for 14 days at room temperature. After 14 days of aging, the un-inoculated and inoculated samples from the High- and Low-level inoculation were gently separately mixed, then 1 ml aliquots were diluted with 9 ml MLB (Difco). Our first goal was to achieve fractional results, which in this case would consist of 5 replicates at the high level of inoculation all yielding positive results, while 20 replicates at the low level of inoculation would yield only 50 ± 25% positive results, and 5 replicates of un-inoculated sample all yielding negative results . Based on this, 5 replicates of toner from the High level, 20 replicates from the Low level, and 5 replicates from the uninoculated samples (negative controls) were pre-enriched for 24 h at 30 °C in MLB, streaked onto BACARA plates and processed simultaneously for molecular analysis. Two equal sets of test portions were taken at the same time in all the replicates and placed into vials, and the first set of test portions proceeded to genomic DNAs extraction without any treatment. From the second set, 1 ml of each test portion was supplemented with 25 µM PMAxx, mixed through tap-spin, and allowed to incubate at room temperature for 10 min in the dark before exposure on the PMA-Lite (Biotium) for 15 min to crosslink free DNA . Immediately following PMAxx treatment, genomic DNAs were extracted, and all genomics (non-treated and treated) were analyzed independently in duplicate using singleplex and multiplex qPCR assays. Figure underlines an overview of the workflow of this portion of our study. Microscopic observation of live/dead B. cereus cells aged for 14 days in toner-samples using BacLight kit To provide a microscopic perspective on the ratio of live/dead cells in the inoculated toner samples, 1 ml from the 14-day aged samples inoculated with B. cereus at the high level was centrifuged at 10,000 × g for 10 min. The supernatant was removed, and the pellet was resuspended in 1 ml of 0.85% sodium chloride (NaCl) and kept at room temperature for 1 h, mixing every 15 min. After centrifugation at 10 000 × g for 10 min, the cells were resuspended in NaCl and stained with the Live/Dead™ Bacterial Viability kit (Cat: L7007, Invitrogen by ThermoFisher Scientific) according to manufacturer’s protocols . Stained cells were viewed under a Zeiss 880 Laser Scanning Microscope (LSM, Carl Zeiss Microscopy, LLC, White Plains, NY). The cells were observed using a Zeiss Axio Observer inverted microscope with × 63 1.4 NA oil immersion plan apochromatic objective. Differential interference contrast (DIC) and confocal fluorescence images were acquired simultaneously. A photomultiplier tube captured the light emitted from a 488-nm argon laser with a 3.7-m pin hole passing through an MBS 488 filter with limits set between 472 and 562 nm for detection of SITO 9 stains for green fluorescence of live cells and between 597 and 669 nm for detection of propidium iodide stains for red fluorescence of bacterial cells with damaged membrane. Zeiss Zen Black software was used to obtain the images with 1024 × 1024-pixel resolution. B. cereus testing in powder-type cosmetic products Each of the 8 powder-types cosmetic products (GC, C-1, C-2, C-3, C-4, O-1, O-2 and RP) was aseptically homogenized, and 1 g was added to 1 ml of Tween 80, before mixing with 8 ml of MLB broth in presence of 10 sterile glass beads to ameliorate homogenization. These samples were diluted, and the 10 −1 –10 −2 dilutions were spread on MLA and BACARA plates, before enrichment at 30 °C for 24 h. Spread plating was used instead of spiral plating, to prevent powder particles from clogging the stylus of the spiral plater. All these enriched samples were streaked onto BACARA plates to screen for B. cereus presence, according to BAM Chapter 23 and processed simultaneously to extract bacterial DNA for B. cereus qPCR analysis, in duplicate. This will permit cross-method comparisons. Statistical analysis Data were analyzed using a linear mixed effects model and p -values were adjusted to account for multiple comparisons. Cutoff points were established using a logistic regression model, such that the cutoff Ct value gives a > 50% chance of a sample being negative. Inclusivity and exclusivity cutoff values were established using the upper bound of the 95% confidence interval of the 95th quantile estimate . The linear mixed effects model was fit using the nlme package version 3.1–145 . Quantile regression was done using the quantreg package version 5.55. All other analyses were done using R version 3.6.0 . Ethics declarations This research does not involve human, animal, seed or plant samples.
The 212 bacterial strains used in this study were obtained from cosmetics, food, environmental sources, as well as from the American Type Culture Collection (Table ). All 212 strains used in this study were maintained at − 80ºC in broth supplemented with 20% glycerol. Each strain was aseptically streaked onto Tryptic soy agar (TSA) (Difco™, Franklin Lakes, NJ) and incubated for 24 h at 30ºC, from which an isolated colony was sub-cultured in Nutrient Broth (NB, pH 7.2) (Difco™), then incubated at 30ºC for 24 h. Inclusivity testing was performed using 143 members of the B. cereus group. One of these strains, B. cereus “3A”, previously obtained from eye shadow (Yossa and Jo Huang, personal communication), was used to establish the limit of detection (LOD) for our assays. Exclusivity testing was performed using a panel of 69 strains: 38 non-cereus strains of Bacillus , and 31 strains of non- Bacillus bacteria.
There were two types of cosmetics used in this study: liquid and powder. To assess how well our assays could detect the presence of B. cereus 3A in artificially contaminated samples of liquid-type cosmetics, we purchased facial toner products [N = 3], all of which were labeled “alcohol free”. These facial toners were primarily composed of water and plant-derived compounds and preserved with phenoxyethanol, an antibacterial agent. To assess the performance of our assays on powder-type cosmetics, we selected 8 cosmetic products: Green Clay (GC), Pink Clay (C1–4), Rice Powder (RP), and Tattoo Powder (O1–2), which had been purchased from a retail establishment online. The two clay products contained no preservatives, the rice powder product contained phenoxyethanol, and no information was available about any preservative in the tattoo product. Prior research had determined 6 powder-based products to be contaminated with B. cereus (Yossa and Jo Huang, personal communication); therefore, we classified these as “naturally-contaminated” and did not add any additional bacteria to these powders.
Inclusivity, exclusivity, limits of detection We isolated DNA from overnight cultures of each pure bacterial strain using the MagMAX™ Express 96 Magnetic Particle Processor (ThermoFisher) with PrepSEQ Nucleic Acid Extraction Kit for Food and Environment (ThermoFisher, P/N 4,428,176) using protocol PrepSeq_ResDNA_20011 (Life Technologies, Carlsbad, CA). Total DNA extracted from these pure cultures were used to evaluate the inclusivity (n = 143) and exclusivity (n = 69) of the qPCR assays. To determine the limit of detection of our qPCR assays, we used DNA extracted from the B. cereus 3A strain, which was grown overnight in NB and adjusted to a density of 0.5 ± 0.05 MacFarland (McF; ~ 6.7 log CFU/ml). DNA extractions from cosmetics Two different extraction procedures were used to perform our qPCR assays on the sets of contaminated cosmetics, due to differences in the matrices. The MagMAX™ and PrepSeq kits were only appropriate for liquid cosmetics. Effective DNA extraction from the powder products required using the DNeasy® PowerPlant® Pro kit (Qiagen, Catalog Number 69204) instead, because powder products become thick and dense in the presence of the proteinase/proteinase DNA extraction buffer used in the MagMAX and PrepSeq kits . Primers/probes Table shows the primers we used and the associated publications documenting their first use. These were purchased from Life Technologies (ThermoFisher). The 16S rRNA sequences (P/N 4,331,348, Life Technologies) used here had previously been determined by De Clerck, et al.; their team targeted the 5’ hypervariable fragment, which can be amplified using a universal forward and a B. cereus -specific reverse primer . Our TaqMan probe was labeled with 6 -carboxyfluorescein (FAM) reporter dye at the 5’ end and labeled at the 3’-end with a Black Hole quencher, BHQ1, to reduce background fluorescence . We ordered the PLC primer probe (CCU001SNR, Life Technologies) labeled with a JUN reporter dye at the 5’-end and a non-fluorescent quencher at the 3’-end, to facilitate multiplex qPCR using 4 dyes. QSY PLC- JUN probe is a custom probe, compatible with the TaqMan Multiplex Master Mix (P/N 446,188; Applied Biosystems, Life Technologies) and ordered through Life Technologies. The TaqMan Internal Positive Control (IPC) from Life Technologies (P/N 4,308,323) was included to monitor the PCR progress and ensure that a negative result is not caused by failed PCR in the sample and used with the TaqMan Multiplex Master Mix to amplify both the single target and the multiple target reactions. qPCR reactions of non-treated and PMAxx -treated samples PMAxx -PCR is an innovative technology that allows differentiation between live and dead microorganisms, based on the loss of cell membrane integrity in dead cells . This system uses a DNA-intercalating dye, PMAxx, that disrupts DNA transcription only in dead cells, as their damaged cell membranes permit entry of the dye. After photoactivation with a defined wavelength, PMAxx intercalates and binds covalently to DNA. Subsequent amplification of that modified DNA is inhibited, thereby reducing the amplification signal from dead/damaged cells in comparison to that from live cells. For each qPCR reaction, a 20 μl volume consisting of 10 μl Multiplex Master Mix, 2 μl 10X Exo IPC Mix, 0.4 μl 50X EXO IPC DNA, 1 μl of primer assay, 2 μl of sample/template DNA, and 4.6 μl of sterile deionized water was placed into a 96 well fast plate. Both primers for the multiplex PCR were combined to reach a final concentration representing 5% of the total reaction volume. These qPCR runs were performed on a 7500 Fast qPCR System (Applied Biosystems) under the following conditions: 2 min at 50 °C, then 10 min at 95 °C, followed by 40 cycles of 15 s at 95 °C and 1 min at 60 °C. B. cereus 3A testing in artificially contaminated liquid products Here we used 3 different cosmetic toner products, which we artificially contaminated with B. cereus 3A. For each trial (N = 3), 3 sterile Wheaton bottles (125 ml) were filled with 30 ml of facial toner and were repeatedly spiked with 300 μl of microbial suspension, containing either a High (~ 4 or 3 log CFU/ml) or Low (~ 3 or 2 log CFU/ml) level of B. cereus 3A vegetative cells (optical density of 0.5 ± 0.05 McF), over three consecutive days, and the 3rd bottle of sample remained un-spiked for the negative controls. These repeated inoculations were necessary to enable recovery of B. cereus from products containing preservatives . To mimic a product contamination event more closely, these inoculated toner samples, along with the un-spiked negative control, were aged for 14 days at room temperature. After 14 days of aging, the un-inoculated and inoculated samples from the High- and Low-level inoculation were gently separately mixed, then 1 ml aliquots were diluted with 9 ml MLB (Difco). Our first goal was to achieve fractional results, which in this case would consist of 5 replicates at the high level of inoculation all yielding positive results, while 20 replicates at the low level of inoculation would yield only 50 ± 25% positive results, and 5 replicates of un-inoculated sample all yielding negative results . Based on this, 5 replicates of toner from the High level, 20 replicates from the Low level, and 5 replicates from the uninoculated samples (negative controls) were pre-enriched for 24 h at 30 °C in MLB, streaked onto BACARA plates and processed simultaneously for molecular analysis. Two equal sets of test portions were taken at the same time in all the replicates and placed into vials, and the first set of test portions proceeded to genomic DNAs extraction without any treatment. From the second set, 1 ml of each test portion was supplemented with 25 µM PMAxx, mixed through tap-spin, and allowed to incubate at room temperature for 10 min in the dark before exposure on the PMA-Lite (Biotium) for 15 min to crosslink free DNA . Immediately following PMAxx treatment, genomic DNAs were extracted, and all genomics (non-treated and treated) were analyzed independently in duplicate using singleplex and multiplex qPCR assays. Figure underlines an overview of the workflow of this portion of our study. Microscopic observation of live/dead B. cereus cells aged for 14 days in toner-samples using BacLight kit To provide a microscopic perspective on the ratio of live/dead cells in the inoculated toner samples, 1 ml from the 14-day aged samples inoculated with B. cereus at the high level was centrifuged at 10,000 × g for 10 min. The supernatant was removed, and the pellet was resuspended in 1 ml of 0.85% sodium chloride (NaCl) and kept at room temperature for 1 h, mixing every 15 min. After centrifugation at 10 000 × g for 10 min, the cells were resuspended in NaCl and stained with the Live/Dead™ Bacterial Viability kit (Cat: L7007, Invitrogen by ThermoFisher Scientific) according to manufacturer’s protocols . Stained cells were viewed under a Zeiss 880 Laser Scanning Microscope (LSM, Carl Zeiss Microscopy, LLC, White Plains, NY). The cells were observed using a Zeiss Axio Observer inverted microscope with × 63 1.4 NA oil immersion plan apochromatic objective. Differential interference contrast (DIC) and confocal fluorescence images were acquired simultaneously. A photomultiplier tube captured the light emitted from a 488-nm argon laser with a 3.7-m pin hole passing through an MBS 488 filter with limits set between 472 and 562 nm for detection of SITO 9 stains for green fluorescence of live cells and between 597 and 669 nm for detection of propidium iodide stains for red fluorescence of bacterial cells with damaged membrane. Zeiss Zen Black software was used to obtain the images with 1024 × 1024-pixel resolution. B. cereus testing in powder-type cosmetic products Each of the 8 powder-types cosmetic products (GC, C-1, C-2, C-3, C-4, O-1, O-2 and RP) was aseptically homogenized, and 1 g was added to 1 ml of Tween 80, before mixing with 8 ml of MLB broth in presence of 10 sterile glass beads to ameliorate homogenization. These samples were diluted, and the 10 −1 –10 −2 dilutions were spread on MLA and BACARA plates, before enrichment at 30 °C for 24 h. Spread plating was used instead of spiral plating, to prevent powder particles from clogging the stylus of the spiral plater. All these enriched samples were streaked onto BACARA plates to screen for B. cereus presence, according to BAM Chapter 23 and processed simultaneously to extract bacterial DNA for B. cereus qPCR analysis, in duplicate. This will permit cross-method comparisons. Statistical analysis Data were analyzed using a linear mixed effects model and p -values were adjusted to account for multiple comparisons. Cutoff points were established using a logistic regression model, such that the cutoff Ct value gives a > 50% chance of a sample being negative. Inclusivity and exclusivity cutoff values were established using the upper bound of the 95% confidence interval of the 95th quantile estimate . The linear mixed effects model was fit using the nlme package version 3.1–145 . Quantile regression was done using the quantreg package version 5.55. All other analyses were done using R version 3.6.0 .
We isolated DNA from overnight cultures of each pure bacterial strain using the MagMAX™ Express 96 Magnetic Particle Processor (ThermoFisher) with PrepSEQ Nucleic Acid Extraction Kit for Food and Environment (ThermoFisher, P/N 4,428,176) using protocol PrepSeq_ResDNA_20011 (Life Technologies, Carlsbad, CA). Total DNA extracted from these pure cultures were used to evaluate the inclusivity (n = 143) and exclusivity (n = 69) of the qPCR assays. To determine the limit of detection of our qPCR assays, we used DNA extracted from the B. cereus 3A strain, which was grown overnight in NB and adjusted to a density of 0.5 ± 0.05 MacFarland (McF; ~ 6.7 log CFU/ml).
Two different extraction procedures were used to perform our qPCR assays on the sets of contaminated cosmetics, due to differences in the matrices. The MagMAX™ and PrepSeq kits were only appropriate for liquid cosmetics. Effective DNA extraction from the powder products required using the DNeasy® PowerPlant® Pro kit (Qiagen, Catalog Number 69204) instead, because powder products become thick and dense in the presence of the proteinase/proteinase DNA extraction buffer used in the MagMAX and PrepSeq kits .
Table shows the primers we used and the associated publications documenting their first use. These were purchased from Life Technologies (ThermoFisher). The 16S rRNA sequences (P/N 4,331,348, Life Technologies) used here had previously been determined by De Clerck, et al.; their team targeted the 5’ hypervariable fragment, which can be amplified using a universal forward and a B. cereus -specific reverse primer . Our TaqMan probe was labeled with 6 -carboxyfluorescein (FAM) reporter dye at the 5’ end and labeled at the 3’-end with a Black Hole quencher, BHQ1, to reduce background fluorescence . We ordered the PLC primer probe (CCU001SNR, Life Technologies) labeled with a JUN reporter dye at the 5’-end and a non-fluorescent quencher at the 3’-end, to facilitate multiplex qPCR using 4 dyes. QSY PLC- JUN probe is a custom probe, compatible with the TaqMan Multiplex Master Mix (P/N 446,188; Applied Biosystems, Life Technologies) and ordered through Life Technologies. The TaqMan Internal Positive Control (IPC) from Life Technologies (P/N 4,308,323) was included to monitor the PCR progress and ensure that a negative result is not caused by failed PCR in the sample and used with the TaqMan Multiplex Master Mix to amplify both the single target and the multiple target reactions.
PMAxx -PCR is an innovative technology that allows differentiation between live and dead microorganisms, based on the loss of cell membrane integrity in dead cells . This system uses a DNA-intercalating dye, PMAxx, that disrupts DNA transcription only in dead cells, as their damaged cell membranes permit entry of the dye. After photoactivation with a defined wavelength, PMAxx intercalates and binds covalently to DNA. Subsequent amplification of that modified DNA is inhibited, thereby reducing the amplification signal from dead/damaged cells in comparison to that from live cells. For each qPCR reaction, a 20 μl volume consisting of 10 μl Multiplex Master Mix, 2 μl 10X Exo IPC Mix, 0.4 μl 50X EXO IPC DNA, 1 μl of primer assay, 2 μl of sample/template DNA, and 4.6 μl of sterile deionized water was placed into a 96 well fast plate. Both primers for the multiplex PCR were combined to reach a final concentration representing 5% of the total reaction volume. These qPCR runs were performed on a 7500 Fast qPCR System (Applied Biosystems) under the following conditions: 2 min at 50 °C, then 10 min at 95 °C, followed by 40 cycles of 15 s at 95 °C and 1 min at 60 °C.
3A testing in artificially contaminated liquid products Here we used 3 different cosmetic toner products, which we artificially contaminated with B. cereus 3A. For each trial (N = 3), 3 sterile Wheaton bottles (125 ml) were filled with 30 ml of facial toner and were repeatedly spiked with 300 μl of microbial suspension, containing either a High (~ 4 or 3 log CFU/ml) or Low (~ 3 or 2 log CFU/ml) level of B. cereus 3A vegetative cells (optical density of 0.5 ± 0.05 McF), over three consecutive days, and the 3rd bottle of sample remained un-spiked for the negative controls. These repeated inoculations were necessary to enable recovery of B. cereus from products containing preservatives . To mimic a product contamination event more closely, these inoculated toner samples, along with the un-spiked negative control, were aged for 14 days at room temperature. After 14 days of aging, the un-inoculated and inoculated samples from the High- and Low-level inoculation were gently separately mixed, then 1 ml aliquots were diluted with 9 ml MLB (Difco). Our first goal was to achieve fractional results, which in this case would consist of 5 replicates at the high level of inoculation all yielding positive results, while 20 replicates at the low level of inoculation would yield only 50 ± 25% positive results, and 5 replicates of un-inoculated sample all yielding negative results . Based on this, 5 replicates of toner from the High level, 20 replicates from the Low level, and 5 replicates from the uninoculated samples (negative controls) were pre-enriched for 24 h at 30 °C in MLB, streaked onto BACARA plates and processed simultaneously for molecular analysis. Two equal sets of test portions were taken at the same time in all the replicates and placed into vials, and the first set of test portions proceeded to genomic DNAs extraction without any treatment. From the second set, 1 ml of each test portion was supplemented with 25 µM PMAxx, mixed through tap-spin, and allowed to incubate at room temperature for 10 min in the dark before exposure on the PMA-Lite (Biotium) for 15 min to crosslink free DNA . Immediately following PMAxx treatment, genomic DNAs were extracted, and all genomics (non-treated and treated) were analyzed independently in duplicate using singleplex and multiplex qPCR assays. Figure underlines an overview of the workflow of this portion of our study.
B. cereus cells aged for 14 days in toner-samples using BacLight kit To provide a microscopic perspective on the ratio of live/dead cells in the inoculated toner samples, 1 ml from the 14-day aged samples inoculated with B. cereus at the high level was centrifuged at 10,000 × g for 10 min. The supernatant was removed, and the pellet was resuspended in 1 ml of 0.85% sodium chloride (NaCl) and kept at room temperature for 1 h, mixing every 15 min. After centrifugation at 10 000 × g for 10 min, the cells were resuspended in NaCl and stained with the Live/Dead™ Bacterial Viability kit (Cat: L7007, Invitrogen by ThermoFisher Scientific) according to manufacturer’s protocols . Stained cells were viewed under a Zeiss 880 Laser Scanning Microscope (LSM, Carl Zeiss Microscopy, LLC, White Plains, NY). The cells were observed using a Zeiss Axio Observer inverted microscope with × 63 1.4 NA oil immersion plan apochromatic objective. Differential interference contrast (DIC) and confocal fluorescence images were acquired simultaneously. A photomultiplier tube captured the light emitted from a 488-nm argon laser with a 3.7-m pin hole passing through an MBS 488 filter with limits set between 472 and 562 nm for detection of SITO 9 stains for green fluorescence of live cells and between 597 and 669 nm for detection of propidium iodide stains for red fluorescence of bacterial cells with damaged membrane. Zeiss Zen Black software was used to obtain the images with 1024 × 1024-pixel resolution.
testing in powder-type cosmetic products Each of the 8 powder-types cosmetic products (GC, C-1, C-2, C-3, C-4, O-1, O-2 and RP) was aseptically homogenized, and 1 g was added to 1 ml of Tween 80, before mixing with 8 ml of MLB broth in presence of 10 sterile glass beads to ameliorate homogenization. These samples were diluted, and the 10 −1 –10 −2 dilutions were spread on MLA and BACARA plates, before enrichment at 30 °C for 24 h. Spread plating was used instead of spiral plating, to prevent powder particles from clogging the stylus of the spiral plater. All these enriched samples were streaked onto BACARA plates to screen for B. cereus presence, according to BAM Chapter 23 and processed simultaneously to extract bacterial DNA for B. cereus qPCR analysis, in duplicate. This will permit cross-method comparisons.
Data were analyzed using a linear mixed effects model and p -values were adjusted to account for multiple comparisons. Cutoff points were established using a logistic regression model, such that the cutoff Ct value gives a > 50% chance of a sample being negative. Inclusivity and exclusivity cutoff values were established using the upper bound of the 95% confidence interval of the 95th quantile estimate . The linear mixed effects model was fit using the nlme package version 3.1–145 . Quantile regression was done using the quantreg package version 5.55. All other analyses were done using R version 3.6.0 .
This research does not involve human, animal, seed or plant samples.
Assessing sensitivity of the qPCR using B. cereus 3A The limits of detection (LOD) for our qPCR assays, in singleplex and multiplex formats, were determined in two independent runs with tenfold serial dilutions (d0–d6) from purified genomic DNA of B. cereus 3A at, a density of 0.5 McF equivalents, to ~ 6 log CFU/ml. As shown in Table , the LOD for the singleplex and multiplex assays were found to be ~ 1 log CFU/ml (C T : 36.61 and 36.23) for 16S rRNA detection and 3 log CFU/ml (C T :36.32and 37.72) for PLC detection. A multiplex multicomponent of the tenfold serial dilution (n = 1) is represented in Fig. . Figure shows the multiplex amplification of 16S rRNA and PLC for the tenfold dilution series of B. cereus 3A. Inclusivity and exclusivity of the qPCR assays Inclusivity and exclusivity results are summarized in Table found in supplemental materials, along with the origins for each strain. Inclusivity was determined using 24 h actively grown pure cultures of B. cereus vegetative strains and closely related species belonging to the B. cereus group (N = 143) (Table in supplemental material). Among these 143 strains, inclusivity for the 16S rRNA was 100% for with a C T cutoff value of 23.8–23.9 and for PLC it was 97.9% (140/143) with a C T cutoff value of 26.6–28.2 for both the singleplex and the multiplex qPCR. All the strains belonging to B. cereus group were confirmed by growth on BACARA plates, where these exhibited the typical morphology of orange colonies surrounded by a white halo. Exclusivity was determined using 69 strains: 38 non- B. cereus and 31 non- Bacillus species. Exclusivity for 16S rRNA was 100% for both non- cereus group and non- Bacillus species. However, the exclusivity tests for PLC showed 98.6% (68/69) for non- cereus group (37/38) and 100% for non- Bacillus species (31/31 ). Detection of B. cereus 3A in artificially contaminated liquid-type cosmetics, using culture and molecular methods, in presence or absence of PMAxx -dye. After the pre-enrichment step, low-level inoculation (~ 2 Log CFU/ml) of B. cereus 3A yielded fractional positive results (75% of the test portions were positive) while 100% samples from the high-level inoculation (~ 3 Log CFU/ml) were positive. Observations of the bacterial population in samples inoculated at the High level, aged for 14 days, then stained with the Live/Dead™ Bacterial Viability kit, showed a mixture of live and dead cells, which had likely been killed by the preservatives in the cosmetic (Fig. ). Cells were not detectable in the Low level of inoculation. The negative control, consisting of uninoculated samples, showed neither growth nor DNA amplifications. qPCR C T values of those samples were undetermined. The C T results for16S rRNA and PLC obtained from non-treated and PMAxx -treated samples in singleplex were not significantly ( P > 0.05) different from the C T results in multiplex qPCR. In addition, those results agreed with the results on BACARA plates (Cohen’s Kappa κ = 0.99; Tables , ). As expected, the C T values for non-treated samples were significantly different ( P < 0.05) from those for the PMAxx -treated samples. Non-PMAxx treated test portions from samples that yielded colonies when plated on BACARA (positive samples according to the culture method), yielded mean 16S rRNA 13.97 ± 0.25 in singleplex and 14.95 ± 0.25 in multiplex, and C T values for PLC of 19.99 ± 0.25 in singleplex and 19.50 ± 0.25 in multiplex. Table shows the delta differences across the two conditions, i.e., the mean C T value of the PMAxx -treated samples minus the mean C T value of the non-treated samples. The untreated samples show C T values delta C T value differences of 2.96 and 2.36 cycles for 16S rRNA and PLC, respectively, which are significantly lower ( P < 0.0001) than the C T values of the culture-positive samples that were treated with PMAxx (Table ). Likewise, the mean C T values obtained for the non-treated test portions which did not yield colonies on BACARA plates (culture-negative samples) were significantly lower than the mean C T values of the PMAxx -treated samples with a delta difference of 7.82 and 7.22 for 16S RNA and PLC, respectively, for singleplex and multiplex (Table ). The C T- shifts of the positive samples (live cells) treated with PMAxx ranged between 2–3 C T values: this is due to the presence of a portion of dead cells. On the other hand, the C T- shifts of the negative samples (dead cells) were more prominent ranging between 7 and 8 C T values making it clear to distinguish between dead and live cells. PMAxx treatment does preferentially improve the detection of live cells. Results of B. cereus testing in powdered cosmetics using culture-based and molecular methods After completing both the culture-based method and qPCR analyses of the 8 powder-types cosmetic products, we compared the results acquired by each. Figure shows the culture-based method results. We found the MLA plates spread with the powder-type samples, which had been previously confirmed as contaminated, grew a heterogeneous population of bacteria from 6 of 8 samples, and 5 of those 6 samples tested positive for the presence of B. cereus on BACARA plates. Interestingly, sample C3 only revealed B. cereus on BACARA after enrichment, while samples O1 and O2 showed no growth of B. cereus (Fig. ). Our qPCR assays of the 6 samples which had tested positive according to the culture-based method did show amplification for both 16S rRNA and PLC B. cereus targets, giving C T values between 14.96 and 19.46 for the singleplex and 20.29–21.72 for the multiplex assay (Table ). Surprisingly, although Sample O-1 had tested negative according to the culture-based method, our qPCR assays showed amplification of the 16S rRNA (C T value > 33) and PLC (C T value > 37) gene targets in one of our two replicates. However, as these replicates had been pre-enriched and still showed no growth of B. cereus when streaked on the BACARA plates, we classified the result as negative.
B. cereus 3A The limits of detection (LOD) for our qPCR assays, in singleplex and multiplex formats, were determined in two independent runs with tenfold serial dilutions (d0–d6) from purified genomic DNA of B. cereus 3A at, a density of 0.5 McF equivalents, to ~ 6 log CFU/ml. As shown in Table , the LOD for the singleplex and multiplex assays were found to be ~ 1 log CFU/ml (C T : 36.61 and 36.23) for 16S rRNA detection and 3 log CFU/ml (C T :36.32and 37.72) for PLC detection. A multiplex multicomponent of the tenfold serial dilution (n = 1) is represented in Fig. . Figure shows the multiplex amplification of 16S rRNA and PLC for the tenfold dilution series of B. cereus 3A.
Inclusivity and exclusivity results are summarized in Table found in supplemental materials, along with the origins for each strain. Inclusivity was determined using 24 h actively grown pure cultures of B. cereus vegetative strains and closely related species belonging to the B. cereus group (N = 143) (Table in supplemental material). Among these 143 strains, inclusivity for the 16S rRNA was 100% for with a C T cutoff value of 23.8–23.9 and for PLC it was 97.9% (140/143) with a C T cutoff value of 26.6–28.2 for both the singleplex and the multiplex qPCR. All the strains belonging to B. cereus group were confirmed by growth on BACARA plates, where these exhibited the typical morphology of orange colonies surrounded by a white halo. Exclusivity was determined using 69 strains: 38 non- B. cereus and 31 non- Bacillus species. Exclusivity for 16S rRNA was 100% for both non- cereus group and non- Bacillus species. However, the exclusivity tests for PLC showed 98.6% (68/69) for non- cereus group (37/38) and 100% for non- Bacillus species (31/31 ). Detection of B. cereus 3A in artificially contaminated liquid-type cosmetics, using culture and molecular methods, in presence or absence of PMAxx -dye. After the pre-enrichment step, low-level inoculation (~ 2 Log CFU/ml) of B. cereus 3A yielded fractional positive results (75% of the test portions were positive) while 100% samples from the high-level inoculation (~ 3 Log CFU/ml) were positive. Observations of the bacterial population in samples inoculated at the High level, aged for 14 days, then stained with the Live/Dead™ Bacterial Viability kit, showed a mixture of live and dead cells, which had likely been killed by the preservatives in the cosmetic (Fig. ). Cells were not detectable in the Low level of inoculation. The negative control, consisting of uninoculated samples, showed neither growth nor DNA amplifications. qPCR C T values of those samples were undetermined. The C T results for16S rRNA and PLC obtained from non-treated and PMAxx -treated samples in singleplex were not significantly ( P > 0.05) different from the C T results in multiplex qPCR. In addition, those results agreed with the results on BACARA plates (Cohen’s Kappa κ = 0.99; Tables , ). As expected, the C T values for non-treated samples were significantly different ( P < 0.05) from those for the PMAxx -treated samples. Non-PMAxx treated test portions from samples that yielded colonies when plated on BACARA (positive samples according to the culture method), yielded mean 16S rRNA 13.97 ± 0.25 in singleplex and 14.95 ± 0.25 in multiplex, and C T values for PLC of 19.99 ± 0.25 in singleplex and 19.50 ± 0.25 in multiplex. Table shows the delta differences across the two conditions, i.e., the mean C T value of the PMAxx -treated samples minus the mean C T value of the non-treated samples. The untreated samples show C T values delta C T value differences of 2.96 and 2.36 cycles for 16S rRNA and PLC, respectively, which are significantly lower ( P < 0.0001) than the C T values of the culture-positive samples that were treated with PMAxx (Table ). Likewise, the mean C T values obtained for the non-treated test portions which did not yield colonies on BACARA plates (culture-negative samples) were significantly lower than the mean C T values of the PMAxx -treated samples with a delta difference of 7.82 and 7.22 for 16S RNA and PLC, respectively, for singleplex and multiplex (Table ). The C T- shifts of the positive samples (live cells) treated with PMAxx ranged between 2–3 C T values: this is due to the presence of a portion of dead cells. On the other hand, the C T- shifts of the negative samples (dead cells) were more prominent ranging between 7 and 8 C T values making it clear to distinguish between dead and live cells. PMAxx treatment does preferentially improve the detection of live cells.
B. cereus testing in powdered cosmetics using culture-based and molecular methods After completing both the culture-based method and qPCR analyses of the 8 powder-types cosmetic products, we compared the results acquired by each. Figure shows the culture-based method results. We found the MLA plates spread with the powder-type samples, which had been previously confirmed as contaminated, grew a heterogeneous population of bacteria from 6 of 8 samples, and 5 of those 6 samples tested positive for the presence of B. cereus on BACARA plates. Interestingly, sample C3 only revealed B. cereus on BACARA after enrichment, while samples O1 and O2 showed no growth of B. cereus (Fig. ). Our qPCR assays of the 6 samples which had tested positive according to the culture-based method did show amplification for both 16S rRNA and PLC B. cereus targets, giving C T values between 14.96 and 19.46 for the singleplex and 20.29–21.72 for the multiplex assay (Table ). Surprisingly, although Sample O-1 had tested negative according to the culture-based method, our qPCR assays showed amplification of the 16S rRNA (C T value > 33) and PLC (C T value > 37) gene targets in one of our two replicates. However, as these replicates had been pre-enriched and still showed no growth of B. cereus when streaked on the BACARA plates, we classified the result as negative.
Current microbiological testing standards for cosmetics in the United States are based on the detection of CFUs, which have not been easily represented by the results of molecular methods such as qPCR. Here we have assessed the utility of qPCR assays developed for B. cereus targeting 16S rRNA and PLC targets, in singleplex and multiplex, in pure culture and in cosmetic products. Overall, our qPCR detection assays gave results consistent with those from culture-based methods. Further, while both singleplex and multiplex assays were successfully applied to naturally and artificially contaminated cosmetic products, using the intercalating PMAxx treatment improved the detection of viable B. cereus cells, as opposed to simply detecting targets regardless of cell viability. This advance helps bridge the gap between cultural and molecular detection of pathogens in cosmetics. To the best of our knowledge, this is the first time that PMAxx has been used as part of a cosmetic microbiological method, and it significantly improved the detection of positive samples. PMAxx dyes efficiently shifted the C T values obtained for 16S rRNA and PLC to higher values, resulting in delta differences of 2.96 and 2.36, respectively, for positive samples, and 7.82 and 7.22, respectively, for negative samples, which is considered as successful . In general, delta C T values were significantly different for the positive and negative samples, and moreover, the effects of PMAxx were strong among the negative samples. C T values of the negative samples were completely shifted towards the end of the cycle or undetermined indicating that PMAxx inhibited the amplification of the DNA from those cells killed by resident preservatives. This was confirmed by the microscopic imaging of the live/dead stained cells from the 14-day aged samples at the High-level of inoculation, which showed a mixture of dead (red) and viable (green) B. cereus cells in those samples. Similar observations have been made in analyses of PMA-treated UHT milks and in complex microflora . Therefore, our qPCR assays, individually or combined, might be used for rapid evaluation of cosmetic raw materials and rapid testing of finished products for the presence of B. cereus and will thus reduce the likelihood that products from being contaminated with B. cereus reach consumers. Importantly, by using PMAxx with qPCR viable cells can be preferentially detected, removing one of the obstacles to using qPCR assays as rapid tests to support traditional bacterial culture identification in the microbiological safety assessment of cosmetics. The speed of these molecular methods could allow the detection of presumably positive samples, 24 h before the culture-based results.
Supplementary Information.
|
Diversity in Clinical Pharmacology: A Call to Action | 9dc48fd4-eba7-4066-9407-bb2e7faf5804 | 10024964 | Pharmacology[mh] | As a multidisciplinary scientific and medical field dedicated to the discovery, development, and optimal utilization of therapeutics, the field of clinical pharmacology is well positioned to further its contributions to the understanding of drug responses for all patient populations. Although analytical tools and technologies are readily available to explore the intrinsic and extrinsic factors that impact pharmacokinetic (PK)/drug responses across patient populations, the data generated in most clinical genomics programs and clinical trials are inadequate. Despite broad agreement that greater diversity is required to elucidate the influence of the human genome on PK/drug responses, clinical genomic studies are conducted in populations of predominantly European ancestry. The underrepresentation of diverse populations within genomic studies not only limits our ability to understand all drivers of variability in PK/drug responses but also results in missed opportunities to identify new associations with population-enriched variants, pinpoint causal variants for functional follow-up, improve genetic risk prediction accuracy for all populations, and characterize shared vs. unique genetic and environmental risk factors that can influence health outcomes. Randomized, controlled trials are the gold standard for establishing the efficacy and safety of pharmacotherapeutic agents, and both novel and established therapies rely on clinical trials to inform new indications, guideline incorporation, and routine clinical practice. Because many characteristics of therapies (e.g., efficacy, safety, dose–response, and absorption, distribution, metabolism, and elimination) vary substantially across individuals, the lack of diversity in clinical trials is challenging. Precision medicine attempts to address this shortcoming by applying the concept that intrinsic factors (e.g., genetic variants) may explain the deviation of an individual’s drug response from the population-wide mean effect. Because precision medicine relies on clinical trials with diverse representation to capture the full range of potential responses to treatment, the exclusion of specific groups from study populations prevents precision medicine from reaching its full potential. Consequently, we cannot precisely characterize and predict how therapies that are proven efficacious in well-represented groups may affect underrepresented groups. Thus, for scientific, medical, and social justice reasons, increasing clinical trial diversity is necessary to work toward improved outcomes for all individuals and populations. Due to longstanding structural barriers in clinical research and clinical trials, the enrollment of women, older adults, individuals from diverse racial and ethnic groups, members of the LGBTQIA+ community, and those from rural communities is disproportionately low. For example, although it has been established that population disparities are present in cholesterol management, a large portion (41%) of the randomized, controlled trials cited by the 2018 Guideline on the Management of Blood Cholesterol did not report race/ethnicity data. Among the trials that did report such data, Black and Hispanic trial participation rates were only 41% and 58%, respectively, of the expected level based on 2019 US Census representation. Moreover, the 2020 Drug Trials Snapshot from the US Food and Drug Administration (FDA), which summarizes clinical trial diversity metrics across the therapies approved in 2021, found that while sponsors were able to enroll patients from diverse racial/ethnic backgrounds, there were instances in which representation remained low. The multifactorial and complex reasons for the lack of diversity in clinical trials must be thoroughly understood before they can be addressed. Cultural differences impact the willingness of diverse communities to participate in trials, with generational trauma, fear of exploitation, and socioeconomic constraints contributing to the reluctance to enroll in clinical research. Instances of research misconduct in underrepresented populations have also contributed to sentiments of distrust among various communities. , Moreover, industry norms that relate to biospecimen collection, data sharing, and return of participant results may be opposed by (or forbidden within) certain populations. To overcome these barriers, respect for cultural differences and increased research in underrepresented populations is required, and it is our shared responsibility to create partnerships among patient stakeholder groups, to increase clinical trial visibility, and to convey the benefits of clinical trial participation. Recently, regulatory bodies including those in the US federal government have implemented initiatives aimed at increasing participant diversity in clinical trials. In this regard, offices that focus on improving minority health and women’s health have been established within the Department of Health and Human Services and sub-agencies such as the National Institutes of Health (NIH) and FDA, with one example of progress represented by the recent FDA draft guidance, Diversity Plans to Improve Enrollment of Participants from Underrepresented Racial and Ethnic Populations in Clinical Trials Guidance for Industry . This guidance encourages sponsors to develop and submit participation plans with strategies to recruit underrepresented racial and ethnic populations early in the drug development process. Collectively, these initiatives acknowledge the disproportionate disease burden often faced by underrepresented racial and ethnic patients and the lack of access to health care within their communities. They also provide sponsors with a clear baseline framework for diversity, equity, and inclusion (DEI) planning, which is a critical first step toward developing industry-wide standards for clinical studies. It is imperative to consider DEI when it comes to the current and future clinical pharmacology workforce. There is ample evidence to demonstrate the value and benefits of a diverse workforce, ranging from better business performance to improved and bolder decision making, which in part, relies on a diversity of viewpoints that arises from a variety of professional and life experiences. Indeed, the best and most creative ideas come from a diverse and inclusive space, and this is particularly relevant for scientific innovations. The extent of multicultural exposure and experience is positively correlated with creativity, problem solving, and idea generation. Meaningful conversations occur only when everyone feels empowered to speak up, ask questions, and share ideas. The impact of not aligning with social, educational, ideological, or phenotypical standards of the majority culture creates biases both externally and internally. From an early age, people are exposed to social comparisons that define how they identify within their environments. These comparisons are marked in those who differ from the majority culture and often present as internalized inadequacy. No one should feel obligated to conform to a social construct at the cost of erasing his/her/their authentic identity . Nonetheless, to “fit in” at work, some individuals feel the need to present a different, less authentic self at work, only feeling comfortable being fully themselves away from the office. While biases exist globally, when it comes to professionalism, the expectation is the same—that individuals adhere to unexamined norms, expectations, and patterns in appearance, speech, and workstyle that are formed through a lens of the majority culture. Such standards are presented as “normal,” traditional, or superior among the various ethnic, racial, and regional identities present in work environments. By openly discussing and interrogating the biased lens through which professional environments are constructed, individuals can help to build safe spaces within which to air microaggressions and address perceptions of stereotypes, with the goal of creating inclusive work environments. This can be achieved within the clinical pharmacology community by examining and articulating our own experiences of feeling included (or excluded), and by being empathetic and aware that our unique experiences shape our behaviors. In these inclusive environments, role models can speak up and advocates can motivate diverse individuals to be heard and to foster happiness, productivity, creativity, and innovation for everyone in the workplace. As key contributors to the advancement and optimal use of therapeutics, clinical pharmacologists have a vested interest and shared responsibility to ensure that future research strategies prioritize DEI. Our call to action to the clinical pharmacology community regarding scientific and clinical trial population considerations is threefold: to foster DEI in trial populations by predefining enrollment targets based on the intended patient population, effectively managing recruitment goals during trials, and increasing the diversity of colleagues associated with the oversight and conduct of clinical studies; to develop partnerships that enable site selection in areas with higher densities of underrepresented subgroups; and to ensure that genomic studies are representative of diverse populations. Our call to action regarding workforce considerations regarding workforce considerations is to be authentic, empathetic, and ever-evolving. When everyone represents themselves authentically, different perceptions and values can be shared without being hindered by others’ expectations. By being our authentic selves, we can foster the acceptance of our colleagues’ similarities and differences and embrace the values that individuals bring to our community. By cultivating a culture of empathy, we can be challenged to understand others’ viewpoints, reactions, and experiences, as seen through each other’s eyes. With active listening, genuine curiosity, and validation of feelings, we can strengthen workplace relationships, and deepen social connections while avoiding misinformed judgments and stereotypes. This recognition and flexibility to continually evolve can allow us to adapt, grow, and change—and to work toward expanding equal opportunities for all employees. In closing, we believe that we are at a critical juncture to incorporate DEI into our science, clinical trials, and work environments in innovative and intentional ways. This will lead to more inclusive research that, in turn, will ultimately lead to more equitable health care for all patients, and to a workforce that will be positioned and supported to flourish now and into the future. supplementary table 1 |
Recommendations of the Spanish brachytherapy group of the Spanish Society of Radiation Oncology and the Spanish Society of Medical Physics for interstitial high-dose-rate brachytherapy for gynaecologic malignancies | 305dafea-fb24-4672-af8f-098451531c59 | 10025210 | Internal Medicine[mh] | Brachytherapy (BT) plays a key role in the treatment of gynaecological malignancies, especially for cervical and endometrial cancer. Various BT techniques are available. However, interstitial brachytherapy (ISBT) is considered essential in the treatment of cancers of the cervix (early stage or locally-advanced), vagina, and vulva , and for the treatment of recurrent disease. In the present document, we describe the clinical indications, applicators, and physics-related aspects (dosimetry, reconstruction, and prescription) defined by an expert group of radiation oncologists and medical physicists from the Spanish Brachytherapy Group (GEB; Grupo Español de Braquiterapia ) and the Spanish Society of Medical Physics (SEFM). These recommendations are based a literature review, professional experience, and in-depth discussions held at the annual consensus meeting of the two societies, which took place on October, 22 2021 at the Catalan Institute of Oncology in Barcelona, Spain. The topic of this meeting was “Interstitial Gynaecological Brachytherapy”. Here we present our consensus-based recommendations. Prior to the meeting, a brief, nine-item electronic survey was distributed to all radiation oncology departments in Spain and Portugal ( n = 174) by the Spanish Society of Radiation Oncology (SEOR). In Spain, there are 124 radiation oncology departments; of these, 72 offer brachytherapy. Of these 72 centres, 43 are public, 17 private, eight public–private, and four are foundations . A total of 36 complete surveys were returned, a response rate of 50% (half of the centres offering brachytherapy). The survey questions and responses are shown in Table .
Cervix intracavitary/interstitial hybrid HDR- BT Image-guided brachytherapy (IGBT), in conjunction with external beam radiotherapy (EBRT) and concurrent chemotherapy, is the current standard treatment for cervical cancer. The EMBRACE I study was performed to evaluate the role of IGBT in locally-advanced cervical cancer. This large, prospective study ( n = 1341 patients) offers the highest level of evidence available at present . Importantly, the results of the study validated the GEC-ESTRO and ICRU (International Commission on Radiation Units and Measurements) recommendations . At a median follow-up of 51 months (interquartile range, 20–64), the actuarial 5-year local control rate was 92 (95% confidence interval: 90–93). EMBRACE I showed an absolute improvement of 14–17% in both local and pelvic control in patients with stage IIIB disease (FIGO) compared to previously reports . This result is similar to that achieved with ISBT. Moreover, compared to previous studies with a similar stage distribution, the 5-year overall survival (OS) rate in EMBRACE I was superior (74 vs 67%) . In cervical cancer, ISBT is indicated for stages IIB-III and IVA. An analysis of the retroEMBRACE study (a retrospective study of patients treated with IGBT based on computed tomography [CT] or magnetic resonance imaging [MRI] before initiation of the EMBRACE study) showed that the patients that benefitted the most were those with large volume, high-risk clinical target volume (HR-CTV) at the time of BT. Local control in patients with HR-CTV > 30 cc was 10% higher for ISBT than for intracavitary BT alone , without any increase in late urinary or gastrointestinal toxicity. The use of ISBT has grown substantially in recent years. For example, in the retroEMBRACE study, 23% of patients were treated with intracavitary or interstitial BT, while up to 43% of patients in EMBRACE 1 received ISBT. Cervix perineal templates/freehand implants The potential coverage allowed by intracavitary/interstitial hybrid applicators is scant if any of the following are present: Medial or distal parametrial extension (up to pelvic wall) Unresponsive bulky disease Cervical tumours with vaginal extension to the middle or lower third Bladder or rectal involvement (stage IV) Or in certain clinical situations, as follows: Cervical cancer in patients who are not suitable for an intrauterine component due to unfavourable topography Presence of poor geometric conditions: very narrow vaginas Previous history of total/subtotal hysterectomy where the gynaecological tandem cannot be used In these cases, it is recommended to add a larger interstitial component . The implant can be performed using transperineal templates or the “freehand” technique, with guided placement and planning by transrectal ultrasound (US), CT, or MRI . Only a limited number of institutions perform interstitial perineal implants, possibly due to the invasive nature of the implants, insufficient experience with the technique, and/or due to the scant literature . Nonetheless, now that MRI is available for BT planning, and considering the need to provide better coverage of locally-advanced tumours in which conventional applicators (including hybrids with an interstitial component) do not allow offer good coverage, it, it is essential that patients have access to this technique when indicated. Primary vaginal malignancies/ vaginal recurrences Primary vaginal cancer is a rare cancer, accounting for only 3% of all gynaecological malignancies. Due to its rarity, there is a notable lack of data on the optimal therapeutic management of this cancer, which represents a major challenge to improving treatment. Historically, surgery was the treatment of choice for primary vaginal cancer. However, due to the need for an extensive resection, organ sparing was not possible, leading to severe morbidity with a negative impact on quality of life . Data from the historical series at centres where surgery was the standard treatment show 5-year OS rates ranging from 47 to 74%, with a median OS for stage I and stage II disease of 82 and 53%, respectively . Given these data, there is wide agreement that surgery is the technique of choice in small tumours (< 2 cm) that are limited to the upper third of the vagina (17). Importantly, treatment with definitive EBRT or even BT alone has shown very good results in terms of local control and disease-specific survival, ranging from 83 to 100% . Since primary vaginal cancer is etiologically similar to cervical cancer, the same treatment strategies were implemented, which is why an organ-sparing approach consisting of radiochemotherapy followed by BT became the standard treatment in primary vaginal cancer. A large study based on the SEER (Surveillance, Epidemiology, and End Results) database found that the median OS in women who received BT was almost twice as long as those who received EBRT alone (6.1 vs. 3.6 years). Moreover, the addition of BT reduced the risk of death by 13%. On the multivariate analysis, BT was an independent predictor of survival . The introduction of MRI-based IGBT in the treatment of cervical cancer allowed for dose escalation, which resulted in better local control rates and reduced morbidity . In vaginal cancer, a few studies have reported 2D-radiograph-based BT outcomes, with good local control, especially in stage T1 disease . Several small, single institutional series have introduced the terms of IGBT in vaginal cancer, showing encouraging results, with 2-year local control rates ranging from 82 to 93% with less morbidity than previous studies . In recent years, the Gynaecological Working Group of the Groupe Européen de Curiethérapie and the European Society for Radiotherapy and Oncology (GYN GEC–ESTRO) introduced the terms and target concept for IGBT in vaginal cancer . Recently, that task group conducted a retrospective, multicentre study involving 148 patients with primary vaginal cancer treated with IGBT, showing a good local control rate (83%), with especially strong results in large advanced stage tumours (T3 and T4a) compared to previous reports . These results, although preliminary, support the role of IGBT in primary vaginal cancer based on the good local control rates in large tumours with less morbidity due to better sparing of organs at risk (OAR). Vulvar cancer Vulvar cancer is rare, accounting for approximately 4% of all gynaecologic malignancies worldwide . Although primary surgery (radical excision/vulvectomy with selective sentinel node biopsy and/or bilateral inguinofemoral lymphadenectomy) is the cornerstone of treatment for this cancer especially for early-stage disease , recurrence rates are high. The main factors influencing local recurrence are nodal involvement and insufficient surgical margins ; however, other factors may also contribute to increased risk of recurrence, including stromal invasion , lymphovascular and/or perineural invasion , tumour size, and the presence of associated preneoplastic lesions and human papillomavirus . In locally-advanced vulvar cancer, the standard treatment—radical surgery—is not always feasible due to the difficulty of achieving clear surgical margins without performing mutilating surgery and/or because the lymph nodes are fixed to the fascia, muscle, or vascular structures. In these cases, the treatment of choice is definitive radiotherapy with or without neoadjuvant chemotherapy . BT is indicated for the treatment of vulvar cancer in three clinical scenarios , as follows: (1) postoperative adjuvant BT for patients with early-stage disease who have unfavourable histological prognostic factors. In this case, treatment options are definitive BT or BT combined with EBRT; (2) boost BT to the primary tumour after EBRT in locally-advanced vulvar tumours not suitable for upfront surgery; (3) local recurrence after primary surgery or previous irradiation.
Image-guided brachytherapy (IGBT), in conjunction with external beam radiotherapy (EBRT) and concurrent chemotherapy, is the current standard treatment for cervical cancer. The EMBRACE I study was performed to evaluate the role of IGBT in locally-advanced cervical cancer. This large, prospective study ( n = 1341 patients) offers the highest level of evidence available at present . Importantly, the results of the study validated the GEC-ESTRO and ICRU (International Commission on Radiation Units and Measurements) recommendations . At a median follow-up of 51 months (interquartile range, 20–64), the actuarial 5-year local control rate was 92 (95% confidence interval: 90–93). EMBRACE I showed an absolute improvement of 14–17% in both local and pelvic control in patients with stage IIIB disease (FIGO) compared to previously reports . This result is similar to that achieved with ISBT. Moreover, compared to previous studies with a similar stage distribution, the 5-year overall survival (OS) rate in EMBRACE I was superior (74 vs 67%) . In cervical cancer, ISBT is indicated for stages IIB-III and IVA. An analysis of the retroEMBRACE study (a retrospective study of patients treated with IGBT based on computed tomography [CT] or magnetic resonance imaging [MRI] before initiation of the EMBRACE study) showed that the patients that benefitted the most were those with large volume, high-risk clinical target volume (HR-CTV) at the time of BT. Local control in patients with HR-CTV > 30 cc was 10% higher for ISBT than for intracavitary BT alone , without any increase in late urinary or gastrointestinal toxicity. The use of ISBT has grown substantially in recent years. For example, in the retroEMBRACE study, 23% of patients were treated with intracavitary or interstitial BT, while up to 43% of patients in EMBRACE 1 received ISBT.
The potential coverage allowed by intracavitary/interstitial hybrid applicators is scant if any of the following are present: Medial or distal parametrial extension (up to pelvic wall) Unresponsive bulky disease Cervical tumours with vaginal extension to the middle or lower third Bladder or rectal involvement (stage IV) Or in certain clinical situations, as follows: Cervical cancer in patients who are not suitable for an intrauterine component due to unfavourable topography Presence of poor geometric conditions: very narrow vaginas Previous history of total/subtotal hysterectomy where the gynaecological tandem cannot be used In these cases, it is recommended to add a larger interstitial component . The implant can be performed using transperineal templates or the “freehand” technique, with guided placement and planning by transrectal ultrasound (US), CT, or MRI . Only a limited number of institutions perform interstitial perineal implants, possibly due to the invasive nature of the implants, insufficient experience with the technique, and/or due to the scant literature . Nonetheless, now that MRI is available for BT planning, and considering the need to provide better coverage of locally-advanced tumours in which conventional applicators (including hybrids with an interstitial component) do not allow offer good coverage, it, it is essential that patients have access to this technique when indicated.
Primary vaginal cancer is a rare cancer, accounting for only 3% of all gynaecological malignancies. Due to its rarity, there is a notable lack of data on the optimal therapeutic management of this cancer, which represents a major challenge to improving treatment. Historically, surgery was the treatment of choice for primary vaginal cancer. However, due to the need for an extensive resection, organ sparing was not possible, leading to severe morbidity with a negative impact on quality of life . Data from the historical series at centres where surgery was the standard treatment show 5-year OS rates ranging from 47 to 74%, with a median OS for stage I and stage II disease of 82 and 53%, respectively . Given these data, there is wide agreement that surgery is the technique of choice in small tumours (< 2 cm) that are limited to the upper third of the vagina (17). Importantly, treatment with definitive EBRT or even BT alone has shown very good results in terms of local control and disease-specific survival, ranging from 83 to 100% . Since primary vaginal cancer is etiologically similar to cervical cancer, the same treatment strategies were implemented, which is why an organ-sparing approach consisting of radiochemotherapy followed by BT became the standard treatment in primary vaginal cancer. A large study based on the SEER (Surveillance, Epidemiology, and End Results) database found that the median OS in women who received BT was almost twice as long as those who received EBRT alone (6.1 vs. 3.6 years). Moreover, the addition of BT reduced the risk of death by 13%. On the multivariate analysis, BT was an independent predictor of survival . The introduction of MRI-based IGBT in the treatment of cervical cancer allowed for dose escalation, which resulted in better local control rates and reduced morbidity . In vaginal cancer, a few studies have reported 2D-radiograph-based BT outcomes, with good local control, especially in stage T1 disease . Several small, single institutional series have introduced the terms of IGBT in vaginal cancer, showing encouraging results, with 2-year local control rates ranging from 82 to 93% with less morbidity than previous studies . In recent years, the Gynaecological Working Group of the Groupe Européen de Curiethérapie and the European Society for Radiotherapy and Oncology (GYN GEC–ESTRO) introduced the terms and target concept for IGBT in vaginal cancer . Recently, that task group conducted a retrospective, multicentre study involving 148 patients with primary vaginal cancer treated with IGBT, showing a good local control rate (83%), with especially strong results in large advanced stage tumours (T3 and T4a) compared to previous reports . These results, although preliminary, support the role of IGBT in primary vaginal cancer based on the good local control rates in large tumours with less morbidity due to better sparing of organs at risk (OAR).
Vulvar cancer is rare, accounting for approximately 4% of all gynaecologic malignancies worldwide . Although primary surgery (radical excision/vulvectomy with selective sentinel node biopsy and/or bilateral inguinofemoral lymphadenectomy) is the cornerstone of treatment for this cancer especially for early-stage disease , recurrence rates are high. The main factors influencing local recurrence are nodal involvement and insufficient surgical margins ; however, other factors may also contribute to increased risk of recurrence, including stromal invasion , lymphovascular and/or perineural invasion , tumour size, and the presence of associated preneoplastic lesions and human papillomavirus . In locally-advanced vulvar cancer, the standard treatment—radical surgery—is not always feasible due to the difficulty of achieving clear surgical margins without performing mutilating surgery and/or because the lymph nodes are fixed to the fascia, muscle, or vascular structures. In these cases, the treatment of choice is definitive radiotherapy with or without neoadjuvant chemotherapy . BT is indicated for the treatment of vulvar cancer in three clinical scenarios , as follows: (1) postoperative adjuvant BT for patients with early-stage disease who have unfavourable histological prognostic factors. In this case, treatment options are definitive BT or BT combined with EBRT; (2) boost BT to the primary tumour after EBRT in locally-advanced vulvar tumours not suitable for upfront surgery; (3) local recurrence after primary surgery or previous irradiation.
Intracavitary/interstitial hybrid HDR-BT Intracavitary applicators consist of an intrauterine tube (IUT) and the vaginal component (ovoid, ring, or cylinder) to which the interstitial part can be attached in different ways (see below). The various commercially-available applicators are described below. Ovoids, intrauterine tube and interstitial needles Elekta (Stockholm, Sweden) Utrecht applicator The original Utrecht applicator is composed of two ovoids and an IUT (4 or 6 mm in diameter). In addition, up to five flexible plastic catheters measuring 294 mm in length with blunt/round tips or sharp tips can be placed in each ovoid, two in the central part to complete the effect of the IUT and three laterally to extend to the parametrial level with an output angulation of 15º and 25º. The IUT has a unique length with a cervical stop where the ovoids hook, thus allowing the applicator to be adapted to the hysterometry of each patient. Geneva applicator This applicator is superior to the older Utrecht-type applicators, especially for patients with smaller anatomies. This modular applicator has a greater size range of ovoid tubes, with a fixed diameter IUT in six different lengths and three angulations. It allows for the placement of at least five needles per ovoid, and more needles can be inserted in larger diameter ovoids. In addition, a central interstitial needle can be added to expand treatment options after hysterectomy. The different components are easily assembled due to the new clip-on system.
Intracavitary applicators consist of an intrauterine tube (IUT) and the vaginal component (ovoid, ring, or cylinder) to which the interstitial part can be attached in different ways (see below). The various commercially-available applicators are described below.
Elekta (Stockholm, Sweden) Utrecht applicator The original Utrecht applicator is composed of two ovoids and an IUT (4 or 6 mm in diameter). In addition, up to five flexible plastic catheters measuring 294 mm in length with blunt/round tips or sharp tips can be placed in each ovoid, two in the central part to complete the effect of the IUT and three laterally to extend to the parametrial level with an output angulation of 15º and 25º. The IUT has a unique length with a cervical stop where the ovoids hook, thus allowing the applicator to be adapted to the hysterometry of each patient. Geneva applicator This applicator is superior to the older Utrecht-type applicators, especially for patients with smaller anatomies. This modular applicator has a greater size range of ovoid tubes, with a fixed diameter IUT in six different lengths and three angulations. It allows for the placement of at least five needles per ovoid, and more needles can be inserted in larger diameter ovoids. In addition, a central interstitial needle can be added to expand treatment options after hysterectomy. The different components are easily assembled due to the new clip-on system.
Utrecht applicator The original Utrecht applicator is composed of two ovoids and an IUT (4 or 6 mm in diameter). In addition, up to five flexible plastic catheters measuring 294 mm in length with blunt/round tips or sharp tips can be placed in each ovoid, two in the central part to complete the effect of the IUT and three laterally to extend to the parametrial level with an output angulation of 15º and 25º. The IUT has a unique length with a cervical stop where the ovoids hook, thus allowing the applicator to be adapted to the hysterometry of each patient. Geneva applicator This applicator is superior to the older Utrecht-type applicators, especially for patients with smaller anatomies. This modular applicator has a greater size range of ovoid tubes, with a fixed diameter IUT in six different lengths and three angulations. It allows for the placement of at least five needles per ovoid, and more needles can be inserted in larger diameter ovoids. In addition, a central interstitial needle can be added to expand treatment options after hysterectomy. The different components are easily assembled due to the new clip-on system.
The interstitial component is coupled to a Fletcher-type titanium applicator with a 3 mm IUT with different lengths and a 30º angulation. The pairs of ovoids have holes that act as a guide or template for the interstitial component, allowing for insertion of sharp or blunt tip needles. The needles are 2 mm in diameter and 320 mm in length, parallel to the IUT. The ovoids allow for the placement of 4, 6 or 8 needles depending on the size of the ovoid. Ring, intrauterine tube, and interstitial needles Elekta (Stockholm, Sweden) Vienna applicator This applicator has capacity for up to seven titanium needles (26 mm ring), and up to nine needles in the larger diameter (30 mm and 34 mm) rings. Needles can be plastic or rigid metallic with an angulation of 60º to the vaginal axis. The rigid needles (1.9 mm in diameter and 240 mm in length) are pre-curved with penetration distances of 30 mm, 40 mm and 50 mm, and are also located parallel to the IUT of fixed lengths. Vienna II applicator The Vienna II ring applicator was developed to treat patients with residual distal parametrial disease that cannot be adequately covered by hybrid ovoid or ring applicators . The Vienna II is the same ring applicator as Vienna, but with an additional piece (a hood that is covered to the vaginal ring). This extra piece serves to help guide interstitial needles with an oblique direction that is 20º relative to the IUT. Venezia applicator This applicator consists of an IUT, two interstitial semicircular tubes forming an easy-to-assemble ring, two vaginal capsules that adhere to the corresponding semicircular tube, and a perineal template through which the needles are inserted in parallel from the perineum (freehand) to the parametrium. The perineal bar is a tool used to fix the applicator to the patient. This hybrid applicator allows users to insert interstitial needles in parallel or obliquely (12º) to the intrauterine tube depending on the patient’s anatomy. This applicator has the capacity to insert up to 134 6F (2 mm) plastic needles (length: 294 mm), both in parallel and divergently angled to the IUT. Each crescent can accommodate up to six needles, and 122 in the perineal insole. This design offers an important advantage for dose distribution in BT in complex volumes due to the large number of channels, their separation and their orientation, allowing for the delivery of an optimised dosimetric coverage in terms of compliance with the HR-CTV, which can reduce doses to the OARs.
Elekta (Stockholm, Sweden) Vienna applicator This applicator has capacity for up to seven titanium needles (26 mm ring), and up to nine needles in the larger diameter (30 mm and 34 mm) rings. Needles can be plastic or rigid metallic with an angulation of 60º to the vaginal axis. The rigid needles (1.9 mm in diameter and 240 mm in length) are pre-curved with penetration distances of 30 mm, 40 mm and 50 mm, and are also located parallel to the IUT of fixed lengths. Vienna II applicator The Vienna II ring applicator was developed to treat patients with residual distal parametrial disease that cannot be adequately covered by hybrid ovoid or ring applicators . The Vienna II is the same ring applicator as Vienna, but with an additional piece (a hood that is covered to the vaginal ring). This extra piece serves to help guide interstitial needles with an oblique direction that is 20º relative to the IUT. Venezia applicator This applicator consists of an IUT, two interstitial semicircular tubes forming an easy-to-assemble ring, two vaginal capsules that adhere to the corresponding semicircular tube, and a perineal template through which the needles are inserted in parallel from the perineum (freehand) to the parametrium. The perineal bar is a tool used to fix the applicator to the patient. This hybrid applicator allows users to insert interstitial needles in parallel or obliquely (12º) to the intrauterine tube depending on the patient’s anatomy. This applicator has the capacity to insert up to 134 6F (2 mm) plastic needles (length: 294 mm), both in parallel and divergently angled to the IUT. Each crescent can accommodate up to six needles, and 122 in the perineal insole. This design offers an important advantage for dose distribution in BT in complex volumes due to the large number of channels, their separation and their orientation, allowing for the delivery of an optimised dosimetric coverage in terms of compliance with the HR-CTV, which can reduce doses to the OARs.
Vienna applicator This applicator has capacity for up to seven titanium needles (26 mm ring), and up to nine needles in the larger diameter (30 mm and 34 mm) rings. Needles can be plastic or rigid metallic with an angulation of 60º to the vaginal axis. The rigid needles (1.9 mm in diameter and 240 mm in length) are pre-curved with penetration distances of 30 mm, 40 mm and 50 mm, and are also located parallel to the IUT of fixed lengths. Vienna II applicator The Vienna II ring applicator was developed to treat patients with residual distal parametrial disease that cannot be adequately covered by hybrid ovoid or ring applicators . The Vienna II is the same ring applicator as Vienna, but with an additional piece (a hood that is covered to the vaginal ring). This extra piece serves to help guide interstitial needles with an oblique direction that is 20º relative to the IUT. Venezia applicator This applicator consists of an IUT, two interstitial semicircular tubes forming an easy-to-assemble ring, two vaginal capsules that adhere to the corresponding semicircular tube, and a perineal template through which the needles are inserted in parallel from the perineum (freehand) to the parametrium. The perineal bar is a tool used to fix the applicator to the patient. This hybrid applicator allows users to insert interstitial needles in parallel or obliquely (12º) to the intrauterine tube depending on the patient’s anatomy. This applicator has the capacity to insert up to 134 6F (2 mm) plastic needles (length: 294 mm), both in parallel and divergently angled to the IUT. Each crescent can accommodate up to six needles, and 122 in the perineal insole. This design offers an important advantage for dose distribution in BT in complex volumes due to the large number of channels, their separation and their orientation, allowing for the delivery of an optimised dosimetric coverage in terms of compliance with the HR-CTV, which can reduce doses to the OARs.
This applicator consists of intrauterine tubes of different lengths with a ring angulation of 60º and 90º to the vaginal axis. The applicator incorporates a sheath that increases the mucosal source distance while also allowing for the insertion of interstitial vectors. This applicator has the capacity for 16 plastic needles, 2 mm in diameter and 320 mm in length, parallel to the IUT.
This applicator consists of three intrauterine tubes with different angulations and three different lengths. It has a thinner IUT (3.5 cm in diameter) with a length of 60 mm. The applicator can attach a 30 mm ring with a double channel, allowing for the incorporation of a rectal retractor. This applicator has the capacity for up to eight plastic needles (diameter: 1.7 mm, length: 300 mm) parallel to the IUT or with an angle of 5º. The manufacturer (Eckert & Ziegler Bebig) markets the “Tulip Applicator Family”, which allows clinicians to adapt the endocavitary applicators to a guide mold, which allows for the guided placement of plastic needles in different central and lateral positions at various different angles. This system can be used with HDR afterloaders from other companies. The capsules are individual consumables and cannot be used for more than 24 h. Perineal templates The classic manufactured applicators, which were designed for CT scans, are being modified for use with MRI. There are perineal templates for use with rigid needles (aluminium or titanium) that minimise implant deviation, which is important considering the long path that the needles must travel from the guide in the perineum (minimum: 10 cm). This long distance can be an inconvenience with plastic catheters, especially if the objective is to add the necessary obliquity to ensure coverage of distal parametrial disease. CT compatible applicators Martinez Universal Perineal Interstitial Template (MUPIT) This template is designed for gynaecological, prostate, anorectal, and perineal implants. The MUPIT consists of a double perineal template, a vaginal and rectal cylinder, and hollow 17-gauge guides separated by 6–10 mm. It has a total of 111 holes, which permit straight or angled vectors in horizontal planes, perpendicular to the plane of the perineal template. These are angled distally 14º, achieving adequate implantation of the parametria and greater lateral coverage of the CTV. The metal needles vary in length (range, 14–20 cm) and are held with stoppers to prevent cranio-caudal displacement during the procedure. A second template is used to strengthen vector fixation. The disadvantages include the lack of an intrauterine tube and the need for CT-based planning. Syed-Neblett applicator The modified Syed-Neblett device (Alfa Omega services, Bellflower CA) is based on the same principle as the original applicator and is only compatible with CT. It consists of a perineal template and a vaginal obturator with hollow 17-gauge guides of various lengths that allow users to combine interstitial sources and a vaginal cylinder with or without an IUT. MRI-compatible applicators Venezia Applicator See description above. MAC applicator The MAC (acronym for the “Mick-Alektiar-Cohen” collaboration) applicator from Eckert & Ziegler Bebig (Germany), consists of a cylinder + IUT, is compatible with MRI and can be used with plastic needles of various diameters up to 2 mm. It has 36 concentric channels for the placement of unilateral or bilateral needles in parametria and can be used with a 0º or 30º IUT. Kelowna template Varian Medical Systems has developed the Kelowna gynaecological templates for interstitial implants, with a universal cylinder as a vaginal obturator. A 25 mm central hole in the template allows for the insertion of an ultrasound probe to guide the placement of needles of various sizes and lengths (113, 200, or 320 mm), whose tips can be sharp or blunt. The needles can be made of polyetheretherketone (PEEK), steel, or titanium. Several teams with extensive experience in perineal ISBT have developed their own (i.e., non-commercial) MRI-compatible applicators that include an IUT. These applicators use rigid titanium or plastic, straight or angled needles, and can provide distal parametrial coverage through the adaptation of commercially-available IUTs. In Spain, two well-known prototypes are the Benidorm and Pamplona templates (Fig. ). 3D printing for the individualised treatment of cervical carcinoma In recent years, notable advances have taken place in terms of the development of three-dimensional (3D) printers for medical supplies. These printers allow clinicians to design patient-specific applicators adapted to the patient's individual anatomy and tumour response. This allows for individualised solutions not available with standard commercial applicators. It important to note that these applicators must be produced in a healthcare environment to ensure their suitability for clinical use in patients. The Navarra Hospital Complex developed a 3D printed endocavitary-interstitial applicator to treat the entire length of the vagina irregularly and to insert parametrial needles. This applicator is shown below in Fig. . Interstitial BT without perineal template (“freehand”) Interstitial implants can also be performed freehand (i.e., without the use of perineal templates) guided by vaginal and rectal touch, ultrasound or laparoscopy. The disadvantage of freehand needle insertion is the complexity of the technique. In addition, msince the geometry is not predefined, it is very user-dependent, which could lead to imprecise and unstable positioning, poor reproducibility, and potential loss of parallelism between the needles.
The classic manufactured applicators, which were designed for CT scans, are being modified for use with MRI. There are perineal templates for use with rigid needles (aluminium or titanium) that minimise implant deviation, which is important considering the long path that the needles must travel from the guide in the perineum (minimum: 10 cm). This long distance can be an inconvenience with plastic catheters, especially if the objective is to add the necessary obliquity to ensure coverage of distal parametrial disease. CT compatible applicators Martinez Universal Perineal Interstitial Template (MUPIT) This template is designed for gynaecological, prostate, anorectal, and perineal implants. The MUPIT consists of a double perineal template, a vaginal and rectal cylinder, and hollow 17-gauge guides separated by 6–10 mm. It has a total of 111 holes, which permit straight or angled vectors in horizontal planes, perpendicular to the plane of the perineal template. These are angled distally 14º, achieving adequate implantation of the parametria and greater lateral coverage of the CTV. The metal needles vary in length (range, 14–20 cm) and are held with stoppers to prevent cranio-caudal displacement during the procedure. A second template is used to strengthen vector fixation. The disadvantages include the lack of an intrauterine tube and the need for CT-based planning. Syed-Neblett applicator The modified Syed-Neblett device (Alfa Omega services, Bellflower CA) is based on the same principle as the original applicator and is only compatible with CT. It consists of a perineal template and a vaginal obturator with hollow 17-gauge guides of various lengths that allow users to combine interstitial sources and a vaginal cylinder with or without an IUT. MRI-compatible applicators Venezia Applicator See description above. MAC applicator The MAC (acronym for the “Mick-Alektiar-Cohen” collaboration) applicator from Eckert & Ziegler Bebig (Germany), consists of a cylinder + IUT, is compatible with MRI and can be used with plastic needles of various diameters up to 2 mm. It has 36 concentric channels for the placement of unilateral or bilateral needles in parametria and can be used with a 0º or 30º IUT. Kelowna template Varian Medical Systems has developed the Kelowna gynaecological templates for interstitial implants, with a universal cylinder as a vaginal obturator. A 25 mm central hole in the template allows for the insertion of an ultrasound probe to guide the placement of needles of various sizes and lengths (113, 200, or 320 mm), whose tips can be sharp or blunt. The needles can be made of polyetheretherketone (PEEK), steel, or titanium. Several teams with extensive experience in perineal ISBT have developed their own (i.e., non-commercial) MRI-compatible applicators that include an IUT. These applicators use rigid titanium or plastic, straight or angled needles, and can provide distal parametrial coverage through the adaptation of commercially-available IUTs. In Spain, two well-known prototypes are the Benidorm and Pamplona templates (Fig. ). 3D printing for the individualised treatment of cervical carcinoma In recent years, notable advances have taken place in terms of the development of three-dimensional (3D) printers for medical supplies. These printers allow clinicians to design patient-specific applicators adapted to the patient's individual anatomy and tumour response. This allows for individualised solutions not available with standard commercial applicators. It important to note that these applicators must be produced in a healthcare environment to ensure their suitability for clinical use in patients. The Navarra Hospital Complex developed a 3D printed endocavitary-interstitial applicator to treat the entire length of the vagina irregularly and to insert parametrial needles. This applicator is shown below in Fig. . Interstitial BT without perineal template (“freehand”) Interstitial implants can also be performed freehand (i.e., without the use of perineal templates) guided by vaginal and rectal touch, ultrasound or laparoscopy. The disadvantage of freehand needle insertion is the complexity of the technique. In addition, msince the geometry is not predefined, it is very user-dependent, which could lead to imprecise and unstable positioning, poor reproducibility, and potential loss of parallelism between the needles.
Martinez Universal Perineal Interstitial Template (MUPIT) This template is designed for gynaecological, prostate, anorectal, and perineal implants. The MUPIT consists of a double perineal template, a vaginal and rectal cylinder, and hollow 17-gauge guides separated by 6–10 mm. It has a total of 111 holes, which permit straight or angled vectors in horizontal planes, perpendicular to the plane of the perineal template. These are angled distally 14º, achieving adequate implantation of the parametria and greater lateral coverage of the CTV. The metal needles vary in length (range, 14–20 cm) and are held with stoppers to prevent cranio-caudal displacement during the procedure. A second template is used to strengthen vector fixation. The disadvantages include the lack of an intrauterine tube and the need for CT-based planning. Syed-Neblett applicator The modified Syed-Neblett device (Alfa Omega services, Bellflower CA) is based on the same principle as the original applicator and is only compatible with CT. It consists of a perineal template and a vaginal obturator with hollow 17-gauge guides of various lengths that allow users to combine interstitial sources and a vaginal cylinder with or without an IUT.
Venezia Applicator See description above. MAC applicator The MAC (acronym for the “Mick-Alektiar-Cohen” collaboration) applicator from Eckert & Ziegler Bebig (Germany), consists of a cylinder + IUT, is compatible with MRI and can be used with plastic needles of various diameters up to 2 mm. It has 36 concentric channels for the placement of unilateral or bilateral needles in parametria and can be used with a 0º or 30º IUT. Kelowna template Varian Medical Systems has developed the Kelowna gynaecological templates for interstitial implants, with a universal cylinder as a vaginal obturator. A 25 mm central hole in the template allows for the insertion of an ultrasound probe to guide the placement of needles of various sizes and lengths (113, 200, or 320 mm), whose tips can be sharp or blunt. The needles can be made of polyetheretherketone (PEEK), steel, or titanium. Several teams with extensive experience in perineal ISBT have developed their own (i.e., non-commercial) MRI-compatible applicators that include an IUT. These applicators use rigid titanium or plastic, straight or angled needles, and can provide distal parametrial coverage through the adaptation of commercially-available IUTs. In Spain, two well-known prototypes are the Benidorm and Pamplona templates (Fig. ).
In recent years, notable advances have taken place in terms of the development of three-dimensional (3D) printers for medical supplies. These printers allow clinicians to design patient-specific applicators adapted to the patient's individual anatomy and tumour response. This allows for individualised solutions not available with standard commercial applicators. It important to note that these applicators must be produced in a healthcare environment to ensure their suitability for clinical use in patients. The Navarra Hospital Complex developed a 3D printed endocavitary-interstitial applicator to treat the entire length of the vagina irregularly and to insert parametrial needles. This applicator is shown below in Fig. .
Interstitial implants can also be performed freehand (i.e., without the use of perineal templates) guided by vaginal and rectal touch, ultrasound or laparoscopy. The disadvantage of freehand needle insertion is the complexity of the technique. In addition, msince the geometry is not predefined, it is very user-dependent, which could lead to imprecise and unstable positioning, poor reproducibility, and potential loss of parallelism between the needles.
Perineal templates Perineal implants with templates present some differences compared to intracavitary/hybrid HDR-BT applicators. Before performing this technique, the first step is to determine exactly how many needles are needed, and their position, obliquity, and depth. Implant displacement must be taken into account when switching the patient from the gynaecological to the supine position, and when repositioning the needles to the cranio-caudal direction, as this can lead to insufficient coverage of the CTV. During repositioning, it is important to consider this movement when inserting the needles and also the needle offset, which will indicate the first position of the source and the start of irradiation. Ideally, needle insertion should be guided by ultrasound, marking the gross tumour volume (GTV) with clips or fiducial markers. The bladder should be filled with radiopaque contrast to allow for bladder visualisation to reduce the risk of perforation. The templates have a central vaginal cylinder, which maintains the geometric rigidity of the needles. However, one of the drawbacks of the commercial templates is that only one cylinder size is available (length: 13 cm; diameter: 3 cm), which is problematic in patients with a narrow, short, or poorly distensible vagina. If the vaginal cylinder cannot be placed, it can serve as a guide to place the vectors, and be removed later. The template is sutured to the perineum. The needles have a stop to prevent displacement. Perineal implants are usually performed only once and then left in place for several days to administer several fractions (delivered twice daily). For this reason, preventive measures are necessary (i.e., anti-bedsore techniques, analgesia, antithrombotics, and antibiotics). The patient cannot move nor raise the headrest above 30º. It is essential that the staff receive specialised training to avoid complications while the implant remains in place. Interstitial vulvar implants This inpatient procedure requires hospital admission. Bowel preparation prior to the implant is recommended. The procedure is performed in the operating room with the patient in the lithotomy position under general or spinal anaesthesia. Bladder catheterisation is recommended. A thorough clinical examination under anaesthesia with disease mapping is recommended prior to implantation. Before planning, it is essential to consider the implant design; if possible, the tubes should be inserted anterior-posteriorly, rather than transversally, to prevent the tube from bending when the legs are lowered from the lithotomy position. Interstitial implants can be performed with rigid needles or plastic tubes, with the latter being more comfortable for the patient. The procedure is usually performed freehand. The target volume encompasses the GTV—or the tumour bed in case of postoperative BT—plus a safety margin of 5–15 mm. Vector needles are inserted parallel to the labia or the vulvectomy scar in one or two planes to cover the residual tumour size in all three dimensions. Spacing should be guided by the Paris system rules. The distance between needles ranges from 10 to 15 mm. Special care is needed to prevent hot spots (200%). If necessary, extra catheters can be added to the surface of the tumour to ensure adequate coverage. The needles can be replaced by plastic catheters, fixed with buttons to the skin at the entry and exit points to ensure their stability, and can be combined with a vaginal cylinder or vaginal packing as a stay. In deep tumours involving the vagina, urethra, paravaginal or paraurethral space, a combined technique consisting of a vaginal cylinder and rigid needles with or without a perineal template should be used. In these cases, vaginal and endorectal ultrasound may be useful to guide the placement of rigid needles. Analgesics, prophylactic antibiotics, and anti-oedema measures are recommended throughout the hospital stay until treatment finalisation. Implant removal should be performed carefully after treatment completion, using all necessary aseptic precautions. After implant remove, the patient can be discharged.
Perineal implants with templates present some differences compared to intracavitary/hybrid HDR-BT applicators. Before performing this technique, the first step is to determine exactly how many needles are needed, and their position, obliquity, and depth. Implant displacement must be taken into account when switching the patient from the gynaecological to the supine position, and when repositioning the needles to the cranio-caudal direction, as this can lead to insufficient coverage of the CTV. During repositioning, it is important to consider this movement when inserting the needles and also the needle offset, which will indicate the first position of the source and the start of irradiation. Ideally, needle insertion should be guided by ultrasound, marking the gross tumour volume (GTV) with clips or fiducial markers. The bladder should be filled with radiopaque contrast to allow for bladder visualisation to reduce the risk of perforation. The templates have a central vaginal cylinder, which maintains the geometric rigidity of the needles. However, one of the drawbacks of the commercial templates is that only one cylinder size is available (length: 13 cm; diameter: 3 cm), which is problematic in patients with a narrow, short, or poorly distensible vagina. If the vaginal cylinder cannot be placed, it can serve as a guide to place the vectors, and be removed later. The template is sutured to the perineum. The needles have a stop to prevent displacement. Perineal implants are usually performed only once and then left in place for several days to administer several fractions (delivered twice daily). For this reason, preventive measures are necessary (i.e., anti-bedsore techniques, analgesia, antithrombotics, and antibiotics). The patient cannot move nor raise the headrest above 30º. It is essential that the staff receive specialised training to avoid complications while the implant remains in place.
This inpatient procedure requires hospital admission. Bowel preparation prior to the implant is recommended. The procedure is performed in the operating room with the patient in the lithotomy position under general or spinal anaesthesia. Bladder catheterisation is recommended. A thorough clinical examination under anaesthesia with disease mapping is recommended prior to implantation. Before planning, it is essential to consider the implant design; if possible, the tubes should be inserted anterior-posteriorly, rather than transversally, to prevent the tube from bending when the legs are lowered from the lithotomy position. Interstitial implants can be performed with rigid needles or plastic tubes, with the latter being more comfortable for the patient. The procedure is usually performed freehand. The target volume encompasses the GTV—or the tumour bed in case of postoperative BT—plus a safety margin of 5–15 mm. Vector needles are inserted parallel to the labia or the vulvectomy scar in one or two planes to cover the residual tumour size in all three dimensions. Spacing should be guided by the Paris system rules. The distance between needles ranges from 10 to 15 mm. Special care is needed to prevent hot spots (200%). If necessary, extra catheters can be added to the surface of the tumour to ensure adequate coverage. The needles can be replaced by plastic catheters, fixed with buttons to the skin at the entry and exit points to ensure their stability, and can be combined with a vaginal cylinder or vaginal packing as a stay. In deep tumours involving the vagina, urethra, paravaginal or paraurethral space, a combined technique consisting of a vaginal cylinder and rigid needles with or without a perineal template should be used. In these cases, vaginal and endorectal ultrasound may be useful to guide the placement of rigid needles. Analgesics, prophylactic antibiotics, and anti-oedema measures are recommended throughout the hospital stay until treatment finalisation. Implant removal should be performed carefully after treatment completion, using all necessary aseptic precautions. After implant remove, the patient can be discharged.
Planning images The type of applicator is usually defined at diagnosis, based on clinical examination and previous MRI. Locally-advanced tumours may have different growth patterns: expansive or infiltrative and responsive, and may produce anatomical changes and changes in the adjacent topography. Whenever possible, MRI should be performed prior to BT to allow for selection of the applicator that offers the most complete coverage. The routine use of preplanning will facilitate the BT procedure in locally-advanced cervical tumours by offering the potential to obtain a geometrically optimal implant, although these are not currently available. The main drawbacks to this approach are the anatomical changes that occur when placing the IUT in the implant and straightening the uterus, and the scant studies on this procedure . The gold standard imaging technique for cervical carcinoma is MRI, which is used for diagnosis, to assess treatment response, and for BT treatment planning. In 2012, the GEC-ESTRO published recommendations (4th edition) on the use of MRI, which included recommendation on how to prepare patients for the scan, as well as technical aspects and image acquisition protocols . Although these guidelines can serve as a baseline, each department must develop, in collaboration with the radiology department, its own criteria for MRI . In most centres, the MRI units are either 1.5 or 3 T, with 1.5 T MRI being the most common. Ideally, the duration of the imaging study should as short as possible to minimize movement. Intravenous contrast or intracavitary coils are not necessary. Although MRI is superior to CT in cervical carcinoma, not all radiotherapy departments have access to this imaging modality nor the capacity to perform MRI-based dosimetry. Moreover, MRI is more expensive than CT and involves a major learning curve, which explains why CT continues to be used for BT planning in many departments around the world. MRI is also recommended for target volume delineation when using ISBT to treat vulvar tumours. However, due to the complexity of reconstruction of the plastic needles in MRI images, fusion with CT is often required. Image fusion must be highly precise, and CT scans should be acquired with the patient in the supine position (slice thickness: 1–3 mm). To help in the contouring process, a thin wire may be placed on the tumour edge prior to acquisition (this wire may be used for target contouring in each slice). Contouring/prescription Intracavitary/interstitial hybrid HDR-BT The contouring recommendations proposed by the GEC-ESTRO have been accepted by the main brachytherapy societies, including the American Brachytherapy Society (ABS) and the GEB. Contouring of the following structures is recommended: (1) GTV: visible macroscopic tumour at the time of BT, detected by physical examination or as a bright image on the T2 MRI sequence; (2) HR-CTV: the high-risk tumour volume includes the GTV, the entire cervix, and any microscopic residual disease in the initial tumour location (parametrium, vagina), which are visible on the T2 MRI sequence as residual gray areas; (3) IR-CT: the intermediate-risk tumour volume, defined as the initial tumour location. This volume is calculated by expanding the HR-CTV with a 5–10 mm margin in all directions, depending on tumour response and limited by the risk organs; 4) OARs: rectum, sigmoid, bladder and intestinal loops should be contoured. Rectum: This should be contoured as a complete organ from the anal margin to the peritoneal reflection. Sigma: This should be contoured as a complete organ from the peritoneal reflection to two cm above the fundus. Bladder: This is contoured as a complete organ, marked by the end of the bladder neck after the balloon is no longer visible from the bladder catheter. Intestinal loops: These loops should be contoured at least in the 2-cm area surrounding the HR-CTV. Perineal templates Perineal implants have special characteristics, so we cannot always apply the published recommendations for intracavitary/hybrid implants. The process involves a single implant that covers both the disease at diagnosis (IR-CTV) and the disease at the time of BT (GTV and HR-CTV), in which treatment volumes are often much greater than usual, especially in cases with involvement of the distal parametrium or the entire vagina. Planning is usually CT-based, although this makes it more difficult to define the tumour, cervix, GTV, and OARs. In implants that use MRI-compatible applicators, the CTVs are smaller, thus higher doses can be administered with less toxicity. Although perforation of healthy organs is not uncommon, needle removal is not indicated since there is no increase in toxicity if appropriate measures are applied (e.g., antibiotics) and these positions are left unactivated at the time of planning . Primary vaginal malignancies/vaginal recurrences In the last decade, several different multi-institutional consensus statements and guidelines have been developed to define BT volumes in vaginal cancer . Volume definition in BT has evolved from the use of orthogonal plaques for the treatment plan to MRI-guided imaging. As a result, the various guidelines and consensus statements differ in terms of imaging and CTV definition . In 2012, the ABS published guidelines and recommendations for ISBT for vaginal cancer (primary tumours and recurrent disease) . In those guidelines, the CTV was defined as the residual tumour on MRI at the time of BT plus the cervix and the whole vagina. As those guidelines underscore, either intracavitary or interstitial BT may be used, depending on the tumour extension, thickness (< or > 5 mm), location, and morphology . For primary vaginal cancer, the GYN GEC-ESTRO group has adopted the validated target concepts used for cervical cancer. In 2019, the task group published the first recommendations and target concepts for IGBT in primary vaginal cancer, pending evaluation in a prospective, multicentre study. MRI and gynaecological examination at the time of BT play an important role in defining target volumes for IGBT. Based on concepts learned in the treatment of cervical cancer, the following volumes are defined: GTV-T res : gross residual tumour at time of BT on clinical examination and/or imaging CTV-T HR : high-risk clinical target volume, including the residual GTV-T and areas with the presence of pathological tissues. CTV-T IR : intermediate-risk clinical target volume, including all significant microscopic disease adjacent to the HRCTV-T. For the treatment of vaginal recurrences, several multi-institutional consensus statements have been published based on the target concepts previously defined by the GEC-ESTRO . In bulky vaginal tumours (target volume or OARs), the Canadian consensus recommends including the entire circumferential vagina in the CTV-T HR at the GTV level. In addition, the CTV-T IR should include the entire vaginal circumference at the level of the CTV-T HR. . The Canadian consensus also adds a larger margin at the cranio-caudal level, following the recommendations of the GEC-ESTRO consensus. However, there is still some debate regarding tumours with adenocarcinoma histology, with many authors recommending inclusion of the entire vaginal length in the CTV-T IR. At present, work is ongoing to develop consensus terminology for dose prescription. Interstitial vulvar BT Preferably, the GTV should be delineated on MRI. In the adjuvant setting, if no macroscopic disease is present, then no GTV BT is delimited. GTV BT includes postoperative residual vulvar/perineal disease detected on imaging and/or physical examination. In locally-advanced vulvar cancer, the GTV BT includes vulvar/perineal disease detected on imaging/physical examination. In both residual vulvar/perineal disease and in locally-advanced vulvar cancer, the macroscopic volume should be contoured if BT is administered as a boost after EBRT; for primary BT, the tumour volume at diagnosis should be contoured. The target volume encompasses the GTV (or the tumour bed in case of postoperative BT) as well changes detected on MRI if EBRT is administered prior to BT. The target volume also includes all suspicious findings on physical examination with a 10–15 mm margin (excluding certain structures [e.g., urethra, anus]). The larger margin can be applied in the same direction as the tubes. According to the Paris system, dose prescription should be at the reference isodose (85% of the minimal dose rate between planes). Doses should be reported after conversion into radiobiologically-weighted dose equivalent of 2 Gy/fraction (α/β 10 Gy for the tumour, half-time of 1.5 h), EQD2 Gy . Reconstruction Intracavitary/interstitial hybrid HDR-BT In recent years, BT techniques have improved substantially. Probably the most notable improvement is the potential to individualise treatment at any point in the process. A wide variety of applicators are available, thus covering the diversity of different tumour anatomies. In this regard, the recent addition of the interstitial component to standard applicators is an especially notable improvement. In addition, wider access to volumetric imaging techniques (e.g., CT, MRI, and US) has made it possible to adapt dosimetry to anatomical volumes, thus obviating the older points-based dosimetric techniques. These two changes have, in turn, altered the paradigm of applicator reconstruction and dosimetric optimisation. Below we summarize the recommendations of the GEB-SEFM group for both of these points. A wide variety of applicator and imaging modality combinations are available (plastic or metal applicators, CT, MRI or US imaging). As a result, it is necessary to approach each option differently. One of the great challenges for medical physicists is reconstruction of the applicators, which requires proper commissioning and quality control of the applicators (including the different imaging modalities used and the corresponding dummies). These new forms of work introduce new uncertainties that must be considered. In this regard, we recommend that clinicians read the studies by Hellebust et al. and Tanderup et al. . The accuracy of applicator reconstruction is conditioned by the image acquisition parameters, especially the slice thickness. Currently, the maximum recommended slice thickness is ≤ 5 mm (64). However, keeping in mind this limitation, the individual centre may choose to acquire images (both CT and MRI) with smaller thicknesses. When using CT images, it is important to note the risk of large artifacts produced by applicator dummies and markers, which could hinder accurate contouring of the volumes of interest (VOI; target volumes and OARs). Usually, the applicator channel is visible on CT images, so we may consider omitting dummies, referring the dwell positions to the visible tip and walls of the applicator. It is also recommended to determine whether CT image reconstruction can be done through the application of iterative techniques and artifact reduction. In some cases, when working with MRI, there is an added difficulty: the need for a commercially-available dummy. Due to lack of availability, most centres must use customised solutions designed in-house, such as tubes filled with fluids (e.g., saline, paraffin, etc.) that are compatible with and visible on MRI. The diversity of MRI sequences must also be considered. The best MRI sequence to visualise tumour tissues is T2, but this is not the most suitable sequence for applicator reconstruction, for which T1 or CT are better. In this situation, registration of the different image sequences can be assessed. If this option is selected, it is essential to ensure that the registration is performed with a focus on the rigid part of the applicator instead of the VOIs. Regardless of the chosen imaging modality, it is highly recommended to take advantage of the different views (axial, coronal, and sagittal). If different MRI sequences are used, it is important to verify that the patient has not moved between acquisitions; if so, this must be taken into account. The recommended view used to outline the VOIs should be the one preferred for applicator reconstruction, especially if there is any mismatch between the different views. For example, Tanderup et al. recommended using a longitudinal plane to avoid the uncertainty caused by the slice thickness. In this image set, the distance from the centre of the axial slice to the channel centre can be measured. In short, several different reconstruction options are available. These can be summarized as follows: (1) direct reconstruction on CT or T2 MRI slices, or (2) reconstruction on the support image (CT, T1 MRI, or other acquisition sequences) and registration between support and reference images. In recent years, some institutions have started using 3D-MRI acquisitions, which simplifies applicator reconstruction, even in T2-weighted series (by reducing slice thickness). Finally, it is worth noting that several commercial vendors are developing automatic reconstruction methods, such as those based on MRI-tracking of a virtual source. When these tools become available, this will greatly facilitate catheter reconstruction, thus shortening the process. This is important given the most time-consuming task after applicator insertion is applicator reconstruction. Automating this process will likely reduce the risk of mistakes. When all the catheters are visible on the image it is important to verify that the reconstructed channels are correctly assigned to the real channels of the applicator to avoid connection errors.
The type of applicator is usually defined at diagnosis, based on clinical examination and previous MRI. Locally-advanced tumours may have different growth patterns: expansive or infiltrative and responsive, and may produce anatomical changes and changes in the adjacent topography. Whenever possible, MRI should be performed prior to BT to allow for selection of the applicator that offers the most complete coverage. The routine use of preplanning will facilitate the BT procedure in locally-advanced cervical tumours by offering the potential to obtain a geometrically optimal implant, although these are not currently available. The main drawbacks to this approach are the anatomical changes that occur when placing the IUT in the implant and straightening the uterus, and the scant studies on this procedure . The gold standard imaging technique for cervical carcinoma is MRI, which is used for diagnosis, to assess treatment response, and for BT treatment planning. In 2012, the GEC-ESTRO published recommendations (4th edition) on the use of MRI, which included recommendation on how to prepare patients for the scan, as well as technical aspects and image acquisition protocols . Although these guidelines can serve as a baseline, each department must develop, in collaboration with the radiology department, its own criteria for MRI . In most centres, the MRI units are either 1.5 or 3 T, with 1.5 T MRI being the most common. Ideally, the duration of the imaging study should as short as possible to minimize movement. Intravenous contrast or intracavitary coils are not necessary. Although MRI is superior to CT in cervical carcinoma, not all radiotherapy departments have access to this imaging modality nor the capacity to perform MRI-based dosimetry. Moreover, MRI is more expensive than CT and involves a major learning curve, which explains why CT continues to be used for BT planning in many departments around the world. MRI is also recommended for target volume delineation when using ISBT to treat vulvar tumours. However, due to the complexity of reconstruction of the plastic needles in MRI images, fusion with CT is often required. Image fusion must be highly precise, and CT scans should be acquired with the patient in the supine position (slice thickness: 1–3 mm). To help in the contouring process, a thin wire may be placed on the tumour edge prior to acquisition (this wire may be used for target contouring in each slice).
Intracavitary/interstitial hybrid HDR-BT The contouring recommendations proposed by the GEC-ESTRO have been accepted by the main brachytherapy societies, including the American Brachytherapy Society (ABS) and the GEB. Contouring of the following structures is recommended: (1) GTV: visible macroscopic tumour at the time of BT, detected by physical examination or as a bright image on the T2 MRI sequence; (2) HR-CTV: the high-risk tumour volume includes the GTV, the entire cervix, and any microscopic residual disease in the initial tumour location (parametrium, vagina), which are visible on the T2 MRI sequence as residual gray areas; (3) IR-CT: the intermediate-risk tumour volume, defined as the initial tumour location. This volume is calculated by expanding the HR-CTV with a 5–10 mm margin in all directions, depending on tumour response and limited by the risk organs; 4) OARs: rectum, sigmoid, bladder and intestinal loops should be contoured. Rectum: This should be contoured as a complete organ from the anal margin to the peritoneal reflection. Sigma: This should be contoured as a complete organ from the peritoneal reflection to two cm above the fundus. Bladder: This is contoured as a complete organ, marked by the end of the bladder neck after the balloon is no longer visible from the bladder catheter. Intestinal loops: These loops should be contoured at least in the 2-cm area surrounding the HR-CTV. Perineal templates Perineal implants have special characteristics, so we cannot always apply the published recommendations for intracavitary/hybrid implants. The process involves a single implant that covers both the disease at diagnosis (IR-CTV) and the disease at the time of BT (GTV and HR-CTV), in which treatment volumes are often much greater than usual, especially in cases with involvement of the distal parametrium or the entire vagina. Planning is usually CT-based, although this makes it more difficult to define the tumour, cervix, GTV, and OARs. In implants that use MRI-compatible applicators, the CTVs are smaller, thus higher doses can be administered with less toxicity. Although perforation of healthy organs is not uncommon, needle removal is not indicated since there is no increase in toxicity if appropriate measures are applied (e.g., antibiotics) and these positions are left unactivated at the time of planning . Primary vaginal malignancies/vaginal recurrences In the last decade, several different multi-institutional consensus statements and guidelines have been developed to define BT volumes in vaginal cancer . Volume definition in BT has evolved from the use of orthogonal plaques for the treatment plan to MRI-guided imaging. As a result, the various guidelines and consensus statements differ in terms of imaging and CTV definition . In 2012, the ABS published guidelines and recommendations for ISBT for vaginal cancer (primary tumours and recurrent disease) . In those guidelines, the CTV was defined as the residual tumour on MRI at the time of BT plus the cervix and the whole vagina. As those guidelines underscore, either intracavitary or interstitial BT may be used, depending on the tumour extension, thickness (< or > 5 mm), location, and morphology . For primary vaginal cancer, the GYN GEC-ESTRO group has adopted the validated target concepts used for cervical cancer. In 2019, the task group published the first recommendations and target concepts for IGBT in primary vaginal cancer, pending evaluation in a prospective, multicentre study. MRI and gynaecological examination at the time of BT play an important role in defining target volumes for IGBT. Based on concepts learned in the treatment of cervical cancer, the following volumes are defined: GTV-T res : gross residual tumour at time of BT on clinical examination and/or imaging CTV-T HR : high-risk clinical target volume, including the residual GTV-T and areas with the presence of pathological tissues. CTV-T IR : intermediate-risk clinical target volume, including all significant microscopic disease adjacent to the HRCTV-T. For the treatment of vaginal recurrences, several multi-institutional consensus statements have been published based on the target concepts previously defined by the GEC-ESTRO . In bulky vaginal tumours (target volume or OARs), the Canadian consensus recommends including the entire circumferential vagina in the CTV-T HR at the GTV level. In addition, the CTV-T IR should include the entire vaginal circumference at the level of the CTV-T HR. . The Canadian consensus also adds a larger margin at the cranio-caudal level, following the recommendations of the GEC-ESTRO consensus. However, there is still some debate regarding tumours with adenocarcinoma histology, with many authors recommending inclusion of the entire vaginal length in the CTV-T IR. At present, work is ongoing to develop consensus terminology for dose prescription. Interstitial vulvar BT Preferably, the GTV should be delineated on MRI. In the adjuvant setting, if no macroscopic disease is present, then no GTV BT is delimited. GTV BT includes postoperative residual vulvar/perineal disease detected on imaging and/or physical examination. In locally-advanced vulvar cancer, the GTV BT includes vulvar/perineal disease detected on imaging/physical examination. In both residual vulvar/perineal disease and in locally-advanced vulvar cancer, the macroscopic volume should be contoured if BT is administered as a boost after EBRT; for primary BT, the tumour volume at diagnosis should be contoured. The target volume encompasses the GTV (or the tumour bed in case of postoperative BT) as well changes detected on MRI if EBRT is administered prior to BT. The target volume also includes all suspicious findings on physical examination with a 10–15 mm margin (excluding certain structures [e.g., urethra, anus]). The larger margin can be applied in the same direction as the tubes. According to the Paris system, dose prescription should be at the reference isodose (85% of the minimal dose rate between planes). Doses should be reported after conversion into radiobiologically-weighted dose equivalent of 2 Gy/fraction (α/β 10 Gy for the tumour, half-time of 1.5 h), EQD2 Gy .
The contouring recommendations proposed by the GEC-ESTRO have been accepted by the main brachytherapy societies, including the American Brachytherapy Society (ABS) and the GEB. Contouring of the following structures is recommended: (1) GTV: visible macroscopic tumour at the time of BT, detected by physical examination or as a bright image on the T2 MRI sequence; (2) HR-CTV: the high-risk tumour volume includes the GTV, the entire cervix, and any microscopic residual disease in the initial tumour location (parametrium, vagina), which are visible on the T2 MRI sequence as residual gray areas; (3) IR-CT: the intermediate-risk tumour volume, defined as the initial tumour location. This volume is calculated by expanding the HR-CTV with a 5–10 mm margin in all directions, depending on tumour response and limited by the risk organs; 4) OARs: rectum, sigmoid, bladder and intestinal loops should be contoured. Rectum: This should be contoured as a complete organ from the anal margin to the peritoneal reflection. Sigma: This should be contoured as a complete organ from the peritoneal reflection to two cm above the fundus. Bladder: This is contoured as a complete organ, marked by the end of the bladder neck after the balloon is no longer visible from the bladder catheter. Intestinal loops: These loops should be contoured at least in the 2-cm area surrounding the HR-CTV.
Perineal implants have special characteristics, so we cannot always apply the published recommendations for intracavitary/hybrid implants. The process involves a single implant that covers both the disease at diagnosis (IR-CTV) and the disease at the time of BT (GTV and HR-CTV), in which treatment volumes are often much greater than usual, especially in cases with involvement of the distal parametrium or the entire vagina. Planning is usually CT-based, although this makes it more difficult to define the tumour, cervix, GTV, and OARs. In implants that use MRI-compatible applicators, the CTVs are smaller, thus higher doses can be administered with less toxicity. Although perforation of healthy organs is not uncommon, needle removal is not indicated since there is no increase in toxicity if appropriate measures are applied (e.g., antibiotics) and these positions are left unactivated at the time of planning .
In the last decade, several different multi-institutional consensus statements and guidelines have been developed to define BT volumes in vaginal cancer . Volume definition in BT has evolved from the use of orthogonal plaques for the treatment plan to MRI-guided imaging. As a result, the various guidelines and consensus statements differ in terms of imaging and CTV definition . In 2012, the ABS published guidelines and recommendations for ISBT for vaginal cancer (primary tumours and recurrent disease) . In those guidelines, the CTV was defined as the residual tumour on MRI at the time of BT plus the cervix and the whole vagina. As those guidelines underscore, either intracavitary or interstitial BT may be used, depending on the tumour extension, thickness (< or > 5 mm), location, and morphology . For primary vaginal cancer, the GYN GEC-ESTRO group has adopted the validated target concepts used for cervical cancer. In 2019, the task group published the first recommendations and target concepts for IGBT in primary vaginal cancer, pending evaluation in a prospective, multicentre study. MRI and gynaecological examination at the time of BT play an important role in defining target volumes for IGBT. Based on concepts learned in the treatment of cervical cancer, the following volumes are defined: GTV-T res : gross residual tumour at time of BT on clinical examination and/or imaging CTV-T HR : high-risk clinical target volume, including the residual GTV-T and areas with the presence of pathological tissues. CTV-T IR : intermediate-risk clinical target volume, including all significant microscopic disease adjacent to the HRCTV-T. For the treatment of vaginal recurrences, several multi-institutional consensus statements have been published based on the target concepts previously defined by the GEC-ESTRO . In bulky vaginal tumours (target volume or OARs), the Canadian consensus recommends including the entire circumferential vagina in the CTV-T HR at the GTV level. In addition, the CTV-T IR should include the entire vaginal circumference at the level of the CTV-T HR. . The Canadian consensus also adds a larger margin at the cranio-caudal level, following the recommendations of the GEC-ESTRO consensus. However, there is still some debate regarding tumours with adenocarcinoma histology, with many authors recommending inclusion of the entire vaginal length in the CTV-T IR. At present, work is ongoing to develop consensus terminology for dose prescription.
Preferably, the GTV should be delineated on MRI. In the adjuvant setting, if no macroscopic disease is present, then no GTV BT is delimited. GTV BT includes postoperative residual vulvar/perineal disease detected on imaging and/or physical examination. In locally-advanced vulvar cancer, the GTV BT includes vulvar/perineal disease detected on imaging/physical examination. In both residual vulvar/perineal disease and in locally-advanced vulvar cancer, the macroscopic volume should be contoured if BT is administered as a boost after EBRT; for primary BT, the tumour volume at diagnosis should be contoured. The target volume encompasses the GTV (or the tumour bed in case of postoperative BT) as well changes detected on MRI if EBRT is administered prior to BT. The target volume also includes all suspicious findings on physical examination with a 10–15 mm margin (excluding certain structures [e.g., urethra, anus]). The larger margin can be applied in the same direction as the tubes. According to the Paris system, dose prescription should be at the reference isodose (85% of the minimal dose rate between planes). Doses should be reported after conversion into radiobiologically-weighted dose equivalent of 2 Gy/fraction (α/β 10 Gy for the tumour, half-time of 1.5 h), EQD2 Gy .
Intracavitary/interstitial hybrid HDR-BT In recent years, BT techniques have improved substantially. Probably the most notable improvement is the potential to individualise treatment at any point in the process. A wide variety of applicators are available, thus covering the diversity of different tumour anatomies. In this regard, the recent addition of the interstitial component to standard applicators is an especially notable improvement. In addition, wider access to volumetric imaging techniques (e.g., CT, MRI, and US) has made it possible to adapt dosimetry to anatomical volumes, thus obviating the older points-based dosimetric techniques. These two changes have, in turn, altered the paradigm of applicator reconstruction and dosimetric optimisation. Below we summarize the recommendations of the GEB-SEFM group for both of these points. A wide variety of applicator and imaging modality combinations are available (plastic or metal applicators, CT, MRI or US imaging). As a result, it is necessary to approach each option differently. One of the great challenges for medical physicists is reconstruction of the applicators, which requires proper commissioning and quality control of the applicators (including the different imaging modalities used and the corresponding dummies). These new forms of work introduce new uncertainties that must be considered. In this regard, we recommend that clinicians read the studies by Hellebust et al. and Tanderup et al. . The accuracy of applicator reconstruction is conditioned by the image acquisition parameters, especially the slice thickness. Currently, the maximum recommended slice thickness is ≤ 5 mm (64). However, keeping in mind this limitation, the individual centre may choose to acquire images (both CT and MRI) with smaller thicknesses. When using CT images, it is important to note the risk of large artifacts produced by applicator dummies and markers, which could hinder accurate contouring of the volumes of interest (VOI; target volumes and OARs). Usually, the applicator channel is visible on CT images, so we may consider omitting dummies, referring the dwell positions to the visible tip and walls of the applicator. It is also recommended to determine whether CT image reconstruction can be done through the application of iterative techniques and artifact reduction. In some cases, when working with MRI, there is an added difficulty: the need for a commercially-available dummy. Due to lack of availability, most centres must use customised solutions designed in-house, such as tubes filled with fluids (e.g., saline, paraffin, etc.) that are compatible with and visible on MRI. The diversity of MRI sequences must also be considered. The best MRI sequence to visualise tumour tissues is T2, but this is not the most suitable sequence for applicator reconstruction, for which T1 or CT are better. In this situation, registration of the different image sequences can be assessed. If this option is selected, it is essential to ensure that the registration is performed with a focus on the rigid part of the applicator instead of the VOIs. Regardless of the chosen imaging modality, it is highly recommended to take advantage of the different views (axial, coronal, and sagittal). If different MRI sequences are used, it is important to verify that the patient has not moved between acquisitions; if so, this must be taken into account. The recommended view used to outline the VOIs should be the one preferred for applicator reconstruction, especially if there is any mismatch between the different views. For example, Tanderup et al. recommended using a longitudinal plane to avoid the uncertainty caused by the slice thickness. In this image set, the distance from the centre of the axial slice to the channel centre can be measured. In short, several different reconstruction options are available. These can be summarized as follows: (1) direct reconstruction on CT or T2 MRI slices, or (2) reconstruction on the support image (CT, T1 MRI, or other acquisition sequences) and registration between support and reference images. In recent years, some institutions have started using 3D-MRI acquisitions, which simplifies applicator reconstruction, even in T2-weighted series (by reducing slice thickness). Finally, it is worth noting that several commercial vendors are developing automatic reconstruction methods, such as those based on MRI-tracking of a virtual source. When these tools become available, this will greatly facilitate catheter reconstruction, thus shortening the process. This is important given the most time-consuming task after applicator insertion is applicator reconstruction. Automating this process will likely reduce the risk of mistakes. When all the catheters are visible on the image it is important to verify that the reconstructed channels are correctly assigned to the real channels of the applicator to avoid connection errors.
In recent years, BT techniques have improved substantially. Probably the most notable improvement is the potential to individualise treatment at any point in the process. A wide variety of applicators are available, thus covering the diversity of different tumour anatomies. In this regard, the recent addition of the interstitial component to standard applicators is an especially notable improvement. In addition, wider access to volumetric imaging techniques (e.g., CT, MRI, and US) has made it possible to adapt dosimetry to anatomical volumes, thus obviating the older points-based dosimetric techniques. These two changes have, in turn, altered the paradigm of applicator reconstruction and dosimetric optimisation. Below we summarize the recommendations of the GEB-SEFM group for both of these points. A wide variety of applicator and imaging modality combinations are available (plastic or metal applicators, CT, MRI or US imaging). As a result, it is necessary to approach each option differently. One of the great challenges for medical physicists is reconstruction of the applicators, which requires proper commissioning and quality control of the applicators (including the different imaging modalities used and the corresponding dummies). These new forms of work introduce new uncertainties that must be considered. In this regard, we recommend that clinicians read the studies by Hellebust et al. and Tanderup et al. . The accuracy of applicator reconstruction is conditioned by the image acquisition parameters, especially the slice thickness. Currently, the maximum recommended slice thickness is ≤ 5 mm (64). However, keeping in mind this limitation, the individual centre may choose to acquire images (both CT and MRI) with smaller thicknesses. When using CT images, it is important to note the risk of large artifacts produced by applicator dummies and markers, which could hinder accurate contouring of the volumes of interest (VOI; target volumes and OARs). Usually, the applicator channel is visible on CT images, so we may consider omitting dummies, referring the dwell positions to the visible tip and walls of the applicator. It is also recommended to determine whether CT image reconstruction can be done through the application of iterative techniques and artifact reduction. In some cases, when working with MRI, there is an added difficulty: the need for a commercially-available dummy. Due to lack of availability, most centres must use customised solutions designed in-house, such as tubes filled with fluids (e.g., saline, paraffin, etc.) that are compatible with and visible on MRI. The diversity of MRI sequences must also be considered. The best MRI sequence to visualise tumour tissues is T2, but this is not the most suitable sequence for applicator reconstruction, for which T1 or CT are better. In this situation, registration of the different image sequences can be assessed. If this option is selected, it is essential to ensure that the registration is performed with a focus on the rigid part of the applicator instead of the VOIs. Regardless of the chosen imaging modality, it is highly recommended to take advantage of the different views (axial, coronal, and sagittal). If different MRI sequences are used, it is important to verify that the patient has not moved between acquisitions; if so, this must be taken into account. The recommended view used to outline the VOIs should be the one preferred for applicator reconstruction, especially if there is any mismatch between the different views. For example, Tanderup et al. recommended using a longitudinal plane to avoid the uncertainty caused by the slice thickness. In this image set, the distance from the centre of the axial slice to the channel centre can be measured. In short, several different reconstruction options are available. These can be summarized as follows: (1) direct reconstruction on CT or T2 MRI slices, or (2) reconstruction on the support image (CT, T1 MRI, or other acquisition sequences) and registration between support and reference images. In recent years, some institutions have started using 3D-MRI acquisitions, which simplifies applicator reconstruction, even in T2-weighted series (by reducing slice thickness). Finally, it is worth noting that several commercial vendors are developing automatic reconstruction methods, such as those based on MRI-tracking of a virtual source. When these tools become available, this will greatly facilitate catheter reconstruction, thus shortening the process. This is important given the most time-consuming task after applicator insertion is applicator reconstruction. Automating this process will likely reduce the risk of mistakes. When all the catheters are visible on the image it is important to verify that the reconstructed channels are correctly assigned to the real channels of the applicator to avoid connection errors.
Due to the existence of various applicators and the possibilities offered by each planning system, the channel reconstruction procedure may vary. In general, it is recommended to use axes that can be modified by the user and whose orientation is parallel to the reconstruction channel. Before starting channel reconstruction, it is advisable to perform a general review of the channel positions, verifying the distance between channels (for example, between ovoids) and/or expected angles between the IUT and the ring surface. For rigid applicators, manual or applicator library reconstruction methods can be used. For curved and multi-channel applicators, it is highly recommended to use the libraries, since this reduces errors . When using MRI, the reconstruction can be performed relative to the applicator surface if the source path has been adequately characterised in that way during commissioning. Some applicators consist of several modules (e.g., two ovoids, or a ring and IUT); in these cases, the applicator library must allow each catheter to be manipulated separately if the user so desires. Catheter deformability is highly recommended. If the applicator library is not available, an alternative is to create a plan library. In this case, the user must make a “perfect” reconstruction in the preferred image modality, where the source path is easily identifiable. Once the reconstruction has been made, some planning systems allow the reconstruction to be saved as a plan, which can then be imported for use with any patient. When importing a reconstruction, specific applicator points must be identified so that placement of the applicator in the images matches perfectly (as occurs in the applicator library). By contrast, the interstitial needles must be directly reconstructed, since there are very few automatic reconstruction solutions in current planning systems. Most of these methods are based on the contrast differences that the channels produce with respect to the surrounding tissue, so they should work somewhat better with CT images, but not with MRI. When automatic methods are used, the reconstruction should be verified based on the information available from commissioning. During direct reconstruction, the use of too many points can lead to an unrealistic “zig-zag” path, which is highly likely to produce undesirable deviations in dose distribution. When the canal has a rectilinear shape, which happens with standard applicator needles or metallic needles, it is better to utilize a reconstruction that has only two or three points. The direction of treatment (i.e., whether the source stops are entering or exiting the applicator) must also be considered. The actual source position will not exactly match the positions indicated by the markers; the tension of the source cable will cause it to be attached to the external wall when the source enters the applicator during treatment and attached to the internal wall when it leaves. Reconstruction of applicators that combine rigid and interstitial components will probably require a mix of approaches. Perineal templates CT Based reconstruction Traditionally, perineal templates for interstitial cervix BT were based on stainless-steel needles. In the mid-1980s, Martinez et al. developed the MUPIT applicator; however, that applicator was not MR-compatible due to the use of stainless-steel needles. This meant that reconstruction had to be performed on CT-based study sets, which unfortunately, results in poor soft tissue contrast. Due to this poor contrast, together with image artifacts caused by the metallic needles, target visualisation is poor. In turn, this impedes precise delineation of the PTV. In these cases, a preplanning CT study set is useful. First, a CT study set of the patient is acquired and the perineal template is superimposed to the axial CT images to select the most appropriate location of the needles to ensure optimal target coverage. The most difficult part of this reconstruction is to precisely determine the tip location, and thus the first source dwell position. The accuracy of this determination is directly related to the slice thickness in the study set. Some treatment planning systems (TPS) have tools to perform “fine” navigation between slices. The scout view may be useful to precisely determine the needle tip localisation. In the cranio-caudal direction, the scout view has no deformation, thus allowing the user to visualise the needle tip to correlate it with the corresponding axial image (Fig. ). MRI-based reconstruction The use of MRI images for perineal ISBT procedures has several major advantages, including good soft tissue contrast and spatial resolution (3D sequences with isotropic voxel size ≈ 1 mm). Nevertheless, several issues must be considered when using these images . First, the applicators must be MRI compatible. Generally, perineal BT procedures are performed with titanium needles, which can be problematic in MRI images due to geometric distortion and magnetic susceptibility artifacts of the titanium needles. Nevertheless, several authors have developed strategies to overcome these drawbacks . In addition, geometric distortion must be assessed for the same type of sequences used clinically to obtain images suitable for planning. Commercial MRI distortion phantoms are available for these purposes. The signal of titanium needles on MRI images shows a “ballooning” artifact at the tip of the needle, making it difficult to determine the exact distal dwell position (Fig. ). However, there are many different solutions to solve this problem. For example, Haack et al. identified the artifact on MR images and correlated it to the dwell position by performing a registration with a CT study set obtained during the commissioning process. Other authors have suggested a method in which MRI markers (A-vitamin pellets)—which produce a high signal on MRI images—are embedded into the perineal template to identify a reference plane on these images. The direction and free length distance of the needles are used to obtain the needle tip coordinates (Fig. ). Rigid applicators commonly used in BT may be modelled into a TPS applicator library. Otal et al. developed a method to incorporate interstitial components into the TPS gynaecologic rigid applicator library. As mentioned above, a precise source path determination is required during the reconstruction process. In some situations, an additional series of images may be acquired—such as T1W, 3D-MRI or 3D Echo SPGR sequences—to improve applicator visibility. Registration accuracy between T2W and the other type of images must be assessed. Optimisation Intracavitary/interstitial hybrid HDR-BT Due to the rapid technological advances in the treatment of cervical cancer, currently there are no unified recommendations on how to optimise BT. The GEC-ESTRO gynaecology working group announced at the World Congress of Brachytherapy in 2021 that they intend to publish planning recommendations in 2022. However, pending publication of those guidelines, the criteria followed may vary from centre to centre, generally based on the centre’s experience together with published data from reference centres and other studies. The protocols followed by Spanish radiation oncology centres are described below, as reflected in the paper published in 2018 by Perez-Calatayud et al. on behalf of SEFM . Before discussing planning tips, it is important to mention that the participating centres in the study of SEFM agree that inverse optimisation should not routinely be used for this type of implant due to loss of control of the assigned times and because, in some situations, while the dosimetric parameters may be adequate, the implant may be excessively heterogeneous. For these reasons, manual methods are usually recommended. The isodose lines in the three views of the implant should be reviewed. Activation of dwell positions The step size should be based on the source size and activity. If the step size is too short, it could result in insufficient stopping times, which could increase uncertainty and ultimately cause deviations from the planned dosimetry. The activation of positions should refer to the VOIs, encompassing the entire HR-CTV plus 5 mm. If any OARs are located near active positions, deactivation of those positions should be considered. Selection of the margins for activation and the final active positions must be confirmed by the radiation oncologist. Optimisation of dwell positions This is the part of the planning process in which the greatest differences among centres can be observed. Two very different optimisation methods are described below for case in which a discrete number of needles are used. Later in this document, we describe other methods for use when the principal component of the implant is interstitial. In this first method, points are generated on the surface of the HR-CTV and the plan is normalised to these points, initially without optimisation (i.e., all dwell times are equal). Next, the stopping times are modified through manual optimisation to meet the criteria established at the treatment centre (e.g., maximum volume allowed for 200% [V200%], V150%, rectal and bladder doses, and HR-CTV and CTV-IR coverage). In the second method, planning is started with standard loading of the intracavitary component and normalisation to the A points (as established in the ABS 2012 recommendations) . Stopping times are modified to meet the dosimetric criteria for OARs at the treating centre. Next, positions in the needles are activated inside the HR-CTV (taking into account potential OAR hotspots) and short times are assigned (1–4 s) in an effort to maintain the same time at all positions. According to the initial published recommendations, the total time assigned to the needles should be < 20% of the total treatment time; however, this restriction has lost importance over time, as the benefits of increasing the IS/IC time ratio has become increasingly clear . Another point to keep in mind is that recommendations for the current dosimetric parameters have mostly been described for their evaluation in EQD2 (equivalent dose in 2 Gy fractions), by adding EBRT and BT dose. To our knowledge, at present (October 2022), none of the currently available TPS allow for dose evaluation in terms of EQD2 (nor for adding doses with EBRT). Consequently, medical physicists must use spreadsheets to carry out this dose assessment requirement. Finally, it is important to emphasise that optimisation must be carried out in close collaboration with the radiation oncologist, especially evaluation of the EQD2 and dose distribution. Perineal templates Activation of dwell positions Activated dwell positions must assure adequate CTV coverage, which is why dwell positions 3–5 mm beyond the CTV are often activated. In this regard, it is important to keep in mind that the metallic needles (titanium or stainless steel) have a blind end about 1 cm in lenght, which should be considered when inserting the needles. Many planning systems have tools to automatically activate dwell positions to cover the CTV with an adequate margin. The source step is 2–3 mm. Three primary strategies are available for optimisation: (1) inverse planning, (2) point-based optimisation, and (3) geometric optimisation. The experience with the inverse planning modules included in the TPS in these types of implants, dose distributions obtained via inverse planning are similar to those obtained with other methods in terms of target coverage, but more heterogeneous. A points-based optimisation consists of generating a mesh of points on the surface of the CTV (target points) and then prescribing the nominal dose to these points. However, when using this type of optimisation, it is not uncommon to find hot spots within the CTV. In our experience, the best way to control these hot spots is to use geometric optimisation, which results in small hot spots at the needle boundaries that do not overlap between adjacent needles. Small manual (graphical) adjustments are performed to assure adequate coverage of the CTV (Fig. ). Interstitial The implant dwell positions should be selected on the clinical target. Initially, plan optimisation can be performed using automated tools in the TPS, followed by manual optimisation if necessary. Alternatively, manual optimisation can be performed without the need for the automated tools. The dose to 90% of the CTV volume should meet the prescription goals if the CTV is contoured. Otherwise, normal tissue dosimetry should include descriptions of the doses to volumes such as 0.1 cm 3 , 1 cm 3 , and 2 cm 3 of the bladder, urethra, rectum, sigmoid colon, and small bowel, depending on the location of the lesion. Summation of the EBRT and BT doses can be performed using EQD2 (equivalent dose in 2 Gy fractions) with α/β of 10 Gy for tumour and 3 Gy for normal tissues. During optimisation, the dwell times should be reviewed to ensure that there are no unexpectedly high dwell times. The volume of tissues receiving > 150% of the prescription dose should be limited to the area near the interstitial needles. The use of quality indices to assess conformality and homogeneity is recommended. The conformity index (CI) is defined as (CTVref/VCTV) (CTVref/Vref) where CTVref is the CTV volume receiving a dose equal to or greater than the reference dose, VCTV is the CTV volume, and Vref is the volume receiving a dose equal to or greater than the reference dose. The homogeneity index (HI) is defined as the fraction receiving a dose between 100 and 150% of the reference dose .
CT Based reconstruction Traditionally, perineal templates for interstitial cervix BT were based on stainless-steel needles. In the mid-1980s, Martinez et al. developed the MUPIT applicator; however, that applicator was not MR-compatible due to the use of stainless-steel needles. This meant that reconstruction had to be performed on CT-based study sets, which unfortunately, results in poor soft tissue contrast. Due to this poor contrast, together with image artifacts caused by the metallic needles, target visualisation is poor. In turn, this impedes precise delineation of the PTV. In these cases, a preplanning CT study set is useful. First, a CT study set of the patient is acquired and the perineal template is superimposed to the axial CT images to select the most appropriate location of the needles to ensure optimal target coverage. The most difficult part of this reconstruction is to precisely determine the tip location, and thus the first source dwell position. The accuracy of this determination is directly related to the slice thickness in the study set. Some treatment planning systems (TPS) have tools to perform “fine” navigation between slices. The scout view may be useful to precisely determine the needle tip localisation. In the cranio-caudal direction, the scout view has no deformation, thus allowing the user to visualise the needle tip to correlate it with the corresponding axial image (Fig. ). MRI-based reconstruction The use of MRI images for perineal ISBT procedures has several major advantages, including good soft tissue contrast and spatial resolution (3D sequences with isotropic voxel size ≈ 1 mm). Nevertheless, several issues must be considered when using these images . First, the applicators must be MRI compatible. Generally, perineal BT procedures are performed with titanium needles, which can be problematic in MRI images due to geometric distortion and magnetic susceptibility artifacts of the titanium needles. Nevertheless, several authors have developed strategies to overcome these drawbacks . In addition, geometric distortion must be assessed for the same type of sequences used clinically to obtain images suitable for planning. Commercial MRI distortion phantoms are available for these purposes. The signal of titanium needles on MRI images shows a “ballooning” artifact at the tip of the needle, making it difficult to determine the exact distal dwell position (Fig. ). However, there are many different solutions to solve this problem. For example, Haack et al. identified the artifact on MR images and correlated it to the dwell position by performing a registration with a CT study set obtained during the commissioning process. Other authors have suggested a method in which MRI markers (A-vitamin pellets)—which produce a high signal on MRI images—are embedded into the perineal template to identify a reference plane on these images. The direction and free length distance of the needles are used to obtain the needle tip coordinates (Fig. ). Rigid applicators commonly used in BT may be modelled into a TPS applicator library. Otal et al. developed a method to incorporate interstitial components into the TPS gynaecologic rigid applicator library. As mentioned above, a precise source path determination is required during the reconstruction process. In some situations, an additional series of images may be acquired—such as T1W, 3D-MRI or 3D Echo SPGR sequences—to improve applicator visibility. Registration accuracy between T2W and the other type of images must be assessed.
Traditionally, perineal templates for interstitial cervix BT were based on stainless-steel needles. In the mid-1980s, Martinez et al. developed the MUPIT applicator; however, that applicator was not MR-compatible due to the use of stainless-steel needles. This meant that reconstruction had to be performed on CT-based study sets, which unfortunately, results in poor soft tissue contrast. Due to this poor contrast, together with image artifacts caused by the metallic needles, target visualisation is poor. In turn, this impedes precise delineation of the PTV. In these cases, a preplanning CT study set is useful. First, a CT study set of the patient is acquired and the perineal template is superimposed to the axial CT images to select the most appropriate location of the needles to ensure optimal target coverage. The most difficult part of this reconstruction is to precisely determine the tip location, and thus the first source dwell position. The accuracy of this determination is directly related to the slice thickness in the study set. Some treatment planning systems (TPS) have tools to perform “fine” navigation between slices. The scout view may be useful to precisely determine the needle tip localisation. In the cranio-caudal direction, the scout view has no deformation, thus allowing the user to visualise the needle tip to correlate it with the corresponding axial image (Fig. ).
The use of MRI images for perineal ISBT procedures has several major advantages, including good soft tissue contrast and spatial resolution (3D sequences with isotropic voxel size ≈ 1 mm). Nevertheless, several issues must be considered when using these images . First, the applicators must be MRI compatible. Generally, perineal BT procedures are performed with titanium needles, which can be problematic in MRI images due to geometric distortion and magnetic susceptibility artifacts of the titanium needles. Nevertheless, several authors have developed strategies to overcome these drawbacks . In addition, geometric distortion must be assessed for the same type of sequences used clinically to obtain images suitable for planning. Commercial MRI distortion phantoms are available for these purposes. The signal of titanium needles on MRI images shows a “ballooning” artifact at the tip of the needle, making it difficult to determine the exact distal dwell position (Fig. ). However, there are many different solutions to solve this problem. For example, Haack et al. identified the artifact on MR images and correlated it to the dwell position by performing a registration with a CT study set obtained during the commissioning process. Other authors have suggested a method in which MRI markers (A-vitamin pellets)—which produce a high signal on MRI images—are embedded into the perineal template to identify a reference plane on these images. The direction and free length distance of the needles are used to obtain the needle tip coordinates (Fig. ). Rigid applicators commonly used in BT may be modelled into a TPS applicator library. Otal et al. developed a method to incorporate interstitial components into the TPS gynaecologic rigid applicator library. As mentioned above, a precise source path determination is required during the reconstruction process. In some situations, an additional series of images may be acquired—such as T1W, 3D-MRI or 3D Echo SPGR sequences—to improve applicator visibility. Registration accuracy between T2W and the other type of images must be assessed.
Intracavitary/interstitial hybrid HDR-BT Due to the rapid technological advances in the treatment of cervical cancer, currently there are no unified recommendations on how to optimise BT. The GEC-ESTRO gynaecology working group announced at the World Congress of Brachytherapy in 2021 that they intend to publish planning recommendations in 2022. However, pending publication of those guidelines, the criteria followed may vary from centre to centre, generally based on the centre’s experience together with published data from reference centres and other studies. The protocols followed by Spanish radiation oncology centres are described below, as reflected in the paper published in 2018 by Perez-Calatayud et al. on behalf of SEFM . Before discussing planning tips, it is important to mention that the participating centres in the study of SEFM agree that inverse optimisation should not routinely be used for this type of implant due to loss of control of the assigned times and because, in some situations, while the dosimetric parameters may be adequate, the implant may be excessively heterogeneous. For these reasons, manual methods are usually recommended. The isodose lines in the three views of the implant should be reviewed. Activation of dwell positions The step size should be based on the source size and activity. If the step size is too short, it could result in insufficient stopping times, which could increase uncertainty and ultimately cause deviations from the planned dosimetry. The activation of positions should refer to the VOIs, encompassing the entire HR-CTV plus 5 mm. If any OARs are located near active positions, deactivation of those positions should be considered. Selection of the margins for activation and the final active positions must be confirmed by the radiation oncologist. Optimisation of dwell positions This is the part of the planning process in which the greatest differences among centres can be observed. Two very different optimisation methods are described below for case in which a discrete number of needles are used. Later in this document, we describe other methods for use when the principal component of the implant is interstitial. In this first method, points are generated on the surface of the HR-CTV and the plan is normalised to these points, initially without optimisation (i.e., all dwell times are equal). Next, the stopping times are modified through manual optimisation to meet the criteria established at the treatment centre (e.g., maximum volume allowed for 200% [V200%], V150%, rectal and bladder doses, and HR-CTV and CTV-IR coverage). In the second method, planning is started with standard loading of the intracavitary component and normalisation to the A points (as established in the ABS 2012 recommendations) . Stopping times are modified to meet the dosimetric criteria for OARs at the treating centre. Next, positions in the needles are activated inside the HR-CTV (taking into account potential OAR hotspots) and short times are assigned (1–4 s) in an effort to maintain the same time at all positions. According to the initial published recommendations, the total time assigned to the needles should be < 20% of the total treatment time; however, this restriction has lost importance over time, as the benefits of increasing the IS/IC time ratio has become increasingly clear . Another point to keep in mind is that recommendations for the current dosimetric parameters have mostly been described for their evaluation in EQD2 (equivalent dose in 2 Gy fractions), by adding EBRT and BT dose. To our knowledge, at present (October 2022), none of the currently available TPS allow for dose evaluation in terms of EQD2 (nor for adding doses with EBRT). Consequently, medical physicists must use spreadsheets to carry out this dose assessment requirement. Finally, it is important to emphasise that optimisation must be carried out in close collaboration with the radiation oncologist, especially evaluation of the EQD2 and dose distribution. Perineal templates Activation of dwell positions Activated dwell positions must assure adequate CTV coverage, which is why dwell positions 3–5 mm beyond the CTV are often activated. In this regard, it is important to keep in mind that the metallic needles (titanium or stainless steel) have a blind end about 1 cm in lenght, which should be considered when inserting the needles. Many planning systems have tools to automatically activate dwell positions to cover the CTV with an adequate margin. The source step is 2–3 mm. Three primary strategies are available for optimisation: (1) inverse planning, (2) point-based optimisation, and (3) geometric optimisation. The experience with the inverse planning modules included in the TPS in these types of implants, dose distributions obtained via inverse planning are similar to those obtained with other methods in terms of target coverage, but more heterogeneous. A points-based optimisation consists of generating a mesh of points on the surface of the CTV (target points) and then prescribing the nominal dose to these points. However, when using this type of optimisation, it is not uncommon to find hot spots within the CTV. In our experience, the best way to control these hot spots is to use geometric optimisation, which results in small hot spots at the needle boundaries that do not overlap between adjacent needles. Small manual (graphical) adjustments are performed to assure adequate coverage of the CTV (Fig. ). Interstitial The implant dwell positions should be selected on the clinical target. Initially, plan optimisation can be performed using automated tools in the TPS, followed by manual optimisation if necessary. Alternatively, manual optimisation can be performed without the need for the automated tools. The dose to 90% of the CTV volume should meet the prescription goals if the CTV is contoured. Otherwise, normal tissue dosimetry should include descriptions of the doses to volumes such as 0.1 cm 3 , 1 cm 3 , and 2 cm 3 of the bladder, urethra, rectum, sigmoid colon, and small bowel, depending on the location of the lesion. Summation of the EBRT and BT doses can be performed using EQD2 (equivalent dose in 2 Gy fractions) with α/β of 10 Gy for tumour and 3 Gy for normal tissues. During optimisation, the dwell times should be reviewed to ensure that there are no unexpectedly high dwell times. The volume of tissues receiving > 150% of the prescription dose should be limited to the area near the interstitial needles. The use of quality indices to assess conformality and homogeneity is recommended. The conformity index (CI) is defined as (CTVref/VCTV) (CTVref/Vref) where CTVref is the CTV volume receiving a dose equal to or greater than the reference dose, VCTV is the CTV volume, and Vref is the volume receiving a dose equal to or greater than the reference dose. The homogeneity index (HI) is defined as the fraction receiving a dose between 100 and 150% of the reference dose .
Due to the rapid technological advances in the treatment of cervical cancer, currently there are no unified recommendations on how to optimise BT. The GEC-ESTRO gynaecology working group announced at the World Congress of Brachytherapy in 2021 that they intend to publish planning recommendations in 2022. However, pending publication of those guidelines, the criteria followed may vary from centre to centre, generally based on the centre’s experience together with published data from reference centres and other studies. The protocols followed by Spanish radiation oncology centres are described below, as reflected in the paper published in 2018 by Perez-Calatayud et al. on behalf of SEFM . Before discussing planning tips, it is important to mention that the participating centres in the study of SEFM agree that inverse optimisation should not routinely be used for this type of implant due to loss of control of the assigned times and because, in some situations, while the dosimetric parameters may be adequate, the implant may be excessively heterogeneous. For these reasons, manual methods are usually recommended. The isodose lines in the three views of the implant should be reviewed. Activation of dwell positions The step size should be based on the source size and activity. If the step size is too short, it could result in insufficient stopping times, which could increase uncertainty and ultimately cause deviations from the planned dosimetry. The activation of positions should refer to the VOIs, encompassing the entire HR-CTV plus 5 mm. If any OARs are located near active positions, deactivation of those positions should be considered. Selection of the margins for activation and the final active positions must be confirmed by the radiation oncologist. Optimisation of dwell positions This is the part of the planning process in which the greatest differences among centres can be observed. Two very different optimisation methods are described below for case in which a discrete number of needles are used. Later in this document, we describe other methods for use when the principal component of the implant is interstitial. In this first method, points are generated on the surface of the HR-CTV and the plan is normalised to these points, initially without optimisation (i.e., all dwell times are equal). Next, the stopping times are modified through manual optimisation to meet the criteria established at the treatment centre (e.g., maximum volume allowed for 200% [V200%], V150%, rectal and bladder doses, and HR-CTV and CTV-IR coverage). In the second method, planning is started with standard loading of the intracavitary component and normalisation to the A points (as established in the ABS 2012 recommendations) . Stopping times are modified to meet the dosimetric criteria for OARs at the treating centre. Next, positions in the needles are activated inside the HR-CTV (taking into account potential OAR hotspots) and short times are assigned (1–4 s) in an effort to maintain the same time at all positions. According to the initial published recommendations, the total time assigned to the needles should be < 20% of the total treatment time; however, this restriction has lost importance over time, as the benefits of increasing the IS/IC time ratio has become increasingly clear . Another point to keep in mind is that recommendations for the current dosimetric parameters have mostly been described for their evaluation in EQD2 (equivalent dose in 2 Gy fractions), by adding EBRT and BT dose. To our knowledge, at present (October 2022), none of the currently available TPS allow for dose evaluation in terms of EQD2 (nor for adding doses with EBRT). Consequently, medical physicists must use spreadsheets to carry out this dose assessment requirement. Finally, it is important to emphasise that optimisation must be carried out in close collaboration with the radiation oncologist, especially evaluation of the EQD2 and dose distribution.
Activation of dwell positions Activated dwell positions must assure adequate CTV coverage, which is why dwell positions 3–5 mm beyond the CTV are often activated. In this regard, it is important to keep in mind that the metallic needles (titanium or stainless steel) have a blind end about 1 cm in lenght, which should be considered when inserting the needles. Many planning systems have tools to automatically activate dwell positions to cover the CTV with an adequate margin. The source step is 2–3 mm. Three primary strategies are available for optimisation: (1) inverse planning, (2) point-based optimisation, and (3) geometric optimisation. The experience with the inverse planning modules included in the TPS in these types of implants, dose distributions obtained via inverse planning are similar to those obtained with other methods in terms of target coverage, but more heterogeneous. A points-based optimisation consists of generating a mesh of points on the surface of the CTV (target points) and then prescribing the nominal dose to these points. However, when using this type of optimisation, it is not uncommon to find hot spots within the CTV. In our experience, the best way to control these hot spots is to use geometric optimisation, which results in small hot spots at the needle boundaries that do not overlap between adjacent needles. Small manual (graphical) adjustments are performed to assure adequate coverage of the CTV (Fig. ).
The implant dwell positions should be selected on the clinical target. Initially, plan optimisation can be performed using automated tools in the TPS, followed by manual optimisation if necessary. Alternatively, manual optimisation can be performed without the need for the automated tools. The dose to 90% of the CTV volume should meet the prescription goals if the CTV is contoured. Otherwise, normal tissue dosimetry should include descriptions of the doses to volumes such as 0.1 cm 3 , 1 cm 3 , and 2 cm 3 of the bladder, urethra, rectum, sigmoid colon, and small bowel, depending on the location of the lesion. Summation of the EBRT and BT doses can be performed using EQD2 (equivalent dose in 2 Gy fractions) with α/β of 10 Gy for tumour and 3 Gy for normal tissues. During optimisation, the dwell times should be reviewed to ensure that there are no unexpectedly high dwell times. The volume of tissues receiving > 150% of the prescription dose should be limited to the area near the interstitial needles. The use of quality indices to assess conformality and homogeneity is recommended. The conformity index (CI) is defined as (CTVref/VCTV) (CTVref/Vref) where CTVref is the CTV volume receiving a dose equal to or greater than the reference dose, VCTV is the CTV volume, and Vref is the volume receiving a dose equal to or greater than the reference dose. The homogeneity index (HI) is defined as the fraction receiving a dose between 100 and 150% of the reference dose .
Cervix intracavitary/interstitial hybrid HDR-BT Currently, the most common treatment scheme is the one described in the EMBRACE study: 4 fractions of 7 Gy (two implants of 2 fractions each), administered at weeks 6 and 7 of the treatment. Alternatively, this can be administered weekly, with four implants of one fraction each. As noted in clinical guidelines, other schemes have also been used (e.g., 5 fractions of 6 Gy) . Dose recommendations are based on clinical results of the retroEMBRACE and EMBRACE studies. These two large studies (> 2000 patients) have provided sufficient data to make dose recommendations to the GTV and HR-CTV that correlate with local control. Numerous reports have described the correlation between OAR doses and treatment-related morbidity. In most Spanish centers, we the dose recommendations, including the recommended limits (minimum doses for the GTV and HR-CTV; maximum for OARs) and the optimal doses from EMBRACE II PROTOCOL ( https://www.embracestudy.dk ) . For the sigmoid and bowel structures, these dose constraints are valid in case of non-mobile bowel loops resulting in the situation that the most exposed volume is located at a similar part of the organ Cervix perineal templates Most published recommendations and guidelines—including the those from the ABS—recommend administering five or more fractions . Nonetheless, the dose and fractionation scheme are highly dependent on the treating centre’s experience. The schemes reported to date include the following: 3.5 Gy × 9 fractions (fr); 4.25 Gy × 7 fr; 5 Gy × 5 fr; 3 Gy × 9 fr; 4.5 Gy × 5 fr; and 4 Gy × 6 fr. Given the lack of standardisation, the fractionation schedule must be combined with the EBRT dose using the EQD2 formula. Primary vaginal malignancies/vaginal recurrences Most consensus statements recommend that clinicians to determine the prescription based on their previous experience in cervical and vaginal cancer. Doses should be converted to EQD2 using the linear quadratic model (α/β ratio of 10 Gy for the target, and 3 Gy for OARs). For pulsed dose rate (PDR) BT, a repair half-time of 1.5 h should be applied for both the target and the OARs. Although precise recommendations for dose and fractionation have not yet been defined, some authors suggest that the combined total dose (EBRT plus BT) to the residual tumour should be > 70 Gy . The latest report of the GYN GEC-ESTRO vaginal cancer task group reported higher local control rates in large tumours when the dose above 80 Gy. In terms of dose/fractionation and number of applications, most centres perform a single application with 2–3 fractions of 6–7 Gy each (median CTV dose: 79 Gy). In cases with vaginal recurrence in a previously irradiated area, reirradiation may be possible, but the treatment must be individualised. Unfortunately, no recommendations are currently available in this clinical setting, and the only available published data are retrospective. In 2020, the ABS published a literature review on reirradiation in gynaecologic cancers, showing that HDR doses > 40 Gy achieved reasonably good local control and toxicity outcomes. In terms of dose and fractionation, the scheme used in many centres is 4–6 Gy in 5–10 fractions, administered twice daily. To our knowledge, no reports have been published to date on the use of PDR in this setting . Interstitial vulvar BT Currently, several different fractionation schemes are available for interstitial vulvar BT. However, all of the current data are based on retrospective studies (3, 51–53). As mentioned above, doses should be reported after conversion into EQD2 of 2 Gy/fraction (α/β 10 Gy for the tumour, half-time of 1.5 h). When BT is administered as an adjuvant treatment (i.e., boost) after EBRT, we recommend a dose of 18–21 Gy (6–7 fractions of 3 Gy/ twice daily), which is equivalent to an EQD2 Gy: α/β10/3 19.5/21.6–23.6/27.3. A five-fraction schedule of 3.5 Gy twice daily can also be used. If adjuvant ISBT is administered alone, the recommended dose is 40.5 Gy (9 fractions of 4.5 Gy each, twice daily). When ISBT is used to deliver a boost to the primary vulvar tumour after EBRT, we recommend a dose of 21–24 Gy (7–8 fractions of 3 Gy/twice daily), which is equivalent to EQD2 Gy: α/β10/3 22.7/25.2–26/28.8. Alternatively, a schedule of 6–7 fractions of 3.5 Gy or 5 fractions of 4 Gy (both administered twice daily) can also be used. If BT is delivered as monotherapy, the recommended dose is 45 Gy (10 fractions of 4.5 Gy/ twice daily). The volumes receiving 90% (V90), 100% (V100), 150% (V150), and 200% (V200) of the prescribed dose should be reported. For postoperative recurrences, the recommendations are the same as in de novo tumours . In recurrences after previous irradiation, it is essential to personalise the treatment approach, especially given the scant published data in this clinical setting . The total dose (EBRT or BT alone, or combined treatment) should be expressed in radiobiologically-equivalent doses of 2 Gy/fraction; always taking into account the 2 Gy biologically-effective dose (BED) received for the previous treatment and the time elapsed between the two treatments. The most important OAR is the urethra. The recommended approach to evaluating the dose administered dose to this organ is to assess D2cc and D0.1 cc based on the dose-volume histogram derived from the 3D dose distribution. Other OARs that should be considered are the anus and the clitoris. Given the lack of concrete data in the literature, the ALARA criteria ( as low as reasonably achievable ) should be followed.
Currently, the most common treatment scheme is the one described in the EMBRACE study: 4 fractions of 7 Gy (two implants of 2 fractions each), administered at weeks 6 and 7 of the treatment. Alternatively, this can be administered weekly, with four implants of one fraction each. As noted in clinical guidelines, other schemes have also been used (e.g., 5 fractions of 6 Gy) . Dose recommendations are based on clinical results of the retroEMBRACE and EMBRACE studies. These two large studies (> 2000 patients) have provided sufficient data to make dose recommendations to the GTV and HR-CTV that correlate with local control. Numerous reports have described the correlation between OAR doses and treatment-related morbidity. In most Spanish centers, we the dose recommendations, including the recommended limits (minimum doses for the GTV and HR-CTV; maximum for OARs) and the optimal doses from EMBRACE II PROTOCOL ( https://www.embracestudy.dk ) . For the sigmoid and bowel structures, these dose constraints are valid in case of non-mobile bowel loops resulting in the situation that the most exposed volume is located at a similar part of the organ
Most published recommendations and guidelines—including the those from the ABS—recommend administering five or more fractions . Nonetheless, the dose and fractionation scheme are highly dependent on the treating centre’s experience. The schemes reported to date include the following: 3.5 Gy × 9 fractions (fr); 4.25 Gy × 7 fr; 5 Gy × 5 fr; 3 Gy × 9 fr; 4.5 Gy × 5 fr; and 4 Gy × 6 fr. Given the lack of standardisation, the fractionation schedule must be combined with the EBRT dose using the EQD2 formula.
Most consensus statements recommend that clinicians to determine the prescription based on their previous experience in cervical and vaginal cancer. Doses should be converted to EQD2 using the linear quadratic model (α/β ratio of 10 Gy for the target, and 3 Gy for OARs). For pulsed dose rate (PDR) BT, a repair half-time of 1.5 h should be applied for both the target and the OARs. Although precise recommendations for dose and fractionation have not yet been defined, some authors suggest that the combined total dose (EBRT plus BT) to the residual tumour should be > 70 Gy . The latest report of the GYN GEC-ESTRO vaginal cancer task group reported higher local control rates in large tumours when the dose above 80 Gy. In terms of dose/fractionation and number of applications, most centres perform a single application with 2–3 fractions of 6–7 Gy each (median CTV dose: 79 Gy). In cases with vaginal recurrence in a previously irradiated area, reirradiation may be possible, but the treatment must be individualised. Unfortunately, no recommendations are currently available in this clinical setting, and the only available published data are retrospective. In 2020, the ABS published a literature review on reirradiation in gynaecologic cancers, showing that HDR doses > 40 Gy achieved reasonably good local control and toxicity outcomes. In terms of dose and fractionation, the scheme used in many centres is 4–6 Gy in 5–10 fractions, administered twice daily. To our knowledge, no reports have been published to date on the use of PDR in this setting .
Currently, several different fractionation schemes are available for interstitial vulvar BT. However, all of the current data are based on retrospective studies (3, 51–53). As mentioned above, doses should be reported after conversion into EQD2 of 2 Gy/fraction (α/β 10 Gy for the tumour, half-time of 1.5 h). When BT is administered as an adjuvant treatment (i.e., boost) after EBRT, we recommend a dose of 18–21 Gy (6–7 fractions of 3 Gy/ twice daily), which is equivalent to an EQD2 Gy: α/β10/3 19.5/21.6–23.6/27.3. A five-fraction schedule of 3.5 Gy twice daily can also be used. If adjuvant ISBT is administered alone, the recommended dose is 40.5 Gy (9 fractions of 4.5 Gy each, twice daily). When ISBT is used to deliver a boost to the primary vulvar tumour after EBRT, we recommend a dose of 21–24 Gy (7–8 fractions of 3 Gy/twice daily), which is equivalent to EQD2 Gy: α/β10/3 22.7/25.2–26/28.8. Alternatively, a schedule of 6–7 fractions of 3.5 Gy or 5 fractions of 4 Gy (both administered twice daily) can also be used. If BT is delivered as monotherapy, the recommended dose is 45 Gy (10 fractions of 4.5 Gy/ twice daily). The volumes receiving 90% (V90), 100% (V100), 150% (V150), and 200% (V200) of the prescribed dose should be reported. For postoperative recurrences, the recommendations are the same as in de novo tumours . In recurrences after previous irradiation, it is essential to personalise the treatment approach, especially given the scant published data in this clinical setting . The total dose (EBRT or BT alone, or combined treatment) should be expressed in radiobiologically-equivalent doses of 2 Gy/fraction; always taking into account the 2 Gy biologically-effective dose (BED) received for the previous treatment and the time elapsed between the two treatments. The most important OAR is the urethra. The recommended approach to evaluating the dose administered dose to this organ is to assess D2cc and D0.1 cc based on the dose-volume histogram derived from the 3D dose distribution. Other OARs that should be considered are the anus and the clitoris. Given the lack of concrete data in the literature, the ALARA criteria ( as low as reasonably achievable ) should be followed.
The dosimetrist is involved in many steps of gynaecological BT. Some of the main responsibilities are as follows: Quality control of applicators and transfer tubes: Applicators: verify that there is no deformity and verify the distal positions and offset to ensure that their real behaviour is exactly as planned in the reconstruction. Transfer tubes: check that no folds or creases are present, and that the tubes properly connect to the indexer and applicator. Patient treatment: Active identification of the patient to ensure an exact match between the planned treatment session and the specific patient. Connect the transfer tubes to the applicators according to the dosimetric indications. For all fractions, check that the applicators are in perfect condition and shape and that the source can pass through the applicators without any risk. Create a comfortable and safe climate for the patient before and after the treatment to avoid radiological incidents and no patient movement during treatment delivery.
Conclusions of the Spanish survey: Of the centres that perform gynaecologic brachytherapy, two-thirds (67%) report using the interstitial technique, although all of the professionals surveyed recognise the importance of the interstitial component. Among the centres that employ an interstitial component in gynaecologic BT ( n = 24), the most common applications are for the cervix (96%), vagina (87.5%), vulva (75%), and relapses (81%). All of the centres (100%) have HDR. Two centres routinely use PDR. 81.5% of the centres use MRI for gynaecological BT planning, at least for the first fraction in the first implant. In 59.3% of centres, image control is taken before the next fraction for the same implant. Three centres (11.1%) use only one fraction per implant (i.e., a new implant is performed for each fraction). The most common applicator type is the Utrecht ( n = 18; 75%), but other applicators are also employed, including freehand needles, tubes, and personalised templates (see responses to question 5) Relapses: if the target area has not been previously irradiated (EBRT), all of the centres administer combined radiotherapy (EBRT and BT). For reirradiation, most centres (69.2%) use BT alone Fractionation: Slightly more than half ( n = 14/24; 58%) of centres use four fractions of 7 Gy, following GEC-ESTRO recommendations. Among the other centres, a wide range of different fractionation schemes are used. In recurrent disease, hen BT is used as monotherapy for reirradiation, the administered dose must be ≥ 40–50 Gy in 9–10 fractions, taking into account the EQD2 (see response to question 8). Planning: 42% of centres use the Manchester system (modified pear), and 54% use inverse planning. For interstitial vulvar implants, 37.5% of centres use the Paris system. After an in-depth discussion of the survey results, the working group reached the following consensus-based recommendations: In cervix BT, the interstitial component is important— even when the HR-CTV is small—to improve dose distribution, especially doses to the OARs. Centres that lack experience in treating large cervical tumours or other gynaecologic cancers (vagina, vulva, and/or recurrences) should refer patients to experienced, high volume centres. Ideally, MRI should be performed after completion of EBRT but prior to starting BT to assess residual disease and to perform preplanning ( especially in relapses) for IGBT. During BT implantation, we recommend ultrasound guidance, both transabdominal and transrectal/transvaginal (depending on the clinician’s experience), for image-guided adaptive BT. Fiducial markers are recommended, if appropriate, to mark the borders. Each centre should select the type of applicators they are most familiar with (Utrecht, Ring, etc.), without precluding the use of freehand needles or personalised templates. Before administration of the subsequent BT fraction, imaging (either CT or x-ray) should be performed for control purposes. In vaginal and vulvar tumours, interstitial needles/plastic tubes are recommended, together with perineal templates or freehand needles/tubes, in accordance with GEC-ESTRO recommendations. For the treatment of recurrent disease, if the target/tumour area has not previously received full EBRT doses, a small field (EBRT) should be applied to cover the macroscopic disease with margins. However, if the entire EBRT dose has been applied previously, including the full dose to the OAR, then we recommend administering BT alone. For cervical cancer, we recommend the GEC-ESTRO scheme: 45 Gy of EBRT + BT in 4 fractions of 7 Gy in two implants (to administer ≥ 85 Gy to D90 HR-CTV); if BT monotherapy is used for reirradiation, we recommend ≥ 40–50 Gy in 9–10 fractions (equivalent to EQD2 60 Gy). For cervical cancer planning, we recommend starting with the Manchester system (modified pear) and adding a weight of no more than 20% for the interstitial needles. Detailed recommendations developed by the SEFM are available (Pérez-Calatayud 2018). For inverse planning techniques, the late effects in hot spots have not been well-characterised. Therefore, caution is warranted. For pure interstitial implants, we recommend using the modified Paris System. We also recommend revising the source stopping times to ensure there are no great source steps.
Of the centres that perform gynaecologic brachytherapy, two-thirds (67%) report using the interstitial technique, although all of the professionals surveyed recognise the importance of the interstitial component. Among the centres that employ an interstitial component in gynaecologic BT ( n = 24), the most common applications are for the cervix (96%), vagina (87.5%), vulva (75%), and relapses (81%). All of the centres (100%) have HDR. Two centres routinely use PDR. 81.5% of the centres use MRI for gynaecological BT planning, at least for the first fraction in the first implant. In 59.3% of centres, image control is taken before the next fraction for the same implant. Three centres (11.1%) use only one fraction per implant (i.e., a new implant is performed for each fraction). The most common applicator type is the Utrecht ( n = 18; 75%), but other applicators are also employed, including freehand needles, tubes, and personalised templates (see responses to question 5) Relapses: if the target area has not been previously irradiated (EBRT), all of the centres administer combined radiotherapy (EBRT and BT). For reirradiation, most centres (69.2%) use BT alone Fractionation: Slightly more than half ( n = 14/24; 58%) of centres use four fractions of 7 Gy, following GEC-ESTRO recommendations. Among the other centres, a wide range of different fractionation schemes are used. In recurrent disease, hen BT is used as monotherapy for reirradiation, the administered dose must be ≥ 40–50 Gy in 9–10 fractions, taking into account the EQD2 (see response to question 8). Planning: 42% of centres use the Manchester system (modified pear), and 54% use inverse planning. For interstitial vulvar implants, 37.5% of centres use the Paris system.
In cervix BT, the interstitial component is important— even when the HR-CTV is small—to improve dose distribution, especially doses to the OARs. Centres that lack experience in treating large cervical tumours or other gynaecologic cancers (vagina, vulva, and/or recurrences) should refer patients to experienced, high volume centres. Ideally, MRI should be performed after completion of EBRT but prior to starting BT to assess residual disease and to perform preplanning ( especially in relapses) for IGBT. During BT implantation, we recommend ultrasound guidance, both transabdominal and transrectal/transvaginal (depending on the clinician’s experience), for image-guided adaptive BT. Fiducial markers are recommended, if appropriate, to mark the borders. Each centre should select the type of applicators they are most familiar with (Utrecht, Ring, etc.), without precluding the use of freehand needles or personalised templates. Before administration of the subsequent BT fraction, imaging (either CT or x-ray) should be performed for control purposes. In vaginal and vulvar tumours, interstitial needles/plastic tubes are recommended, together with perineal templates or freehand needles/tubes, in accordance with GEC-ESTRO recommendations. For the treatment of recurrent disease, if the target/tumour area has not previously received full EBRT doses, a small field (EBRT) should be applied to cover the macroscopic disease with margins. However, if the entire EBRT dose has been applied previously, including the full dose to the OAR, then we recommend administering BT alone. For cervical cancer, we recommend the GEC-ESTRO scheme: 45 Gy of EBRT + BT in 4 fractions of 7 Gy in two implants (to administer ≥ 85 Gy to D90 HR-CTV); if BT monotherapy is used for reirradiation, we recommend ≥ 40–50 Gy in 9–10 fractions (equivalent to EQD2 60 Gy). For cervical cancer planning, we recommend starting with the Manchester system (modified pear) and adding a weight of no more than 20% for the interstitial needles. Detailed recommendations developed by the SEFM are available (Pérez-Calatayud 2018). For inverse planning techniques, the late effects in hot spots have not been well-characterised. Therefore, caution is warranted. For pure interstitial implants, we recommend using the modified Paris System. We also recommend revising the source stopping times to ensure there are no great source steps.
The interstitial component in gynaecological BT plays an essential role in the treatment of cancers of the cervix, vagina and vulva, both in primary tumours and recurrent disease. Appropriate training is highly recommended.
|
A comparative analysis of public and private dental benefit payer types for the provision and outcomes of root canal therapy on permanent teeth of children and adolescents in Massachusetts | 30f1747a-1d82-4e7d-8af6-b2c94a4ed298 | 10026184 | Dental[mh] | Data sources We used data from the Center for Health Information and Analysis, a state agency of Massachusetts. This Massachusetts All-Payer Claims Database includes dental claims and information on member eligibility, provider, and insurance type, which are collected from health insurance payers licensed to operate in the Commonwealth of Massachusetts. Release Version 7.0 was inclusive for years 2013 through 2017 and contained records for 1,516,624 children aged 6 through 18 years; 539,966 (36%) were Medicaid beneficiaries, and 976,658 (64%) were enrolled in private insurance plans. This research is part of a data use agreement approved by the Center for Health Information and Analysis (1491238-1). This study was approved by New York University School of Medicine’s Institutional Review Board (i19-01436). Variables The analytic data set contained the following information for each patient: unique identification number, dates of enrollment and disenrollment, dental insurance payer type (private vs public or Medicaid), age at date of treatment, sex, ZIP Code of residence, date of treatment, procedure code for the treatment provided, tooth number treated, provider’s specialty code, and payment information for the procedures (that is, the maximum amount contractually allowed and that an insurer will pay for the procedure or the amount paid by the insurer). Our study included children and adolescents aged 6 through 18 years who had undergone initial root canal therapy on a permanent tooth. The term initial denotes the first observed root canal treatment for a patient in the cohort for a particular tooth. Codes from the American Dental Association’s CDT 2017: Current Dental Terminology (CDT) were used to identify the endodontic treatment procedures for analysis (D3310, D3320, D3330). Furthermore, CDT codes were used to identify an adverse event after initial root canal therapy. Adverse events were defined as nonsurgical endodontic retreatment (D3346, D3347, D3348), surgical endodontic retreatment or apicoectomy (D3410, D3421, D3425), or tooth extraction (D7140, D7210), and indicated failure of the initial root canal therapy. Provider specialty codes were used to determine the provider type, either individual providers or facilities. Individual providers included general (nonspecialist) dentists and specialist dentists. Dental specialists included endodontists (specialists in performing root canal therapy), pediatric dentists, prosthodontists, periodontists, orthodontists, and oral surgeons. Individual providers were dichotomized into endodontist and other individual providers for the analyses. Facilities included provider specialty codes identified as clinic or center and hospitals. Statistical analyses Data analysis was completed using SAS software, Version 9.4 (SAS Institute) and R, Version 4.0 (R Core Team). Provision of initial root canal therapy was analyzed at the person level. Patients who received initial root canal therapy within the study period of 2013 through 2017 and who had complete payer-type data were included in the final data set. Treatment outcomes of initial rootcanal therapy were measured at the tooth level. For tooth-level analysis, patients were excluded if they had less than 1 year of insurance enrollment after treatment or records of treatment were missing the associated tooth number. To ensure external validity, χ 2 tests and t tests were used to compare characteristics between those excluded due to missing tooth number and those included in the sample. Multiple logistic regression was used to measure the association between payer-type and initial root canal therapy at the person level, with adjustments for age, sex, tooth type (anterior, premolar, molar), and provider type. For the evaluation of procedural outcomes at the tooth level, initial root canal therapies were considered successful unless there was an adverse event (tooth extraction, endodontic retreatment, apicoectomy) or they were censored at an identified lapse in the patient’s insurance enrollment. The Kaplan-Meier method was used to estimate procedural survival according to payer type. Cox proportional hazard regression was used to evaluate the hazard of adverse event occurrence after initial root canal therapy, according to payer type (public or private), using a model that adjusted for the following covariates: age, sex, tooth type, and provider type. Member identification was also included in this model to account for person-level clustering and to generate estimates with robust SEs. In addition, this methodology was used to assess procedural survival according to provider type in a subset of the sample limited to procedures completed by an individual provider, with the adjusted Cox proportional hazard model controlling for age, sex, tooth type, and payer type. The payment data (amounts allowed and paid via the insurer) for initial root canal therapy were summarized according to both payer and tooth type. Wilcoxon rank-sum tests were used to test for the statistical significance of differences in paid and allowed amounts according to payer type for each tooth type (that is, anterior, premolar, and molar). Statistical significance level ( α ) of .05 was used for all analyses.
We used data from the Center for Health Information and Analysis, a state agency of Massachusetts. This Massachusetts All-Payer Claims Database includes dental claims and information on member eligibility, provider, and insurance type, which are collected from health insurance payers licensed to operate in the Commonwealth of Massachusetts. Release Version 7.0 was inclusive for years 2013 through 2017 and contained records for 1,516,624 children aged 6 through 18 years; 539,966 (36%) were Medicaid beneficiaries, and 976,658 (64%) were enrolled in private insurance plans. This research is part of a data use agreement approved by the Center for Health Information and Analysis (1491238-1). This study was approved by New York University School of Medicine’s Institutional Review Board (i19-01436).
The analytic data set contained the following information for each patient: unique identification number, dates of enrollment and disenrollment, dental insurance payer type (private vs public or Medicaid), age at date of treatment, sex, ZIP Code of residence, date of treatment, procedure code for the treatment provided, tooth number treated, provider’s specialty code, and payment information for the procedures (that is, the maximum amount contractually allowed and that an insurer will pay for the procedure or the amount paid by the insurer). Our study included children and adolescents aged 6 through 18 years who had undergone initial root canal therapy on a permanent tooth. The term initial denotes the first observed root canal treatment for a patient in the cohort for a particular tooth. Codes from the American Dental Association’s CDT 2017: Current Dental Terminology (CDT) were used to identify the endodontic treatment procedures for analysis (D3310, D3320, D3330). Furthermore, CDT codes were used to identify an adverse event after initial root canal therapy. Adverse events were defined as nonsurgical endodontic retreatment (D3346, D3347, D3348), surgical endodontic retreatment or apicoectomy (D3410, D3421, D3425), or tooth extraction (D7140, D7210), and indicated failure of the initial root canal therapy. Provider specialty codes were used to determine the provider type, either individual providers or facilities. Individual providers included general (nonspecialist) dentists and specialist dentists. Dental specialists included endodontists (specialists in performing root canal therapy), pediatric dentists, prosthodontists, periodontists, orthodontists, and oral surgeons. Individual providers were dichotomized into endodontist and other individual providers for the analyses. Facilities included provider specialty codes identified as clinic or center and hospitals.
Data analysis was completed using SAS software, Version 9.4 (SAS Institute) and R, Version 4.0 (R Core Team). Provision of initial root canal therapy was analyzed at the person level. Patients who received initial root canal therapy within the study period of 2013 through 2017 and who had complete payer-type data were included in the final data set. Treatment outcomes of initial rootcanal therapy were measured at the tooth level. For tooth-level analysis, patients were excluded if they had less than 1 year of insurance enrollment after treatment or records of treatment were missing the associated tooth number. To ensure external validity, χ 2 tests and t tests were used to compare characteristics between those excluded due to missing tooth number and those included in the sample. Multiple logistic regression was used to measure the association between payer-type and initial root canal therapy at the person level, with adjustments for age, sex, tooth type (anterior, premolar, molar), and provider type. For the evaluation of procedural outcomes at the tooth level, initial root canal therapies were considered successful unless there was an adverse event (tooth extraction, endodontic retreatment, apicoectomy) or they were censored at an identified lapse in the patient’s insurance enrollment. The Kaplan-Meier method was used to estimate procedural survival according to payer type. Cox proportional hazard regression was used to evaluate the hazard of adverse event occurrence after initial root canal therapy, according to payer type (public or private), using a model that adjusted for the following covariates: age, sex, tooth type, and provider type. Member identification was also included in this model to account for person-level clustering and to generate estimates with robust SEs. In addition, this methodology was used to assess procedural survival according to provider type in a subset of the sample limited to procedures completed by an individual provider, with the adjusted Cox proportional hazard model controlling for age, sex, tooth type, and payer type. The payment data (amounts allowed and paid via the insurer) for initial root canal therapy were summarized according to both payer and tooth type. Wilcoxon rank-sum tests were used to test for the statistical significance of differences in paid and allowed amounts according to payer type for each tooth type (that is, anterior, premolar, and molar). Statistical significance level ( α ) of .05 was used for all analyses.
The final study sample included 27,217 children aged 6 through 18 years who had undergone initial root canal therapy in a permanent tooth from 2013 through 2017 ( ). There were 9,202 children with private insurance who had undergone 11,874 initial root canal therapies. Medicaid and CHIP beneficiaries included 17,985 children who had undergone 25,627 initial root canal therapies. Mean (SD) number of initial root canal therapies per patient was 1.38 (0.94) (median, 1; interquartile range [IQR], 1-1; range, 1-25). Children and adolescents with public dental insurance were more likely to have undergone more than 1 initial root canal therapy (adjusted odds ratio, 1.91; 95% CI, 1.73 to 2.11) ( ). Median age at time of initial root canal therapy was 15 years (IQR, 13-17 years). Molars were the most frequently treated tooth type regardless of payer type ( ). Dental care provider types for the initial root canal therapy procedures are reported for the first treatment each patient received ( ). Children with public-payer insurance were more likely than those with private insurance to have had their endodontic treatment performed in a facility setting ( P < .001). Dental care providers, classified as individual providers performed 12,507 (46%) of the initial root canal therapies, including nonspecialist dentists (51.2%), endodontists (46.2%), pediatric dentists (1%), and other dental specialists (1%). Children with private insurance were more likely to have an endodontist perform their root canal therapy ( P < .0001). After excluding treatments with missing tooth numbers, the sample included 1,497 private-payer beneficiaries having undergone 1,654 initial root canal therapies and 13,523 public-payer beneficiaries having undergone 18,434 initial root canal therapies. Between those included in the sample and those excluded due to missing tooth number, statistically significant differences were found according to sex and provider type, but not age and tooth type ( P < .0001). Median follow-up period was 33 months (IQR, 22-43 months; range, 12-59 months) for the privately insured and 34 months (IQR, 22-46 months; range, 12-59 months) for public-payer beneficiaries. The adjusted Cox proportional hazards models in indicates that public-payer beneficiaries had a greater hazard of experiencing an adverse event (adjusted hazard ratio [HR], 2.19; 95% CI, 1.53 to 3.14). In the provider-type analysis, patients who received treatment from endodontists had superior outcomes relative to patients who received treatment from other dentists in the unadjusted model (HR, 0.34; 95% CI, 0.12 to 0.92). However, after adjusting for covariates, the statistical significance of this association did not hold (adjusted HR, 0.41; 95% CI, 0.13 to 1.22). When we evaluated procedural survival rates according to payer type, a log-rank test revealed statistically significant differences between the public and privately insured patients ( P < .0001; ). A sensitivity analysis of the Kaplan-Meier survival estimates and adjusted HRs on cases censored according to eligibility did not reveal any statistically or clinically significant differences in treatment outcomes. All 20,088 initial root canal therapies were evaluated at 1 year of follow-up; survival rates were 98.0% (95% CI, 97.8% to 98.2%) among public beneficiaries and 99.2% among those privately insured (95% CI, 98.8% to 99.6%). At 3 years follow-up, 9,433 procedures could be evaluated; survival rates were 94.0% (95% CI, 93.6% to 94.4%) among public beneficiaries and 97.4% among privately insured children (95% CI, 96.5% to 98.3%). For initial root canal therapies that failed during the study period (n = 1,295 [6.4%]), the first identified adverse event was tooth extraction in 78% of cases (n = 1,016) and surgical or nonsurgical retreatment in 21.5% of cases (n = 279). No statistically significant difference was detected for type of adverse events according to payer type ( P = .721). Reimbursement amounts for procedures according to payer and tooth type, as indicated by CDT code (D3310 for anterior, D3320 for premolar, D3330 for molar) for the first treatment each patient received are reported in . Amounts allowed and paid by the insurer were significantly higher for those privately insured, regardless of tooth type ( P < .001).
Federal efforts to expand public dental insurance coverage have reduced disparities in dental visit use according to payer type for children and adolescents with public and private dental insurance. – , , However, differences in the dental procedure mix for children and adolescents are prevalent; public-payer beneficiaries are more likely to undergo a higher share of therapeutic dental services, including endodontic treatment. , , Our findings support those of investigators reporting that pediatric public-payer beneficiaries receive more endodontic treatment, including root canal therapy, than their counterparts with private dental coverage. , , Although the survival rates of initial root canal therapy performed on the permanent teeth of children and adolescents were high for both groups at 1 and 3 years postoperatively, our findings indicate potential disparities in treatment outcomes, as beneficiaries of public dental insurance were more likely to experience procedural failures after root canal therapy than those with private-payer insurance. Our study was subject to some limitations. First, there was a large number of missing data related to tooth number in the private-payer cohort. Tooth number was required for the treatment outcome analysis at the tooth level, but not the analysis of the provision of treatment at the person level. After eliminating cases that were missing the tooth number, 71% of the otherwise eligible cohort of private-payer procedures were lost. This differential loss of cases according to payer type may increase the possibility of selection bias for analysis of treatment outcomes according to payer type and limit generalizability to all children and adolescents in Massachusetts. We considered approaches to overcome the missing tooth number issue for the tooth-level analysis, such as assuming that any subsequent adverse event procedure that followed the initial root canal therapy (that is, extraction, endodontic retreatment, or apicoectomy) was an indication of procedural failure. However, in the end, we could not apply this less conservative analytic approach because tooth extraction is the most common adverse event observed, and the CDT code that identifies this procedure does not include tooth type (that is, anterior, premolar, or molar), as is done for endodontic treatments. Without the tooth number, there was too high a risk that a downstream, unrelated tooth extraction may be incorrectly associated with the initial root canal therapy. Another factor potentially limiting the generalizability of our findings is that we have included data from only 1 state, Massachusetts. However, the findings for public-payer (Medicaid, CHIP) beneficiaries in our study were comparable with findings from a cohort of pediatric Medicaid beneficiaries from New York. To our knowledge, a comparable study does not yet exist for a cohort of privately insured children. Second, as in all studies using dental claims data, the initial diagnosis for the treated teeth could not be determined due to the nonclinical nature of the administrative data and the absence of diagnostic codes in dentistry. This is particularly relevant to our study, as endodontic diagnosis is known to be associated with endodontic treatment outcomes. , Despite these limitations, our study results provide insights regarding access to oral health care for children with public or private dental insurance in the Commonwealth of Massachusetts related to the prevalence of endodontic treatment and access to dental care providers and treatment settings. The results of our study, particularly the realized differences in the provision of root canal therapy, may reflect existing disparities in oral health status between public and private-payer beneficiaries. , Realized disparities in procedural outcomes of root canal therapy according to payer type may also be attributable to differences in treatment setting or dentist provider types. In previous studies using similar methodology, root canal therapies performed by endodontists have been associated with improved outcomes. , It was difficult to make this distinction in our study, as payer type and treatment setting or provider type were so closely associated ( P < .001) for the subset of the sample for which provider specialty codes could be identified: 94% of public-payer root canal therapy procedure claims included a provider specialty code associated with a facility (clinic, center, hospital) and 97% of private root canal therapy procedure claims included a provider specialty code identifying an individual dental care provider. After adjusting for covariates, including treatment setting, the disparity in treatment outcomes according to payer type was reduced, but still significant. Of the provider specialty codes associated with dental care providers, 51% were general dentists, 46% were identified as endodontists, and 1% were pediatric dentists. The most recent information available on the frequency of endodontic procedures is from the American Dental Association’s Survey of Dental Services Rendered , conducted from 2005 through 2006 and published in 2007, which reported that endodontists performed 25% of all endodontic procedures (adult and pediatric populations). Our findings, therefore, suggest that endodontists may be more likely to treat children requiring root canal therapy than other dental care providers. Nationally, 43% of all dentists, 73% of pediatric dentists, and only 28% of endodontists, participate in Medicaid for child dental services. In our study, none of the root canal therapy procedure claims for publicly insured children were associated with a provider specialty code for an endodontist as an individual dental care provider; however, it is possible that when treated in a facility setting, an endodontist was the provider. Facilities (clinics, centers, hospitals) are likely to be dental schools or federally qualified health centers and may employ endodontists or have postgraduate trainees specializing in endodontics as providers. When these facts are considered, one could question whether access to endodontic treatment is limited for pediatric patients enrolled in public insurance. Whether dentists participate as Medicaid providers is strongly associated with Medicaid reimbursement rates. – The findings of our study highlighted differences in both allowed and paid amounts for root canal therapy procedure claims billed by insurers, with the amounts allowed and paid by private payers higher than those by public payers. In Massachusetts in 2017, Medicaid reimbursement as a percentage of private insurance reimbursements for diagnostic and preventative dental services was 74%. We found that the reimbursement percentage was higher for the specific endodontic procedure of root canal therapy on premolars and molars. This may be due to the fact that endodontic treatment performed on premolars and molars generally have higher levels of case difficulty. In the same year (2017) in Massachusetts, 51% of dentists were enrolled as Medicaid providers, but only 26% of Massachusetts dentists had seen Medicaid patients in the past year. Future research aiming to increase understanding of dentists’ Medicaid participation, particularly for the provision of endodontic services to pediatric populations, will help inform strategies for increasing access to endodontic care for Medicaid enrollees. Surges in the percentage of children enrolled in Medicaid and CHIP during the past 15 years indicate a need for increased attention to not only access to oral health care, but also treatment outcomes for public-payer beneficiaries. ,
Statistically significant differences exist in the provision and outcomes of endodontic treatment between privately and publicly insured children and adolescents in the state of Massachusetts. Public payer beneficiaries were more likely to receive root canal therapy and have poorer treatment outcomes associated with the procedure than those enrolled in private-payer dental insurance plans. Differences in treatment setting, provider types, and payment amounts between public and private insurance are potential contributing factors to this realized disparity.
|
Optimising adherence to secondary prevention medications following acute coronary syndrome utilising telehealth cardiology pharmacist clinics: a matched cohort study | 691d2878-2faf-42f4-b40a-6aa35743d8a0 | 10026199 | Internal Medicine[mh] | Implementing cardiology pharmacist telehealth clinics leads to improved medication adherence, and by extension improved clinical outcomes in patients following an acute coronary syndrome. This model of care is now a permanent component of care for patients following discharge for an acute coronary syndrome across a large regional health service. This model is now being adapted in a rapid access atrial fibrillation clinic model to enhance anticoagulant and anti-arrhythmic management.
Cardiovascular disease represents a major burden on health worldwide. In 2016, 43,963 deaths were attributed to heart disease nationally, equating to 30% of all deaths . The number one cause of cardiovascular disease death and morbidity is acute coronary syndromes (ACS) . Developed parts of the world all share in the burden of cardiovascular disease, with similar rates of mortality and morbidity across Europe, North America and Australia . In Australia, due to its large degree of urbanisation, health disparities exist for regional Australians versus their metropolitan counterparts, where there is a higher rate of early deaths from cardiovascular disease compared to metropolitan areas . Advances in access to catheter laboratories and stent technology have all contributed to increased survival following acute coronary syndrome (ACS) events . However, medications remain the mainstay of treatment for secondary prevention of subsequent events . In particular, there are a number of medications directly linked with mortality benefit and reduction of major adverse cardiovascular events (MACE). These medications are referred to as optimal medical therapy (OMT) and consist of four groups: dual anti-antiplatelet therapy (DAPT), which includes aspirin and an adenosine diphosphate (P2Y12) inhibitor (clopidogrel, prasugrel, or ticagrelor); HMG Co-A reductase inhibitors (statins); beta blockers; and angiotensin converting enzyme inhibitors/angiotensin receptor blockers (ACEI/ARB). More recently, angiotensin receptor and neprilysin inhibitors (ARNI) have also been used in place of ACEI in patients with pre-existing heart failure with reduced ejection fraction (HFrEF) . While the randomised control trials (RCT) that have generated clinical outcomes data show that the presence of these medications reduces MACE, the conditions of those trials are not necessarily representative of real-world populations . For example, real-world statin studies often show non-adherence rates from 10 to 45%, with RCT data showing non-adherence rates of 1–2% . These observational, real-world studies provide insight into how these therapies are utilised within the population. A multicentre registry study of 19,704 patients in the United States of America showed that at 90 days post ACS, mean adherence to DAPT was 72%, beta blockers 63%, statin 63% and ACEI 64% . The link between adherence to therapy and clinical outcomes has also been researched extensively within the ACS population. An Australian registry study of 9,735 patients demonstrated that being on less than three secondary prevention medications (sub-optimal medical therapy) was an independent predictor of long-term mortality at 4 years, when compared to being on all OMT medications (8.2% vs. 16.8%, p < 0.001, n = 9375) . There are many factors associated with poor medication adherence, such as socioeconomic status, age, and health literacy . These aspects are also strong determinants of multi-morbidity and cardiovascular outcomes, with multidisciplinary interventions recommended to optimise outcomes . Poor understanding of medications is a contributor to poor medication adherence, illustrating the importance of providing appropriate medication and disease education to patients . In an American study of 5014 statin users, only 1654 (33%) were aware of why they were prescribed a statin and what relevance it had to their cardiovascular health . Telehealth, the use of telephony and video-conferencing for communicating health, has been utilised by pharmacists in ambulatory care and cardiology for many years, using a variety of interventions and impact measures . Telehealth is an established way of increasing services without affecting the quality of care. It has previously been utilised successfully in the management of cardiovascular diseases . The emerging requirement for non-patient facing consultations due to the COVID-19 pandemic also highlighted the need for telehealth-based interventions. There is evidence for the impact that a pharmacist can have as a part of the multi-disciplinary team providing care to patients in the ambulatory care setting . One randomised control trial from Canada utilised pharmacists and telehealth, and involved phone calls of 5–10 min at 1 week, 1 month, 6 months and 9 months post coronary stenting. The data illustrated that at 12 months 87.2% (n = 131/150) of patients in the intervention arm remained on clopidogrel with only 43.1% (n = 65/150) of patients in the control (no phone call follow up) arm . This study seeks to build upon existing literature in targeted medication adherence interventions . Aim This study aimed to explore the effects of a telehealth cardiology pharmacist clinic on patient adherence to secondary prevention medications in the 12 months following ACS. Ethics approval This study was granted approval and waiver of consent by Grampians Health Ballarat Human Research Ethics Committee (Project number LNR/71907/BHSSJOG) and Monash University (28159).
This study aimed to explore the effects of a telehealth cardiology pharmacist clinic on patient adherence to secondary prevention medications in the 12 months following ACS.
This study was granted approval and waiver of consent by Grampians Health Ballarat Human Research Ethics Committee (Project number LNR/71907/BHSSJOG) and Monash University (28159).
Study design This was a retrospective matched cohort study with a 12 month follow up duration from the index PCI. It utilised an existing data source of patients who received PCI with coronary stenting. The study consisted of two groups, an intervention group from 2020 who received a consult in the telehealth cardiology pharmacist clinic, and a control group of patients prior to 2020 who did not receive the intervention. Telehealth cardiology pharmacist clinic The telehealth cardiology pharmacist clinic involves a 20-min consultation with a pharmacist addressing a number of different care elements. The mode of delivery of the service, as well as patient and clinician acceptability has been assessed in a previous study . A detailed best-possible medication history is taken with secondary source verification in line with the Society of Hospital Pharmacists Australia: Standards of Practice in Clinical Pharmacy . Questions regarding cardiac symptoms and a functional assessment are completed, which align with the outcome measures defined in the Melbourne Interventionalist Group (MIG) assessment form. These questions are designed to understand the burden of cardiac symptoms as well as identifying any potential urgent referral points for patients. Education on cardiac health and treatment is provided in line with the National Heart Foundation of Australia National Cardiac Rehabilitation Quality Indicators . Patients are consulted at 1 month, 3 months and 12 months post PCI. The format is repeated for each of the consultations, with a focus on building patient knowledge of cardiac medications and health. When barriers to access of medications are noted, the pharmacist took steps to rectify this by consultation with the cardiology unit and/or the primary care physician. Examples would be arranging new prescriptions, recommending modification of dose or dosage forms, and implementing dose administration aids. Additionally, attendance at cardiac rehab was assessed and referrals made where indicated. Study setting The study was based at a large public regional health service in Victoria, utilising data collected as part of the health services participation in the Victorian Cardiac Outcomes Registry (VCOR) and MIG, as well as data from the telehealth cardiology pharmacist clinic consult letters. Study participants There are approximately 250 PCI and coronary stent patients who present to Grampians Health Ballarat annually with ACS. Patients who received PCI were enrolled into an opt-out MIG registry. Data are collected at the time of PCI and at 30 days. Twelve-month follow up data were available for patients who received PCI with coronary stents in the 2015–2017 cohort. From 2017, routine 12 month follow up data were not collected. Patients in the intervention arm had their baseline MIG data collected at time of PCI and at 30 days. Due to the lack of 12-month MIG data, the telehealth cardiology pharmacist clinic report was used to check if OMT was present at 12 months. In order to allow for comparison between control and intervention groups, set criteria were established to ensure 12-month follow up data were available. It was the intention of this study to enrol all patients who received PCI for ACS during the study period of January 2020 to July 2020. Inclusion criteria: Adults aged 18 years or older, no upper age cut-off Diagnosed with ACS requiring PCI with coronary stenting, inclusive of ST elevation Myocardial Infarction (STEMI) Non-ST elevation Myocardial infarction (NSTEMI) Unstable angina Patients who had 12-month follow up data available for analysis (either via registry or clinic for control and intervention arms respectively) Patient or carer able to participate in telehealth consult or via phone (intervention arm only). Exclusion criteria: Patients with ACS not treated with PCI who were transferred for surgical intervention PCI without stent deployment (balloon angioplasty only) Elective PCI for stable angina Patients with unsuccessful PCI that were escalated to surgical management Patients who chose to opt-out from registry or clinic at any time during the 12-month follow up period. Adherence definitions Adherence to medication classes was broken up into three groups, similar to previous studies . These groups included: Optimal Medical Therapy (OMT) (all four medication groups) Near-optimal Medical Therapy (NMT) (three medication groups) Sub-optimal Medical Therapy (SMT) (less than three medication groups). Patients are specifically asked about adherence to each medication comprising OMT, with adherence defined as taking the medication more than 80% of the time . Study outcomes The primary outcome was the difference of self-reported adherence to all four groups of secondary prevention medications (optimal medication therapy) at 12 months post coronary stenting between a matched cohort of patients who received the intervention and those who did not. Self-reported adherence was determined by a set assessment form by the Melbourne Interventionalists Group, where the pharmacist would ask the patient directly during the telehealth consultation. This questionnaire was identical between the control and intervention groups. Secondary outcomes included the difference in Near-optimal Medical Therapy and Sub-optimal Medical Therapy, and individual medication groups (DAPT, statin, beta blocker and ACEI/ARB/ARNI). Additionally, the difference in Major Adverse Cardiovascular Events (MACE) at 12 months between control and intervention matched cohort was investigated. MACE was defined as stroke, non-fatal myocardial infarction, rehospitalisation or death. To validate the use of self-reported adherence within the study, self-reported adherence was compared to calculated medication possession ratio (MPR) via the patient’s primary pharmacy dispensing records. This outcome was only measured in the intervention group due to dispensing data availability. Statistical analysis Cohort matching was used to reduce potential confounding between the control and intervention arms . As this study is retrospective and non-randomised, the use of individual matching between cohorts provides a method to reduce confounding . The matching criteria were selected due to both their availability within the data set and evidence regarding their significant correlation with changing adherence patterns between participants . In this study, matching was performed using individual matching across criteria: Age stratification at time of percutaneous coronary intervention (PCI) o < 50, 50–59, 60–69, 70–79, 80–89, > 89. Sex o Male, female. Type of acute coronary syndrome (ACS) o STEMI, NSTEMI, unstable angina. Left ventricular dysfunction at PCI defined as by stratified ejection fraction o < 50%, ≥ 50%. Data matching was indexed at the time of the ACS event and baseline MIG data collected. Matching and analysis was performed using Stata ® 17 and Microsoft Excel ®. Based on previous studies, adherence to medications at 12 months post ACS event can vary between 45 and 75% . Based on a population size of approximately 100 patients to have ACS in the 7-month intervention period (January 2020 to July 2020), a margin of error of 5% with an alpha of 0.05 and beta of 0.2, the sample size would need to be 73–78 matched patient pairs (one to one matching). For outcome calculations, McNemar’s Chi-squared analysis for matched data was used between the control and intervention pairs. This was repeated across the primary and secondary outcomes involving matched data. For the adherence measure validation outcome, an R 2 value was calculated between self-reported adherence scores and calculated MPR. A value of 0.75 or greater was considered a substantial correlation.
This was a retrospective matched cohort study with a 12 month follow up duration from the index PCI. It utilised an existing data source of patients who received PCI with coronary stenting. The study consisted of two groups, an intervention group from 2020 who received a consult in the telehealth cardiology pharmacist clinic, and a control group of patients prior to 2020 who did not receive the intervention.
The telehealth cardiology pharmacist clinic involves a 20-min consultation with a pharmacist addressing a number of different care elements. The mode of delivery of the service, as well as patient and clinician acceptability has been assessed in a previous study . A detailed best-possible medication history is taken with secondary source verification in line with the Society of Hospital Pharmacists Australia: Standards of Practice in Clinical Pharmacy . Questions regarding cardiac symptoms and a functional assessment are completed, which align with the outcome measures defined in the Melbourne Interventionalist Group (MIG) assessment form. These questions are designed to understand the burden of cardiac symptoms as well as identifying any potential urgent referral points for patients. Education on cardiac health and treatment is provided in line with the National Heart Foundation of Australia National Cardiac Rehabilitation Quality Indicators . Patients are consulted at 1 month, 3 months and 12 months post PCI. The format is repeated for each of the consultations, with a focus on building patient knowledge of cardiac medications and health. When barriers to access of medications are noted, the pharmacist took steps to rectify this by consultation with the cardiology unit and/or the primary care physician. Examples would be arranging new prescriptions, recommending modification of dose or dosage forms, and implementing dose administration aids. Additionally, attendance at cardiac rehab was assessed and referrals made where indicated.
The study was based at a large public regional health service in Victoria, utilising data collected as part of the health services participation in the Victorian Cardiac Outcomes Registry (VCOR) and MIG, as well as data from the telehealth cardiology pharmacist clinic consult letters.
There are approximately 250 PCI and coronary stent patients who present to Grampians Health Ballarat annually with ACS. Patients who received PCI were enrolled into an opt-out MIG registry. Data are collected at the time of PCI and at 30 days. Twelve-month follow up data were available for patients who received PCI with coronary stents in the 2015–2017 cohort. From 2017, routine 12 month follow up data were not collected. Patients in the intervention arm had their baseline MIG data collected at time of PCI and at 30 days. Due to the lack of 12-month MIG data, the telehealth cardiology pharmacist clinic report was used to check if OMT was present at 12 months. In order to allow for comparison between control and intervention groups, set criteria were established to ensure 12-month follow up data were available. It was the intention of this study to enrol all patients who received PCI for ACS during the study period of January 2020 to July 2020. Inclusion criteria: Adults aged 18 years or older, no upper age cut-off Diagnosed with ACS requiring PCI with coronary stenting, inclusive of ST elevation Myocardial Infarction (STEMI) Non-ST elevation Myocardial infarction (NSTEMI) Unstable angina Patients who had 12-month follow up data available for analysis (either via registry or clinic for control and intervention arms respectively) Patient or carer able to participate in telehealth consult or via phone (intervention arm only). Exclusion criteria: Patients with ACS not treated with PCI who were transferred for surgical intervention PCI without stent deployment (balloon angioplasty only) Elective PCI for stable angina Patients with unsuccessful PCI that were escalated to surgical management Patients who chose to opt-out from registry or clinic at any time during the 12-month follow up period.
Adherence to medication classes was broken up into three groups, similar to previous studies . These groups included: Optimal Medical Therapy (OMT) (all four medication groups) Near-optimal Medical Therapy (NMT) (three medication groups) Sub-optimal Medical Therapy (SMT) (less than three medication groups). Patients are specifically asked about adherence to each medication comprising OMT, with adherence defined as taking the medication more than 80% of the time .
The primary outcome was the difference of self-reported adherence to all four groups of secondary prevention medications (optimal medication therapy) at 12 months post coronary stenting between a matched cohort of patients who received the intervention and those who did not. Self-reported adherence was determined by a set assessment form by the Melbourne Interventionalists Group, where the pharmacist would ask the patient directly during the telehealth consultation. This questionnaire was identical between the control and intervention groups. Secondary outcomes included the difference in Near-optimal Medical Therapy and Sub-optimal Medical Therapy, and individual medication groups (DAPT, statin, beta blocker and ACEI/ARB/ARNI). Additionally, the difference in Major Adverse Cardiovascular Events (MACE) at 12 months between control and intervention matched cohort was investigated. MACE was defined as stroke, non-fatal myocardial infarction, rehospitalisation or death. To validate the use of self-reported adherence within the study, self-reported adherence was compared to calculated medication possession ratio (MPR) via the patient’s primary pharmacy dispensing records. This outcome was only measured in the intervention group due to dispensing data availability.
Cohort matching was used to reduce potential confounding between the control and intervention arms . As this study is retrospective and non-randomised, the use of individual matching between cohorts provides a method to reduce confounding . The matching criteria were selected due to both their availability within the data set and evidence regarding their significant correlation with changing adherence patterns between participants . In this study, matching was performed using individual matching across criteria: Age stratification at time of percutaneous coronary intervention (PCI) o < 50, 50–59, 60–69, 70–79, 80–89, > 89. Sex o Male, female. Type of acute coronary syndrome (ACS) o STEMI, NSTEMI, unstable angina. Left ventricular dysfunction at PCI defined as by stratified ejection fraction o < 50%, ≥ 50%. Data matching was indexed at the time of the ACS event and baseline MIG data collected. Matching and analysis was performed using Stata ® 17 and Microsoft Excel ®. Based on previous studies, adherence to medications at 12 months post ACS event can vary between 45 and 75% . Based on a population size of approximately 100 patients to have ACS in the 7-month intervention period (January 2020 to July 2020), a margin of error of 5% with an alpha of 0.05 and beta of 0.2, the sample size would need to be 73–78 matched patient pairs (one to one matching). For outcome calculations, McNemar’s Chi-squared analysis for matched data was used between the control and intervention pairs. This was repeated across the primary and secondary outcomes involving matched data. For the adherence measure validation outcome, an R 2 value was calculated between self-reported adherence scores and calculated MPR. A value of 0.75 or greater was considered a substantial correlation.
The control cohort consisted of 366 patients in total across the 3-year period, with 335 patients having 12-month follow up data available for analysis. The intervention group contained 107 patients, thirteen of which were excluded as they declined the service, or did not attend any clinic sessions. From these two data sources, 156 patients (78 matched pairs) were matched using the pre-specified criteria detailed in the methods (Fig. ). Following matching, there were no statistically significant differences in the demographic data available from MIG (Table ). Primary outcome There was a significant difference in the number of patients on OMT at 12 months between the groups, in favour of the telehealth cardiology pharmacist clinic cohort. There were 78 matched pairs in the analysis of the primary outcome, with 24 of 78 (31%) patients adherent to OMT at 12 months post PCI in the control group and 34 of 78 (44%) patients in the intervention group demonstrating an absolute difference of 13% ( p = 0.038). Secondary outcome There was no statistically significant difference when comparing matched pairs of participants with near-optimal medical therapy (three out of four post ACS mediation groups present). However, there was a statistically significant absolute reduction in patients with sub-optimal medical therapy (less than three post ACS medication groups present) of 16% (33/78 vs 20/78, p = 0.04). Across each individual medication group, significant differences were seen between matched pairs across all classes except for beta blockers (Table ). There was a significant reduction in major adverse cardiovascular events (MACE) consisting of a 4-point composite outcome of stroke, non-fatal myocardial infarction, rehospitalisation or death. The was an absolute reduction of MACE of 22% (34/78 vs. 17/78, p < 0.01). This was driven primarily through a reduction in hospitalisations (Table ). For the validation of self-reported adherence within the study population, all intervention group participants (n = 98) were included. This outcome compared the self-reported adherence scores to the medication possession ratio (MPR) sourced from participant dispensing histories. The linear regression model demonstrated a R 2 value of 0.84, with 17% (16/94) of participants underestimating and 13% (12/94) participants overestimating adherence when compared to MPR. The remaining 70% (66/94) of participants had MPR within 10% of the self-reported adherence (Fig. ).
There was a significant difference in the number of patients on OMT at 12 months between the groups, in favour of the telehealth cardiology pharmacist clinic cohort. There were 78 matched pairs in the analysis of the primary outcome, with 24 of 78 (31%) patients adherent to OMT at 12 months post PCI in the control group and 34 of 78 (44%) patients in the intervention group demonstrating an absolute difference of 13% ( p = 0.038).
There was no statistically significant difference when comparing matched pairs of participants with near-optimal medical therapy (three out of four post ACS mediation groups present). However, there was a statistically significant absolute reduction in patients with sub-optimal medical therapy (less than three post ACS medication groups present) of 16% (33/78 vs 20/78, p = 0.04). Across each individual medication group, significant differences were seen between matched pairs across all classes except for beta blockers (Table ). There was a significant reduction in major adverse cardiovascular events (MACE) consisting of a 4-point composite outcome of stroke, non-fatal myocardial infarction, rehospitalisation or death. The was an absolute reduction of MACE of 22% (34/78 vs. 17/78, p < 0.01). This was driven primarily through a reduction in hospitalisations (Table ). For the validation of self-reported adherence within the study population, all intervention group participants (n = 98) were included. This outcome compared the self-reported adherence scores to the medication possession ratio (MPR) sourced from participant dispensing histories. The linear regression model demonstrated a R 2 value of 0.84, with 17% (16/94) of participants underestimating and 13% (12/94) participants overestimating adherence when compared to MPR. The remaining 70% (66/94) of participants had MPR within 10% of the self-reported adherence (Fig. ).
This matched retrospective cohort study demonstrated a 13% absolute increase in the degree of adherence to Optimal Medical Therapy (OMT) at 12 months post an acute coronary syndrome (ACS) event. Furthermore, a significant decrease in patients with sub-optimal adherence to medical therapy was observed, which may translate into further reductions in MACE as per previously published work utilising the state-wide MIG registry . The analysis of individual medication groups demonstrate increases in adherence at 12 months when comparing matched pairs, with the exception of beta blockers. Given the trend of other therapies, one possible explanation of why beta blockers did not see an increase was due to a change in recommendations of beta blockers in Non-ST-Elevated Myocardial Infarction (NSTEMI). There has been discussion in guidelines regarding the role of beta blockers in NSTEMI with revascularisation, where beta blockers are no longer recommended in the absence of left ventricular systolic dysfunction . This study was not powered to investigate these individual relationships between ACS types. The significant reduction in MACE was driven primarily by hospital admissions, with a substantial increase in the risk of readmissions in the control group. When considering secondary prevention medications and reductions in MACE, signs of benefit are not fully detected until after 12 months in this setting . The use of self-reported adherence is easy to ascertain from a cost and time perspective, but is challenged with the balance between non-adherence and non-prescription . This study’s use of a secondary outcome of internally validated self-reported adherence within this population provides proof of prescription and dispensing via the review of dispensing records independent to patient self-reporting. The use of two separate measures improves accuracy, particularly when a combination of subjective (self-reported adherence) and objective (medication possession ratio) measures are used . Self-reported outcomes have been documented as being the least reliable, often associated with “white coat adherence” and overestimation . However, the data from this study suggest that this population is just as likely to underestimate as they are to overestimate, with the majority of patients self-reporting within 10% of their calculated MPR. Limitations Limitations in this study include its single-centre setting and although this represents the population of the area the health service operates, this may not reflect Australian or international populations. Telehealth has seen drastic uptake in the COVID-19 era, presenting itself with new challenges in the ambulatory care setting. However, this telehealth model of care was established and validated with patients and clinicians prior to the COVID-19 pandemic, and therefore was not overly affected by changes in practice . Although this study does not replace a randomised study, it has employed various techniques to utilise the benefits randomisation brings to a study. While the matching variables selected have been shown to explain differences in adherence in previous studies, it does not possess the potential power of other methods such as propensity matching . However, this study used population analysis rather than sampling, and randomisation does not control this variance as no sample is drawn. With the population sampled, total adherence within almost all possible recruitment was known. In addition, unmatched comorbidity characteristics showed little difference overall, and adjustment of these variables would have been unlikely to shift the outcome given the high degree of significance. The decision to add therapies is directed by evidence-based guidelines, but also must involve tailoring to the patient’s individual needs and safety. This study treated absence of therapy as non-adherence, however contraindications to therapy or safety outcome guided cessation of therapies may have contributed to what was analysed as non-adherent. This is a common limitation of studies like this, and while the 12-month follow up was the point of interest, dynamic changes in patient medication prescription are not present in the data . However, regardless of therapies prescribed at discharge, the intervention group saw an increase in medication adherence at the end of the follow-up relative to those who did not receive the intervention.
Limitations in this study include its single-centre setting and although this represents the population of the area the health service operates, this may not reflect Australian or international populations. Telehealth has seen drastic uptake in the COVID-19 era, presenting itself with new challenges in the ambulatory care setting. However, this telehealth model of care was established and validated with patients and clinicians prior to the COVID-19 pandemic, and therefore was not overly affected by changes in practice . Although this study does not replace a randomised study, it has employed various techniques to utilise the benefits randomisation brings to a study. While the matching variables selected have been shown to explain differences in adherence in previous studies, it does not possess the potential power of other methods such as propensity matching . However, this study used population analysis rather than sampling, and randomisation does not control this variance as no sample is drawn. With the population sampled, total adherence within almost all possible recruitment was known. In addition, unmatched comorbidity characteristics showed little difference overall, and adjustment of these variables would have been unlikely to shift the outcome given the high degree of significance. The decision to add therapies is directed by evidence-based guidelines, but also must involve tailoring to the patient’s individual needs and safety. This study treated absence of therapy as non-adherence, however contraindications to therapy or safety outcome guided cessation of therapies may have contributed to what was analysed as non-adherent. This is a common limitation of studies like this, and while the 12-month follow up was the point of interest, dynamic changes in patient medication prescription are not present in the data . However, regardless of therapies prescribed at discharge, the intervention group saw an increase in medication adherence at the end of the follow-up relative to those who did not receive the intervention.
This study demonstrated that for patients with acute coronary syndromes, a telehealth cardiology pharmacist clinic added to standard care was able to improve adherence to secondary prevention medications at 12 months. The cardiology pharmacist telehealth clinic increased individual adherence to all drug classes with the exception of beta blockers, and was associated with reduced MACE in the first 12 months following an acute coronary syndrome. This model of care has become a permanent service within Grampians Health, and is being translated into a pharmacist-physician model of care focusing on rapid access atrial fibrillation clinics to improve patient care and utilisation of anticoagulants and anti-arrhythmic agents.
|
In silico designing and expression of novel recombinant construct containing the variable part of | 7b13586a-2d31-4c06-afb5-58b8110770e5 | 10026285 | Anatomy[mh] | INTRODUCTION Breast cancer (BC), as one of the most frequent cancers in women, is rapidly increasing, particularly in low‐ and middle‐income nations. , The cluster of differentiation 44 or CD44, as a tumor‐associated marker, can be used to detect breast cancer stem cells (BCSCs). Normal epithelial cells express CD44, while carcinoma epithelial cells overexpress it. Different CSC properties such as self‐renewability, tumor initiation, metastasis, and chemoradioresistance are regulated by CD44, primarily CD44 variable isoforms (CD44v). As cancer progresses, the total amount spent on therapeutic drugs rises, especially in low‐ and middle‐income countries. Therefore, early intervention is advantageous because it leads to a more successful cure at a lower cost. Although several available therapies are used to treat BC, some tissue stem cells or their early progenitors are thought to be the source of cancer establishment, leading to cancer reappearance and metastasis years after cure. Self‐renewing cells are a tiny minority of tumor cells that may split into multiple cell types, according to recent research. , Cancer stem cells can start a tumor and cause the disease to progress in a limited number of breast cancer populations. In addition, resistance to chemotherapy and radiation contributes to treatment failure and disease relapse; hence, identifying and monitoring BCSCs plays a critical role in prognosis and treatment resistance, potentially opening up new therapeutic options. , Many efforts are currently being undertaken to improve advanced diagnostic and therapeutic procedures as well as develop new super tools, such as discovering markers for BCSCs. Important surface indicators in cancer stem cells have recently been found in several investigations. BCSCs in breast cancer accounts for roughly 2% of all tumor tissue. As a result, getting these cells out of tumor tissue is tough. Many researchers have been hunting for surface breast cancer antigens by identifying the subgroup of BC cells that express CD44 + /CD24 − as a distinctive hallmark of BCSCs. , Furthermore, the CD44 CSC surface marker interacts positively with initiator cells. As a result, the introduction of a complete diagnostic probe for CD44 can lead to early diagnosis and improved treatments. Antibodies or autoantibodies that can react with tumor cells, tissue, and isolated proteins are generated concurrently with the development of tumor‐dependent antigens, while they are scarcely visible at the early stage. Antibodies, on the other hand, are physiologically increased and quantifiable in the early stages of breast cancer, giving them a viable tool for identifying cancerous tissue from healthy tissue. , CD44, which is recognized either as a surface marker of cancer‐initiating cells (CICs) or cancer stem cell antigen, was utilized as a key marker for identifying BCSCs, linking with tumor aggressiveness, metastasis, and recurrence. , , CD44 is a non‐kinase transmembrane glycoprotein having a cytoplasmic carboxyl‐terminal tail and extracellular and transmembrane domains. The CD44 gene has 19 and 20 exons in humans and mice, respectively; the exon number six and the variant number one have no similarity in humans, while the first and last five exons are fixed in all isoforms and encode the shortest isoform of CD44, known as standard CD44 (CD44s). Furthermore, central exons can be alternatively cleaved and constructed with the 10 exons found in the conventional CD44 isoform, known as “variables,” associated with CD44 variant (CD44v) isoforms, which include the middle nine exons. To make a single variant exon, CD44v isoforms can be coupled with other variant exons that code for extracellular domain peptides. , Different CD44v isoforms affect the structure of cell surface as well as the receptors for cytokines and growth factors. CD44's biological activity is influenced by these distinct binding motifs. Hence, cancer cells frequently express a large number of CD44 variations, particularly in the late stages of progression. , , Cell–cell and cell‐substrate interactions, such as proliferation, differentiation, cell migration, cell adhesion, signaling, and survival, are all aided by CD44. Meanwhile, proteolytic enzymes may release this molecule as a cell surface glycoprotein into circulation. As a result, the amount of soluble CD44 proteins in cancer patients' serum correlates with tumor growth and metastasis in various carcinomas. , , CD44 has a high level of expression on bulk tumor cells and CICs, while it has a low level of expression in healthy tissue and differentiated breast tissue, according to the findings. Up‐regulation of CD44, according to immunohistological research, influences the pace of cell invasion, tumor growth, and metastasis in a variety of malignancies, including breast cancer. Furthermore, patients with CD44 overexpression have had unfavorable outcomes. Breast cancer treatment frequently has favorable results when compared to other types of early‐stage cancer treatment, making this initial identification critical, and CD44 is the diagnostic target in the early stages of the disease. , We assumed that BCSC biomarkers are visible in patients' serums because the most extensively utilized biomarkers, HER2 and MUC1, are employed in breast cancer. Because of the heterogeneity created by the complexity of expression CD44v isoforms, it is critical to screen breast cancers and employ these isoforms as diagnostic biomarkers. CD44v was employed as a BCSC marker in this study to look at the link between serum levels and pathological variables. Our objective was to create a recombinant antigen containing a short common section of variable area exons the extracellular domain (rCD44v) that may be utilized as a coating antigen in immunoassay methods to detect anti‐CD44v antibodies. Also, in this study, we showed that polyclonal antibodies obtained from mice are able to detect CD44v antigen in the serum of breast cancer patients by ELISA method, and we further determined that above antibodies have demonstrated this property in patient's tissue by immunohistochemistry.
MATERIALS AND METHODS 2.1 Design of antigen structure The CD44 variants sequences were retrieved in FASTA format from UniProt protein databases (P16070) and aligned using Clustal W software to find the conserved areas of human CD44. Nearly 100 amino acids were taken from various areas of the CD44 extracellular domain regions to make a single synthetic protein. The modified protein's physicochemical characteristics were investigated using the ExPASy server's ProtParam command. The special specialized server GOR‐IV and I‐TASSER predicted the secondary and tertiary structures of proteins, respectively. Likewise, the 3D‐protein models were examined by comparative modeling like Mod‐Base and Loop server. To predict continuous B‐cell epitopes, the amino acid sequences were examined using the Bcepred, and ABCpred programs. Simultaneously, to predict discontinuous B‐cell epitopes, Discotope 1.2 and SEPPA were applied. To predict antigenic sequence binding MHC Class I and T‐cell epitopes, CTLPRED was used. HLApred, MHCpred, and ProPred were used to predict peptides from the recombinant protein binding MHC Class ll. The amino acid sequence of hCD44 was reversely translated into nucleotide sequence, and codons were optimized by the General biosystem service. Ultimately, the secondary structure of mRNA was predicted and analyzed by the mfold program. , , 2.2 Cases and controls A total of 67 tissue and serum samples were taken, 30 healthy women with an average age of 43 and an age range of 26 to 60 years, four patients with benign breast tumors with an average age of 37, and an age range of 26 to 49 years, and 33 breast cancer patients with an average age of 49 and an age range of 28 to 71 years, before and after treatment including surgery, post‐operative radiotherapy, and, or chemotherapy from Imam Khomeini Hospital (Tehran, Iran). We screened 25 samples of positive antigens and 12 samples of negative antigens for CD44 by immunohistochemistry using mouse monoclonal Ab anti‐human CD44 (DAKO, Catalogue # M 7082). The BC patients encompassed 4 benign tumor persons with hyperplasia and fibroadenoma and 33 persons with adenocarcinoma. Meanwhile, the healthy samples included 30 nonpregnant and non‐lactating samples (Table ). 2.3 Expression and purification of CD44v To create the pET28‐CD44v construct, the 107‐amino‐acid designed sequence was synthesized and cloned into the pET28a (General biosystem service). GeNetBio kit was used to extract plasmids (Catalogue # K‐1000). A PCR reaction with T7 universal primers was carried out to validate the pET28a‐CD44 construct. Furthermore, the sequence of construct was presented and confirmed by General biosystem service. The construct was used to convert E. coli BL21 (DE3) competent cells using a thermal shock technique. Four milliliter of pET28a‐CD44 harboring E. coli BL21 (DE3) was inoculated into 200 ml LB broth encompassing 25 μg/ml kanamycin (Sigma‐Aldrich, Catalogue # K1637) and incubated at 170 rpm and 37°C for 2 h to reach OD600 to 0.7–0.9, then 1 mM IPTG (Sinaclon, Catalogue # CL5812) was added. The six His‐tagged fusion protein of CD44 (rCD44v) was subjected to Ni‐NTA column chromatography (Qiagen, Catalogue # 30210) under native conditions. Different concentrations of imidazole(Merck, Catalogue # 104716) followed by a pH gradient procedure were used to elute the rCD44v antigen. Non‐induced recombinant cells were considered as the negative control; 15% SDS‐PAGE was used to check the purity, and protein content was estimated by the Bradford method. 2.4 Recombinant protein confirmation by western blotting The purified antigen was electrophoresed on gel 15% SDS‐PAGE and transferred to a polyvinylidene difluoride (PVDF) membrane using transfer buffer. After running, the membrane was blocked by 5% BSA (Sigma, Catalogue # 7030) at 37°C for 1 h. Thereupon, the PVDF membrane was washed three times with 0.05% tween 20 in PBS buffer (PBST). The recombinant CD44v was incubated with a mouse antibody against His‐tag conjugated with HRP (1/2000, Roche, Catalogue # 11965085001) in PBST buffer and was shaken at 37°C for 1 h and then washed with PBST buffer several times. In the next step, the detection was performed using an HRP staining solution (DAB) (Diagnostic Biosystems, Catalogue # K047). The chromogenic process was finally stopped by washing twice with pure water. 2.5 Preparation of anti‐CD44v serum Twenty micrograms of recombinant CD44v protein was injected subcutaneously into the back of necks of 5‐week‐old female BALB/c mice with complete Freund's adjuvant (Sigma, Catalogue # F5881). On days 14 and 28, an incomplete Freund's adjuvant (Sigma, Catalogue # F5506) was utilized as a booster dosage in future injections. Furthermore, as a negative control, full and incomplete Freund's adjuvant was administered in the same way as previously. The indirect ELISA technique was used to detect antibodies against the rCD44v antigen, and then the mouse serum was collected and kept at −70°C for future usage. 2.6 Immunohistochemical analysis The prepared mouse polyclonal antibody (pAb) against rCD44v and monoclonal antibody (mAb) against hCD44 (DAKO, Catalogue # M 7082) were used to stain the breast cancer tissue according to standard Immunohistochemical (IHC) protocol. Slides were incubated with mouse pAb against rCD44v, mAb against human CD44 diluted with TBST (Tris‐borate saline‐ tween 20) in a 1:100 ratio. After washing the slides by TBST for 5 min and they were incubated with HRP conjugated anti‐mouse antibody (Diagnostic Biosystems, Catalogue # PVP100D) for 30 min. The slides were incubated with 3, 39‐diaminobenzidine tetrahydrochloride (DAB) (Diagnostic Biosystems, Catalogue # PVP100D) and instantly washed by tap water after color development. The slides that were not incubated with anti‐rCD44v antibodies were considered as a second control. All slides were immersed in hematoxylin for 3 min and then rinsed with running water. After that, dewatering, clarification, and gluing of the slide were performed, respectively, and ultimately the slides were evaluated under a light microscopy (Olympus BX51). , 2.7 Determination of serum IgG antibody against rCD44v Specific antibody responses in mouse samples and patients' sera were measured using ELISA, in triplicate. To do so, 5 μg of purified rCD44v was diluted in coating buffer (Na 2 CO 3 , NaHCO 3 , pH 9.6) and held at 4°C for 16 h on 96‐well plates (Nunc). After that, the plates were washed three times with PBST and blocked for 1 h at 37°C with 5% BSA (Sigma, Catalogue # P 7030). The mouse control and test serum samples, mAb against CD44 (DAKO, Catalogue # M 7082) diluted (1:50) by PBST, and 37 patients' sera samples diluted (1:25) by PBST, were poured into wells and incubated at 37°C for 1 h. In the following, they were washed by PBST several times and conjugated HRP with goat anti‐mouse IgG and anti‐human IgG secondary Abs (1/2000 in PBST) (Razi BioTech, Catalogue # AP8036 and Catalogue # AP7071, Kermanshah, Iran) was added to the wells at 37°C for 1 h and washed three times by PBST. Afterward, 100 μl of substrate solution containing 0.06% (wt/vol) O‐phenylenediaminedihydrochloride (OPD) (Sigma, Catalogue # P1526) and 0.06% (vol/vol) H 2 O 2 (DNAbiotec, Catalogue # DB9651, Pretoria, South Africa) were added to the plates and placed at ambient temperature for 15 min in the dark. Ultimately, the reaction was terminated with 100 μl of 2 M H 2 SO 4 , and the OD at the wavelength of 492 nm was recorded by an ELISA reader (BioTek, ELX800). , , 2.8 Quantitative measurement of serum CD44v by ELISA Microplates coated with 2 μg rCD44v were serially diluted to 1:100 by coating buffer (pH 9.6) and placed at 4°C for 16 h. Next, the microplates were blocked with 5% BSA solution at 37°C for 1 h. After washing by PBST, mice serums containing anti‐rCD44v antibodies were poured into each well at a dilution of 1/100, then incubated at 37°C for 1 h, and washed again. Next, anti‐mouse IgG conjugated with peroxidase (Razi BioTech, Catalogue # AP8036) was applied to the wells at a dilution of 1/2000 and incubated for 1 h at 37°C, followed by the addition of OPD as a peroxidase substrate, the color was developed, and OD492 was recorded. A standard curve was created to measure the relative values of rCD44v in serum samples, and the optical density of CD44v antigen in patients' serum was measured using the standard curve. 2.9 Statistical analysis The data are obtained from three independent experiments and analyzed using SPSS version 17.0 (SPSS Inc.), as mean ± standard deviation. For multigroup comparisons, the statistical significance was determined using the Two‐way ANOVA test ( p ‐value >.05).
Design of antigen structure The CD44 variants sequences were retrieved in FASTA format from UniProt protein databases (P16070) and aligned using Clustal W software to find the conserved areas of human CD44. Nearly 100 amino acids were taken from various areas of the CD44 extracellular domain regions to make a single synthetic protein. The modified protein's physicochemical characteristics were investigated using the ExPASy server's ProtParam command. The special specialized server GOR‐IV and I‐TASSER predicted the secondary and tertiary structures of proteins, respectively. Likewise, the 3D‐protein models were examined by comparative modeling like Mod‐Base and Loop server. To predict continuous B‐cell epitopes, the amino acid sequences were examined using the Bcepred, and ABCpred programs. Simultaneously, to predict discontinuous B‐cell epitopes, Discotope 1.2 and SEPPA were applied. To predict antigenic sequence binding MHC Class I and T‐cell epitopes, CTLPRED was used. HLApred, MHCpred, and ProPred were used to predict peptides from the recombinant protein binding MHC Class ll. The amino acid sequence of hCD44 was reversely translated into nucleotide sequence, and codons were optimized by the General biosystem service. Ultimately, the secondary structure of mRNA was predicted and analyzed by the mfold program. , ,
Cases and controls A total of 67 tissue and serum samples were taken, 30 healthy women with an average age of 43 and an age range of 26 to 60 years, four patients with benign breast tumors with an average age of 37, and an age range of 26 to 49 years, and 33 breast cancer patients with an average age of 49 and an age range of 28 to 71 years, before and after treatment including surgery, post‐operative radiotherapy, and, or chemotherapy from Imam Khomeini Hospital (Tehran, Iran). We screened 25 samples of positive antigens and 12 samples of negative antigens for CD44 by immunohistochemistry using mouse monoclonal Ab anti‐human CD44 (DAKO, Catalogue # M 7082). The BC patients encompassed 4 benign tumor persons with hyperplasia and fibroadenoma and 33 persons with adenocarcinoma. Meanwhile, the healthy samples included 30 nonpregnant and non‐lactating samples (Table ).
Expression and purification of CD44v To create the pET28‐CD44v construct, the 107‐amino‐acid designed sequence was synthesized and cloned into the pET28a (General biosystem service). GeNetBio kit was used to extract plasmids (Catalogue # K‐1000). A PCR reaction with T7 universal primers was carried out to validate the pET28a‐CD44 construct. Furthermore, the sequence of construct was presented and confirmed by General biosystem service. The construct was used to convert E. coli BL21 (DE3) competent cells using a thermal shock technique. Four milliliter of pET28a‐CD44 harboring E. coli BL21 (DE3) was inoculated into 200 ml LB broth encompassing 25 μg/ml kanamycin (Sigma‐Aldrich, Catalogue # K1637) and incubated at 170 rpm and 37°C for 2 h to reach OD600 to 0.7–0.9, then 1 mM IPTG (Sinaclon, Catalogue # CL5812) was added. The six His‐tagged fusion protein of CD44 (rCD44v) was subjected to Ni‐NTA column chromatography (Qiagen, Catalogue # 30210) under native conditions. Different concentrations of imidazole(Merck, Catalogue # 104716) followed by a pH gradient procedure were used to elute the rCD44v antigen. Non‐induced recombinant cells were considered as the negative control; 15% SDS‐PAGE was used to check the purity, and protein content was estimated by the Bradford method.
Recombinant protein confirmation by western blotting The purified antigen was electrophoresed on gel 15% SDS‐PAGE and transferred to a polyvinylidene difluoride (PVDF) membrane using transfer buffer. After running, the membrane was blocked by 5% BSA (Sigma, Catalogue # 7030) at 37°C for 1 h. Thereupon, the PVDF membrane was washed three times with 0.05% tween 20 in PBS buffer (PBST). The recombinant CD44v was incubated with a mouse antibody against His‐tag conjugated with HRP (1/2000, Roche, Catalogue # 11965085001) in PBST buffer and was shaken at 37°C for 1 h and then washed with PBST buffer several times. In the next step, the detection was performed using an HRP staining solution (DAB) (Diagnostic Biosystems, Catalogue # K047). The chromogenic process was finally stopped by washing twice with pure water.
Preparation of anti‐CD44v serum Twenty micrograms of recombinant CD44v protein was injected subcutaneously into the back of necks of 5‐week‐old female BALB/c mice with complete Freund's adjuvant (Sigma, Catalogue # F5881). On days 14 and 28, an incomplete Freund's adjuvant (Sigma, Catalogue # F5506) was utilized as a booster dosage in future injections. Furthermore, as a negative control, full and incomplete Freund's adjuvant was administered in the same way as previously. The indirect ELISA technique was used to detect antibodies against the rCD44v antigen, and then the mouse serum was collected and kept at −70°C for future usage.
Immunohistochemical analysis The prepared mouse polyclonal antibody (pAb) against rCD44v and monoclonal antibody (mAb) against hCD44 (DAKO, Catalogue # M 7082) were used to stain the breast cancer tissue according to standard Immunohistochemical (IHC) protocol. Slides were incubated with mouse pAb against rCD44v, mAb against human CD44 diluted with TBST (Tris‐borate saline‐ tween 20) in a 1:100 ratio. After washing the slides by TBST for 5 min and they were incubated with HRP conjugated anti‐mouse antibody (Diagnostic Biosystems, Catalogue # PVP100D) for 30 min. The slides were incubated with 3, 39‐diaminobenzidine tetrahydrochloride (DAB) (Diagnostic Biosystems, Catalogue # PVP100D) and instantly washed by tap water after color development. The slides that were not incubated with anti‐rCD44v antibodies were considered as a second control. All slides were immersed in hematoxylin for 3 min and then rinsed with running water. After that, dewatering, clarification, and gluing of the slide were performed, respectively, and ultimately the slides were evaluated under a light microscopy (Olympus BX51). ,
Determination of serum IgG antibody against rCD44v Specific antibody responses in mouse samples and patients' sera were measured using ELISA, in triplicate. To do so, 5 μg of purified rCD44v was diluted in coating buffer (Na 2 CO 3 , NaHCO 3 , pH 9.6) and held at 4°C for 16 h on 96‐well plates (Nunc). After that, the plates were washed three times with PBST and blocked for 1 h at 37°C with 5% BSA (Sigma, Catalogue # P 7030). The mouse control and test serum samples, mAb against CD44 (DAKO, Catalogue # M 7082) diluted (1:50) by PBST, and 37 patients' sera samples diluted (1:25) by PBST, were poured into wells and incubated at 37°C for 1 h. In the following, they were washed by PBST several times and conjugated HRP with goat anti‐mouse IgG and anti‐human IgG secondary Abs (1/2000 in PBST) (Razi BioTech, Catalogue # AP8036 and Catalogue # AP7071, Kermanshah, Iran) was added to the wells at 37°C for 1 h and washed three times by PBST. Afterward, 100 μl of substrate solution containing 0.06% (wt/vol) O‐phenylenediaminedihydrochloride (OPD) (Sigma, Catalogue # P1526) and 0.06% (vol/vol) H 2 O 2 (DNAbiotec, Catalogue # DB9651, Pretoria, South Africa) were added to the plates and placed at ambient temperature for 15 min in the dark. Ultimately, the reaction was terminated with 100 μl of 2 M H 2 SO 4 , and the OD at the wavelength of 492 nm was recorded by an ELISA reader (BioTek, ELX800). , ,
Quantitative measurement of serum CD44v by ELISA Microplates coated with 2 μg rCD44v were serially diluted to 1:100 by coating buffer (pH 9.6) and placed at 4°C for 16 h. Next, the microplates were blocked with 5% BSA solution at 37°C for 1 h. After washing by PBST, mice serums containing anti‐rCD44v antibodies were poured into each well at a dilution of 1/100, then incubated at 37°C for 1 h, and washed again. Next, anti‐mouse IgG conjugated with peroxidase (Razi BioTech, Catalogue # AP8036) was applied to the wells at a dilution of 1/2000 and incubated for 1 h at 37°C, followed by the addition of OPD as a peroxidase substrate, the color was developed, and OD492 was recorded. A standard curve was created to measure the relative values of rCD44v in serum samples, and the optical density of CD44v antigen in patients' serum was measured using the standard curve.
Statistical analysis The data are obtained from three independent experiments and analyzed using SPSS version 17.0 (SPSS Inc.), as mean ± standard deviation. For multigroup comparisons, the statistical significance was determined using the Two‐way ANOVA test ( p ‐value >.05).
RESULTS 3.1 Structural design and prediction A short common section of variable area exons the extracellular domain of hCD44 containing the N‐terminal 100 amino acid residues (441–540) was chosen for assembly in this investigation. As a fusion tag, six His amino acids were utilized at the C‐terminus (Figure ). With the accession number MT396985, the designed sequence was deposited in Genbank. The recombinant protein's physiochemical characteristics comprised a molecular mass of 12 kDa, a pI of 5.65, and aliphatic indexes of 36.45. Figure shows predictions for the second and third recombinant protein shapes. Secondary structure predictions validated the high fraction of random coils (83.18%) and low extended strand (16.82%) (Figure ), while tertiary structure predictions revealed the immunogenic region was exposed (Figure ) and antigenic residues had VaxiJen cutoff values of >0.5. 3.2 Other structural features Algpred tool and SDAP allergen library were applied to predict the allergenicity of sequences. According to their results, there is no potential for allergenicity of rCD44v antigen. 3.3 Prediction of B and T cell epitope Table summarizes the primary epitope properties, including hydrophilicity, accessibility, stability, antigenicity, polarity, and exposed recombinant protein surfaces. Table also displays the anticipated B and T cell epitopes with the greatest interaction score for both MHC Classes I and II. 3.4 Expression, purification of rCD44v protein The PCR reaction was performed with T7 promoter primer and T7 terminator primer in terms of temperature and time in three steps which are: initial denaturation at 95°C for 5 min, (35 cycles) at 94°C 20 s, 54°C 20 s, 72°C 20 s, and final extension at 72°C for 5 min. The PCR products were examined by electrophoresis on agarose gel 1%, and a cloned gene fragment of size 324 bp was observed (Figure ). The expressed recombinant CD44v was purified by Ni metal affinity chromatography, and fractions were examined by SDS‐PAGE 15%, and results showed a band of about 17 kD (Figure ). The accuracy of recombinant CD44v was verified by reaction with anti‐His tag Abs by western blotting (Figure ). In the negative control, however, there was no reactivity (Figure ). 3.5 Detection of specific mouse Abs against CD44v with ELISA and IHC Immunization of mice with pure rCD44v protein resulted in the formation of specific IgG antibodies. Compared with mice control antisera, an anti‐rCD44v IgG antibody titer can be detected even at 1/1600 dilutions (Figure ). Our findings revealed that rCD44v can be detected by either anti‐CD44 or anti‐His‐tag antibodies. Furthermore, IHC examination exhibited that the mice anti‐CD44 polyclonal Ab, like typical anti‐CD44 monoclonal Ab, can detect CD44 protein expressed in breast tumors (Figure ). 3.6 Detection of CD44v protein in patients' sera ELISA analysis revealed that rCD44v protein was capable of detecting anti‐CD44v antibodies in patients' serum. This means that natural anti‐CD44v antibodies can recognize epitopes on rCD44v (Figure ). In addition, anti‐rCD44v antibody identified CD44v proteins in patients' serum on the plate (Figure and Table ). 3.7 Calculating CD44v antigen in the patients' sera by ELISA ELISA was used for quantitative measurement of CD44v antigen in serum by comparing their optical densities with a standard curve and calculating relative values. We considered the cut‐off value of 631.3 ng/ml as the positive quotient of CD44v, which is values higher than 631.3 ng/ml are considered positive CD44v. Thus, the sensitivity and specificity of this procedure were determined through SPSS analysis 92% and 98%, respectively.
Structural design and prediction A short common section of variable area exons the extracellular domain of hCD44 containing the N‐terminal 100 amino acid residues (441–540) was chosen for assembly in this investigation. As a fusion tag, six His amino acids were utilized at the C‐terminus (Figure ). With the accession number MT396985, the designed sequence was deposited in Genbank. The recombinant protein's physiochemical characteristics comprised a molecular mass of 12 kDa, a pI of 5.65, and aliphatic indexes of 36.45. Figure shows predictions for the second and third recombinant protein shapes. Secondary structure predictions validated the high fraction of random coils (83.18%) and low extended strand (16.82%) (Figure ), while tertiary structure predictions revealed the immunogenic region was exposed (Figure ) and antigenic residues had VaxiJen cutoff values of >0.5.
Other structural features Algpred tool and SDAP allergen library were applied to predict the allergenicity of sequences. According to their results, there is no potential for allergenicity of rCD44v antigen.
Prediction of B and T cell epitope Table summarizes the primary epitope properties, including hydrophilicity, accessibility, stability, antigenicity, polarity, and exposed recombinant protein surfaces. Table also displays the anticipated B and T cell epitopes with the greatest interaction score for both MHC Classes I and II.
Expression, purification of rCD44v protein The PCR reaction was performed with T7 promoter primer and T7 terminator primer in terms of temperature and time in three steps which are: initial denaturation at 95°C for 5 min, (35 cycles) at 94°C 20 s, 54°C 20 s, 72°C 20 s, and final extension at 72°C for 5 min. The PCR products were examined by electrophoresis on agarose gel 1%, and a cloned gene fragment of size 324 bp was observed (Figure ). The expressed recombinant CD44v was purified by Ni metal affinity chromatography, and fractions were examined by SDS‐PAGE 15%, and results showed a band of about 17 kD (Figure ). The accuracy of recombinant CD44v was verified by reaction with anti‐His tag Abs by western blotting (Figure ). In the negative control, however, there was no reactivity (Figure ).
Detection of specific mouse Abs against CD44v with ELISA and IHC Immunization of mice with pure rCD44v protein resulted in the formation of specific IgG antibodies. Compared with mice control antisera, an anti‐rCD44v IgG antibody titer can be detected even at 1/1600 dilutions (Figure ). Our findings revealed that rCD44v can be detected by either anti‐CD44 or anti‐His‐tag antibodies. Furthermore, IHC examination exhibited that the mice anti‐CD44 polyclonal Ab, like typical anti‐CD44 monoclonal Ab, can detect CD44 protein expressed in breast tumors (Figure ).
Detection of CD44v protein in patients' sera ELISA analysis revealed that rCD44v protein was capable of detecting anti‐CD44v antibodies in patients' serum. This means that natural anti‐CD44v antibodies can recognize epitopes on rCD44v (Figure ). In addition, anti‐rCD44v antibody identified CD44v proteins in patients' serum on the plate (Figure and Table ).
Calculating CD44v antigen in the patients' sera by ELISA ELISA was used for quantitative measurement of CD44v antigen in serum by comparing their optical densities with a standard curve and calculating relative values. We considered the cut‐off value of 631.3 ng/ml as the positive quotient of CD44v, which is values higher than 631.3 ng/ml are considered positive CD44v. Thus, the sensitivity and specificity of this procedure were determined through SPSS analysis 92% and 98%, respectively.
DISCUSSION A biomarker demonstrates physiological changes from a healthy state to illness, pharmacological therapy, toxin presence, and other environmental stressors. Cancer‐associated autoantibodies are regarded as ideal biomarkers, since they reflect the immune system's biosensors and are known to respond to cancer development. Immunoglobulins are highly stable in serum samples and can last for a long period after the antigen has been removed. They have a major advantage over other bodily fluid measures when used. Like many other tumors, early diagnosis of breast cancer is critical in responding well to the treatment and lowering the mortality rate. CD44, a surface biomarker extracted from CICs, is among the most effective biomarkers for the early identification of breast cancer. CICs and stem cells have many similarities in terms of niche interaction, epithelial‐mesenchymal transformation (EMT), motility, and apoptosis resistance, all of which rely on CD44. Cell adhesion, required for the growth of multicellular organisms, stem cell adhesion, embryonic migration, and tissue healing, is depended on activities of CD44 on the surface of CICs. CD44v, unlike CD44s, is only found on a few epithelial cells and in a few kinds of cancer. CD44 (possibly CD44v) has important biological activities for CICs, and there is a link between incremental CD44 isoform regulation and tumor development to the metastatic phase and breast cancer recurrence. In parenthesis, the CD44 isoforms implicated are also listed (CD44v3, CD44v4, CD44v5, CD44v6, CD44v7, CD44v10). In this study, we cloned and expressed a novel short common sequence of variable extracellular domain human CD44 (variant isoforms 3, 4, 5, 6, 8, 9, and 10) in the pET28a vector. We have developed a reliable ELISA method for detecting serum auto‐antibodies against CD44v, the advantages of our designed molecule are getting a more significant number of epitopes and greater efficiency in coating plates, natural antigenicity without glycosylation, high specificity, and crucial property as a coating antigen to detect anti‐CD44v antibodies in clinical samples. , To test the sensitivity and specificity of ELISA, we devised two separate techniques, one for detecting anti‐CD44v antibodies and the other for detecting CD44v antigens in patients' sera. When serum samples were diluted at 1:25, the results indicated that both techniques had identical detection rates. But, immune responses create huge amounts of antibodies with extended half‐lives, allowing them to be tracked with great sensitivity. Antigens found on malignant or premalignant lesions, on the other hand, are often generated on a tiny scale and are not detectable in the blood due to dilution or clearance. , , The rate of anti‐CD44 IgG was greater in breast cancer patients' pre‐and post‐surgery than in healthy persons, according to the comparative diagnostic advantage of the CD44. , Senbanjo LT et al. discovered a link between blood soluble CD44 concentrations in patients and tumor load and metastasis in BC patients. Also, Isamu Okamoto et al. discovered significant quantities of CD44 in normal breast specimens, however considerably lower than in tumor tissues. Notably, ectodomain cleavage of CD44 is particularly common in tumors, and tumors expressing a CD44 splice variant have an increase in CD44 cleavage; thus, the high prevalence of CD44 cleavage suggests that it plays a key role in tumorigenesis. , Although comparable effects have been reported in other cancer patients, it has been revealed that therapy successfully lowered circulating anti‐CD44 antibody levels in breast cancer patients. , , Following the therapy, the circulating anti‐CD44 IgG in breast cancer patients was depleted in proportion to the period before treatment. During extended illness and latent postoperative micro‐metastasis initial treatment, the consequence of necrosis or metalloproteases, which release membrane‐bound antigen into the blood, resulting in reduced levels of circulating anti‐CD44 antibodies. , , Although there has been a negative association between circulating anti‐CD44 antibodies and antigen in patients with specific malignant tumors, CD44 antigen exhibited a substantial negative link with the matching antibody in breast cancer patients in our current investigation. Meanwhile, there is no correlation between HER2 and CA15‐3 levels and CD44, based on patient characteristics (Table ) and our findings. , , Our findings suggest that low and high CD44v antigen levels in individual patients may be linked to low and high anti‐CD44v antibody levels, respectively. Synchronized elevation of CD44v antigens and anti‐CD44v antibodies in serum can be expected when breast cancer has progressed to a late stage. As a result, anti‐CD44v antibodies may bind to CD44v antigen in breast cancer patients, implying that recurrence or distant metastasis may play a role in CD44v antigen overexpression. , , To summarize, we used two ELISA techniques to look for benign and malignant breast cancers in blood samples. Recognizing the CD44v antigen, according to our findings, is required in breast cancer, particularly in malignant tumors. In addition, assessing the quantity of anti‐CD44v antibodies in the blood can be a useful tool for identifying breast cancer in its early stages, which leads to better outcomes. To expand the study, further extensive investigation with more instances is required.
Elaheh Gheybi: Conceptualization (equal); data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); writing – original draft (equal); writing – review and editing (equal). Ahmad Asoodeh: Investigation (equal); project administration (equal); supervision (equal); writing – review and editing (equal). Jafar Amani: Project administration (equal); supervision (equal); writing – review and editing (equal).
This study was supported by the Research and Technology Council of the Ferdowsi University of Mashhad, Mashhad, Iran (Grant number: 3/48488, 1397/10/11).
The authors have stated explicitly that there are no conflicts of interest in connection with this article.
The present study was approved by the Ethics Committee of Ferdowsi University of Mashhad (Mashhad, Iran), and conducted according to the 1964 Declaration of Helsinki.
Informed consent was obtained from all individual participants included in the study.
|
Metabolomics in oncology | 0c49f6b9-497b-48d3-b88c-2e7a2de090cd | 10026298 | Internal Medicine[mh] | INTRODUCTION Metabolomics includes the systematic identification and quantification of metabolic products from the human body. In this review, we emphasize its relevance and potential applications in the oncology field. Cancer is one of the leading causes of mortality worldwide and a key deterrent to increasing global life expectancy. According to WHO estimates for 2019, cancer is the main cause of death for adults aged below 70 in most countries. Cancer cells have a faulty metabolism causing uncontrolled proliferation. This altered metabolism generates unique metabolic characteristics that can be used to aid in early cancer detection, personalized treatment, and/or gauge therapeutic response. , Metabolic changes in cancer patient due to treatment, nutrition, and exercise can influence cancer outcomes and patient quality of life. , Metabolites can be evaluated in several body fluids such as blood, plasma and urine and therefore represent a potential non‐invasive tool for the management of cancer patients, eventually providing a novel set of diagnostic biomarkers for tumor status and progression. Moreover, its study can support the management of the anticancer treatment response at an individualized level and also predict failures. Biochemical processes and metabolic pathways can be described in details using metabolomics rather than standard clinical laboratory procedures. In metabolomics, metabolites such as monosaccharides, amino acids, small lipids, co‐factors, vitamin B complexes, energy cycle intermediates, nucleotides, exogenous xenobiotics and more are measured and studied comprehensively. , Biomass, energy and redox balance all play a key role in cell metabolism, which is essential for life. There have been numerous studies demonstrating the importance of metabolic reprogramming in a variety of disorders, including cancer, diabetes, cardiovascular disease, and neurological disease. , In order to uncover the underlying causes of disease and devise new treatments, understanding metabolism is a necessity. The objective of this article is to give a general overview on the current and future prospects on metabolomics and its role in cancer detection, monitoring and treatment. Here, we discuss recent developments in metabolomics and then emphasize on the clinical applications of metabolomics.
REPROGRAMMING OF CANCER CELL METABOLISM Metabolic reprogramming helps cancer cells to survive and proliferate during the course of cancer development. The enhanced growth and proliferation in malignant cells require an increased amount of energy in the form of ATP and other co‐factors. This increased demand for resources is fulfilled through alteration of flux via multiple metabolic pathways. Altered Glycolysis and glucose metabolism (Warburg effect) is the most well‐known and studied pathways in cancer metabolism. Over time multiple other pathways have been found to be altered in cancer cells for example lipid metabolism pathway, glutamine metabolism pathway, amino acid metabolism, citric acid cycle, fatty acid oxidation, one‐carbon metabolism etc. The reprogramming in theses pathways is complex and involves multiple factors. Also depending on the cancer type the reprogramming occurs in various degrees and in contexts with the microenvironmental conditions, providing required plasticity to cancer cell. In this review we have tried to differentiate the metabolic state of cancer cells depending on the stages of cancer progression. For cancer cells to proliferate, invade, and metastasize they need to acquire a different metabolic state. It can broadly be classified into 3 stages. (A) Tumor micro‐environments are typically acidic and hypoxic with a distinct nutrient composition compared to normal tissues, forcing cancer cells to adapt to such conditions in order to survive. (B) During invasion, for survival in blood vessels, cancer cells must reprogram their metabolic state, allowing for anchorage‐independent growth. (C) Lastly, when cancer cells colonize other organs, they must adapt to a completely new metabolic environments compared to primary sites in order to grow. , Understanding the mechanisms underlying this metabolic reprogramming can help in identification of new therapeutic targets for cancer. 2.1 Metabolic reprogramming and tumor microenvironment Tumor microenvironments have altered metabolic mechanism compared to normal tissues. This metabolism is influenced by number of intrinsic and extrinsic factors (Figure ). The classic example of cell‐intrinsic factor is altered glycolysis (also known as Warburg effect), a fast glycolysis event due to the necessity of malignant proliferation. Lack of oxygen in the local tumor microenvironment is frequently caused by tumor cells high proliferative capacity and high energy requirement. However, despite the fact that glycolysis does not give as much energy as aerobic respiration, it is 100 times faster and produces the amino acids and pentose phosphates required by rapidly proliferating cancer cells. , Similarly, another most frequently altered signaling pathway in human cancer is phosphatidylinositol‐3‐kinase (PI3K)/Akt signaling pathway which along with mTOR (mammalian target of rapamycin) controls the uptake of glucose, lipids, nucleotides and amino acids. , Aberrant activation of this pathway through mutation is dominant in many tumor progressions. Other intrinsic factors include altered glutaminolysis, activated mitochondrial electron transport chain (ETC) function , altered tricarboxylic acid (TCA) cycle in many cancer types, mutations in enzymes such as isocitrate dehydrogenases 1 and 2 (IDH1, IDH2), succinate dehydrogenase (SDH) and (fumarate hydratase) FH. , On the other hand, extrinsic factors like transformed nutrient microenvironment due to heterogenous density of blood and lymphatic vessels force cancer cells for metabolic reprogramming in order to survive. , Other extrinsic factors for example hypoxia and acidification also play a critical role in metabolic reprogramming , (Figure ). Hypoxic microenvironments lead to upregulation and stabilization of hypoxia‐inducible factors (HIFs), which are known to regulate expression of several genes (for example SNAIL, ZEB1, TWIST, matrix metalloproteinases, lactate dehydrogenase A and pyruvate dehydrogenase kinase 1 , , , , , ) that contribute to cancer progression, including many involved in cell survival, angiogenesis, glycolysis, cancer invasion, and metastasis. Understanding these metabolic switches and their role in metabolic reprogramming can provide critical insights in designing effective treatment regimens. 2.2 Metabolic reprograming for anchorage‐independent growth For cancer cells to metastasise anchorage independent growth is a requirement, wherein cancer cells must detach from extra cellular matrix, enter blood/lymphatic vessel and survive in anchorage‐independent manner. Interestingly a very small portion of circulating cancer cells are capable of doing this. , Primarily because anchorage independent growth requires metabolic reprogramming, which induce oxidative stress on the cells. A classic example of this metabolic reprogramming is reductive carboxylation of glutamine that reduces excessive reactive oxygen species (ROS) in mitochondria. This process converts glutamine derived α‐ketoglutarate (αKG) to citrate through cytosolic IDH1 enzyme. Another enzyme fatty acid synthase (FASN) is essential in maintaining the IDH1 activity thus indirectly regulating reductive carboxylation. Bueno et al., have recently showed pharmacological inhibition of FASN can prevent tumor progression. Similarly, this reductive carboxylation pathway was also found to be essential in maintaining cell proliferation under hypoxic conditions in renal cell carcinoma (RCC). , Also, pentose phosphate pathway upregulation is observed in multiple cancer types and is associated with anchorage independent growth, invasion and metastasis mainly observed in KRAS‐induced anchorage‐independent growth. Amino acid metabolism also play an important role in Anchorage‐Independent cell survival, it works by altering sphingolipid diversity through deregulation of serine, alanine, and pyruvate. In conclusion decrypting this metabolic network involved in anchorage‐independent growth can provide helpful insights in designing therapeutic strategies to prevent cancer metastasis.
Metabolic reprogramming and tumor microenvironment Tumor microenvironments have altered metabolic mechanism compared to normal tissues. This metabolism is influenced by number of intrinsic and extrinsic factors (Figure ). The classic example of cell‐intrinsic factor is altered glycolysis (also known as Warburg effect), a fast glycolysis event due to the necessity of malignant proliferation. Lack of oxygen in the local tumor microenvironment is frequently caused by tumor cells high proliferative capacity and high energy requirement. However, despite the fact that glycolysis does not give as much energy as aerobic respiration, it is 100 times faster and produces the amino acids and pentose phosphates required by rapidly proliferating cancer cells. , Similarly, another most frequently altered signaling pathway in human cancer is phosphatidylinositol‐3‐kinase (PI3K)/Akt signaling pathway which along with mTOR (mammalian target of rapamycin) controls the uptake of glucose, lipids, nucleotides and amino acids. , Aberrant activation of this pathway through mutation is dominant in many tumor progressions. Other intrinsic factors include altered glutaminolysis, activated mitochondrial electron transport chain (ETC) function , altered tricarboxylic acid (TCA) cycle in many cancer types, mutations in enzymes such as isocitrate dehydrogenases 1 and 2 (IDH1, IDH2), succinate dehydrogenase (SDH) and (fumarate hydratase) FH. , On the other hand, extrinsic factors like transformed nutrient microenvironment due to heterogenous density of blood and lymphatic vessels force cancer cells for metabolic reprogramming in order to survive. , Other extrinsic factors for example hypoxia and acidification also play a critical role in metabolic reprogramming , (Figure ). Hypoxic microenvironments lead to upregulation and stabilization of hypoxia‐inducible factors (HIFs), which are known to regulate expression of several genes (for example SNAIL, ZEB1, TWIST, matrix metalloproteinases, lactate dehydrogenase A and pyruvate dehydrogenase kinase 1 , , , , , ) that contribute to cancer progression, including many involved in cell survival, angiogenesis, glycolysis, cancer invasion, and metastasis. Understanding these metabolic switches and their role in metabolic reprogramming can provide critical insights in designing effective treatment regimens.
Metabolic reprograming for anchorage‐independent growth For cancer cells to metastasise anchorage independent growth is a requirement, wherein cancer cells must detach from extra cellular matrix, enter blood/lymphatic vessel and survive in anchorage‐independent manner. Interestingly a very small portion of circulating cancer cells are capable of doing this. , Primarily because anchorage independent growth requires metabolic reprogramming, which induce oxidative stress on the cells. A classic example of this metabolic reprogramming is reductive carboxylation of glutamine that reduces excessive reactive oxygen species (ROS) in mitochondria. This process converts glutamine derived α‐ketoglutarate (αKG) to citrate through cytosolic IDH1 enzyme. Another enzyme fatty acid synthase (FASN) is essential in maintaining the IDH1 activity thus indirectly regulating reductive carboxylation. Bueno et al., have recently showed pharmacological inhibition of FASN can prevent tumor progression. Similarly, this reductive carboxylation pathway was also found to be essential in maintaining cell proliferation under hypoxic conditions in renal cell carcinoma (RCC). , Also, pentose phosphate pathway upregulation is observed in multiple cancer types and is associated with anchorage independent growth, invasion and metastasis mainly observed in KRAS‐induced anchorage‐independent growth. Amino acid metabolism also play an important role in Anchorage‐Independent cell survival, it works by altering sphingolipid diversity through deregulation of serine, alanine, and pyruvate. In conclusion decrypting this metabolic network involved in anchorage‐independent growth can provide helpful insights in designing therapeutic strategies to prevent cancer metastasis.
METABOLIC REPROGRAMMING TO FORM METASTATIC TUMOR One of the leading causes of death in cancer patients is metastasis. After extravasation, in order to survive cancer cells must reprogramme their metabolic status as per the new site which is distinct from that of the primary site. Thus, to adapt to this new microenvironment multiple enzymes and pathways are deregulated. For example, in metastatic breast cancer increased proline catabolism is observed via proline dehydrogenase (PRODH) upregulation compared to primary breast cancer cases, Similarly asparagine is known to increase the metastatic and invasive capabilities in breast cancer cells. It works through upregulation of asparagine synthetase (ASNS) an enzyme responsible to synthesize asparagine from aspartate. Multiple other proteins are known to be deregulated in many cancer types that support metastases for example deregulation of phosphoglycerate dehydrogenase (PHGDH), α‐ketoglutarate, monocarboxylate transporter 1 (MCT1), pentose phosphate pathway, acetyl‐CoA carboxylase (ACC) etc. Decoding these metabolic pathways and their role in metastatic can be beneficial in designing new targeted therapies against metastatic cancer types.
DRUGS TARGETING CANCER METABOLISM Increased understanding of cancer metabolism has led to development of new drugs targeting tumor metabolism. These metabolism‐based therapies are designed to target specific metabolic pathways that are involved in tumor growth and progression. Such drugs work by either blocking the target enzyme involved in the pathway or providing a metabolic product that alters tumor metabolism. Table provides a brief list of metabolism‐based anti‐cancer drugs along with their targets.
METABOLITE‐BASED BIOMARKERS FOR DIAGNOSIS, PROGNOSIS AND PERSONALIZED CANCER TREATMENT It is projected by 2030 around 17 million people would die per year from cancer. Discovery of sensitive biomarkers for cancer using a tailored strategy is now a focus in cancer research and can be utilized as a detection tool for therapeutic targeting of metabolic enzymes. Early intervention in cancer treatment could lead to better outcomes if these tactics are successfully applied. In future, diagnostic and prognostic biomarkers of disease will play an important role in individualized treatment and precision medicine. Analyzing metabolic phenotypes will allow for the categorization of patients by their metabolic profiles. For example, larkin et al., recently showed metabolomic biomarkers in blood samples can be used to identify cancer with nonspecific symptoms. A similar study by shi et al., identified 61 differential metabolites in the plasma of children with medulloblastoma. Another study by Oshashi et al., showed through metabolic profiling, metabolite levels differ between head and neck squamous cell carcinoma (HNSCC) and normal tissues. Such metabolite insights can significantly enhance diagnostic and prognostic outcomes in cancer patients. Similar screening used for a large metabolomics investigation identified 15 metabolites that differed between colorectal cancer (CRC) tissue and normal tissue near the tumor. In addition to the elevation of lactate, glycerol, and glutamate linked to the Warburg effect, they found high levels of alanine, aspartate, palmitoleic acid, kynurenine, and uracil, as well as high levels of putrescine, cysteine, hypoxanthine and 2‐aminobutyrate, and low levels of myo‐inositol. In this study, four distinct CRC cohorts from Hangzhou, Shanghai, Beijing, and United States were evaluated, therefore it was believed that this panel could identify CRC in patients from different genetic origins, mutations, clinical stages, and geographic backgrounds. Moreover, CRC with a recurrence time frame of 52.9 months and better 5‐year survival rates could be distinguished from those with a shorter time frame (25.9 months) cases. From this metabolomics study, a possible predictive metabolic signature for human colorectal cancer emerged. Another study by Cacciatore et al., profiled plasma samples from 41 South African men with prostate cancer using nuclear magnetic resonance (NMR) spectroscopy. The inflammatory NMR markers, GlycA (glycoproteins containing N‐acetylglucosamine and N‐acetylgalactosamine portion), and GlycB (glycoproteins containing N‐acetylneuraminic acid portion), were quantified along with several other markers. They found the plasma of patients with aggressive and metastatic prostate cancer have extremely high levels of GlycA and GlycB supporting the use of plasma metabolome to improve the stratification of patients with prostate cancer. In an additional study, proton magnetic resonance spectroscopy (MRS) was used to quantify 2‐hydroxyglutarate (2HG) levels in the tumors of 30 patients. Researchers observed 2HG expression correlated with IDH1 or IDH2 mutations, a common mutation in grade 2 and grade 3 gliomas. Gliomas with IDH‐1 and IDH‐2 gene mutations are more likely to have 2HG buildup. Such information could be exploited in designing more selective therapies. An example of the successful integration between clinical application and metabolomics is the study by Tenori et al., Human epidermal growth factor receptor 2 (HER2) is used as a biomarker for precision medicine in HER2 positive metastatic breast malignancies. Its overexpression can be suppressed by the anti‐HER2 drugs. Women with metastatic breast cancer treated with paclitaxel and either anti‐HER2 medication (lapatinib) or a placebo, had their serum samples examined for metabolic profiles. Comparing the paclitaxel plus lapatinib group, researchers observed patients with HER2 positive illness were more likely to respond favorably to treatment compared to control group. Metabolomics can also be used to guide oncological surgery. Recent advances in cancer surgery have been made possible by the introduction of the biomarker‐based iKnife technology. This technology involves the intra‐operative, Rapid Evaporative Ionization Mass Spectrometry (REIMS) coupled to electrosurgical tools to allow for near real‐time characterization of margins during cautery‐led tumor dissection. Upon clinical validation of the metabolomics profiling method, surgeons may immediately be able to determine if tissue is healthy or cancerous based on the “smoke‐based” metabolomics profiling of healthy and diseased cells. 5.1 Metabolic markers in the progression of cancer For early detection and screening of malignant tumors, numerous new tumor markers have emerged as a result of comprehensive development in modern molecular biology techniques; nonetheless, current conventional tumor indicators lack sensitivity and specificity for the early detection of malignancies. As an alternative metabolomics can be used to analyze and validate metabolites as biomarkers for early detection of malignancies or to correctly and sensitively determine tumor progression in clinical context. An overview of metabolome investigations in different types of tumors is provided in Table .
Metabolic markers in the progression of cancer For early detection and screening of malignant tumors, numerous new tumor markers have emerged as a result of comprehensive development in modern molecular biology techniques; nonetheless, current conventional tumor indicators lack sensitivity and specificity for the early detection of malignancies. As an alternative metabolomics can be used to analyze and validate metabolites as biomarkers for early detection of malignancies or to correctly and sensitively determine tumor progression in clinical context. An overview of metabolome investigations in different types of tumors is provided in Table .
DETECTION METHODS There are several methods used for detection in metabolomics, including nuclear magnetic resonance (NMR), mass spectrometry (MS), metabolic flux analysis (MFA) etc. These techniques can be used alone or in combination to identify and quantify the levels of various metabolites in biological samples. Currently metabolic profiling is performed mainly by exploiting 2 principal techniques: NMR and MS. Both these techniques require small amount of sample and can identify and quantify wide range of molecules simultaneously. NMR spectroscopy is a powerful tool for the detection of metabolites in cancer. In this technique magnetic field and radio frequency pulses are applied on the molecules to characterize the resonant frequency of that atomic nucleus. Thus, providing information about the molecular structure, motion, and chemical environment of the molecules. Hydrogen is most commonly targeted nucleus in biological samples. NMR spectroscopy can be used to identify and quantify a wide range of metabolites, including small molecules, lipids, and amino acids, in cancer cells and tissues. One of the advantages of using NMR for metabolomics in cancer is its ability to provide metabolic fingerprints that can be used to distinguish between normal and cancerous tissue. NMR spectroscopy can also be used to monitor changes in metabolite levels during cancer progression and treatment. It has also been used to study a variety of cancer types, including colon, lung, colorectal, endocrine, etc. NMR spectroscopy can also be used in conjunction with other techniques, such as mass spectrometry (MS), to provide a more comprehensive understanding of the metabolic changes associated with cancer. NMR Spectroscopy can further be broadly classified into (a) one‐dimensional, (b) two‐dimensional, and (c) three‐dimensional NMR methods (1D‐NMR, 2D‐NMR, and 3D‐NMR) (Figure ). Limitations for NMR include initial high start‐up cost and requirement of high user skills. On the other hand, MS is a destructive method that relies upon the generation of gas phase ions. These ions are then separated depending on their mass‐to‐charge ratio and the amount of ionization at each point in time is detected. This data can be analyzed to determine the chemical composition of the samples. One of the advantages of using MS for metabolomics in cancer is its high sensitivity and specificity, which allows for the detection of very small amounts of a wide range of metabolites. MS analysis has been used to study a variety of cancer types, and also in identification of different biomarkers in cancer. For example, Han et al., recently showed four lipid‐based biomarkers that may be used for early diagnosis in lung cancer. MS can be coupled with other analytical techniques. Based on this coupling MS can be further classified into (a) ion mobility‐MS (IM‐MS), (b) direct injection‐MS (DI‐MS), (c) liquid chromatography‐MS (LC–MS), (d) capillary electrophoresis‐MS (CE‐MS) and (e) gas chromatography‐MS (GC–MS). Such coupling provides a more inclusive understanding of the metabolic variations related with cancer. Apart from these 2 principal techniques other techniques like metabolic flux analysis (MFA) are gaining popularity in cancer research. In this technique concurrent identification and quantification of metabolic fluxes are interpreted numerically using stoichiometric models. This permits study on huge set of reactions (e.g., transport of metabolites, anabolic reactions, catabolic reactions etc.) that define the make‐up and metabolism in malignant cells. , Figure broadly classifies these detection methods. Overall, choosing and using the right techniques in detection of metabolites in cancer can provide valuable insights into the metabolic changes associated with cancer progression and treatment, and may aid in the development of new diagnostic and therapeutic strategies for cancer.
THE FUTURE OF METABOLOMICS IN ONCOLOGY Metabolomics, encompasses a wide range of metabolite analysis, pattern identification, and statistical analysis. For prognostic or predictive interpretation of illness status, metabolic biomarkers can be widely employed in the clinical setting. Multivariate biomarkers, such as cancer fingerprint, profile or signature, should be identified using metabolomics as a biomarker discovery tool. Oncologists may soon be able to detect cancer earlier, when it is still treatable, evaluate the aggressiveness of cancer for better prognosis and treatment, and forecast therapeutic effectiveness with the use of this technology. Despite the fact that advanced analytic procedures are required, these signatures are practical and accurate. Metabolomics has been shown to be a reliable diagnostic tool for leukemia and breast cancer, respectively. Genome microarrays and serum protein profiles have been used in the past to diagnose colon and ovarian cancers, respectively. Metabolomics can be helpful in malignancies that are difficult to diagnose. Ovarian cancer has already been diagnosed using metabolomic profiles from serum. Ascitic fluid, pancreatic secretions, and bronchoalveolar or pleural fluid may all be useful in the diagnosis of cancers in future. When no other test can provide a conclusive response, metabolomics may save time, money and effort by identifying pathognomonic patterns in these fluids and validating them. Even more promising is the prospect of screening cancers through conveniently accessible bodily fluid. Development and application of metabolite spectrum databases, cross‐validation between NMR and MS metabolites, and correlation with other quantitative assays will all play a role in the future of metabolite research. Finally, the results of metabolomic analyses must be integrated with results from other omics technologies in order to characterize the whole malignant phenotypic range. Misdiagnosis, recurrence of symptoms, and undesirable side effects can restrict the clinical efficacy of current clinical practice in illness control. Biomarkers associated with disease‐specific alterations can be examined to establish a tailored method for treating and monitoring disease development by comparing the metabolomic profiles of two or more disease phenotypes. To date, metabolomics implementation in clinic is still not common. There are still open problems to be solved including the scalability of data interpretation and standardization of sample handling practice. The successful transition from research tool to a clinically implemented tool requires cooperation between several disciplines. To date the major problems to be solved are related to equipment design, experimental validation, standardization of methods, and data interpretability in a reproducible fashion. Strict protocols are needed to eliminate any potential biases that may arise from using metabolomics on a variety of biological fluids, including cerebrospinal fluid (CSF), urine, plasma, sweat etc. The metabolomics profile can be highly variable if the time interval between sample collection and processing is not strictly regulated. Furthermore, the quality of the data obtained is harmed by incorrect storage and repeated freeze–thaw cycles. Overcoming these hurdles can greatly benefit clinical application of metabolomics in the long run. Metabolomics can make cancer precision medicine more feasible. Prior to moving into in vivo testing, in‐silico models can be employed to understand the effects of medicines on metabolic characteristics. Metabolomics has opened new avenues for cancer research and is already influencing cancer diagnosis and treatment in number of different ways.
CONCLUSION Although in infancy metabolomics can make a substantial impact on personalized cancer medicine. Using metabolomics, many cancer phenotypes can be accurately described with individualized metabolic markers. This can be used to identified treatment options and/or predict responsiveness to treatments. Also, it will be vital to combine the outcomes of metabolomic assessments with other omics techniques to characterized the entire spectrum of the malignant phenotypes. Technical challenges like database management, cost and methodical knowhow still persist. Overcoming these challenges in near further can help in designing new treatment régimes with increased sensitivity and specificity. ETHICAL STATEMENT Ethical Approval and Consent to participate is not required for such type of studies. Also, the review has been done in accordance to ethical guidelines and that is has been performed in a responsible way, with no misconduct.
Ethical Approval and Consent to participate is not required for such type of studies. Also, the review has been done in accordance to ethical guidelines and that is has been performed in a responsible way, with no misconduct.
The authors have stated explicitly that there are no conflicts of interest in connection with this article.
Gurparsad Singh Suri: Conceptualization (equal); investigation (equal); methodology (supporting); project administration (equal); resources (equal); supervision (equal); validation (equal); visualization (equal); writing – original draft (equal); writing – review and editing (equal). Gurleen Kaur: Conceptualization (supporting); investigation (supporting); resources (supporting); visualization (supporting). giuseppina carbone: Conceptualization (equal); investigation (equal); project administration (equal); supervision (lead); writing – review and editing (equal). Dheeraj Shinde: Conceptualization (equal); data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); project administration (equal); resources (equal); writing – original draft (lead); writing – review and editing (lead).
|
Pharmacogenetics of | a5c5e4d3-2774-49cc-be78-b56c3f063892 | 10026301 | Pharmacology[mh] | INTRODUCTION Acute myeloid leukemia (AML) comprises a heterogenous group of diseases with dismal outcomes mainly in developing countries. Identification of genetic alterations is key for risk stratification, as it guides to appropriate treatment. Most patients with AML are classified as intermediate or unfavorable risk, usually leading to failure of induction therapy and early relapse after achieving complete remission. The identification of gene variants that participate in the processes of pharmacodynamics and pharmacokinetics (pharmacogenetics), have become relevant in the era of personalized medicine, in terms of the influence they exert on the response to treatment, , , and the incorporation of interindividual genetic differences for the design of more effective diagnostic and therapeutic strategies. , , , , Cytarabine (Cytosine arabinoside Ara‐c) and anthracyclines are routinely used in front‐line therapy for AML. , , , Resistance development to chemotherapy is a major obstacle in AML treatment and is responsible for relapses and increased toxicity in second‐line therapies. Ara‐C is a pyrimidine analog which is converted into ara‐CMP by deoxycytidine kinase ( DCK ), later on is converted to ara‐CDP and into ara‐CTP by cytidine/uridine monophosphate kinase 1 ( CMPK1 ) and nucleoside diphosphate kinase 1 (NME1), respectively. Ara‐CTP is a competitor of deoxycytidine 5′‐triphosphate that acts by inhibiting DNA synthesis. Cytidine deaminase ( CDA ) converts ara‐C to the inactive metabolite uracil arabinoside (ara‐U), which limits the amount of ara‐C to be converted to ara‐CTP. , , , ABCB1 , also known as multi‐drug resistance protein 1 (MDRP1), is 1 of 49 putative members in the superfamily of human adenosine triphosphate (ATP)‐binding cassette (ABC) transporters that encode transporter and channel proteins that function as efflux pumps, , codifies a P‐glycoprotein efflux transporter involved in mediating resistance to several drugs, multidrug resistance phenotype, in cancer , (Figure ). Different studies have shown that single nucleotide variants (SNV) in ABCB1, CDA, DCK, GSTT1 , and GSTM1 genes are related to drug toxicity in patients with AML. Two main SNV in the deoxycytidine kinase ( DCK ) gene (‐360C > T, rs377182313) and (‐201C > A, rs2306744), have been described associated to pharmacogenetic responses. Whether DCK mutations make AML cells resistant to cytarabine is controversial. Cytidine deaminase ( CDA ) irreversibly deaminates cytarabine, its overexpression results in Ara‐C resistance, while decreased expression is associated with toxicity. Two SNV in CDA gene have been found associated with pharmacogenetic responses, (79A > C, rs2072671) and (‐451C > T, rs532545). , For ABCB1 , three SNV have been associated to pharmacogenetic responses, ABCB1 (1236G > A, rs1128503), ABCB1 (2677C > A/T, rs2032582) and ABCB1 (3435G > A, rs1045642). , , Several antineoplastic drugs are metabolized by glutathione S‐transferase (GST) which catalyzes the conjugation of reduced glutathione to electrophilic centers of platinum drugs, anthracyclines, vinca alkaloids, cyclophosphamide, and epipodophylotoxins. The GSTT1 gene encodes the phase II metabolizing enzyme glutathione s‐transferase theta and GSTM1 encodes glutathione S‐transferase mu 1. The most studied variants of GSTT1 and GSTM1 are the null variant, which results from the complete or partial deletion of these genes. It has been suggested that individuals lacking GSTT1 and or GSTM1 have an impaired ability to detoxify environmental xenobiotics and are thus at elevated risk for cellular damage and resultant cancer. In the present study we analyzed 6 SNV in ABCB1 gene (3435C > T, rs1045642), (1236G > A, rs1128503), (2677G > T/A, rs2032582); CDA gene (79A > C, rs2072671), (‐451C > T, rs532545); DCK gene (‐201C > A, rs2306744), and the presence or absence of the GSTT1 and GSTM1 null alleles and their correlation with clinical outcomes and toxicity in a cohort of pediatric patients with AML from two pediatric cancer treatment centers in Colombia. MATERIALS AND METHODS 2.1 Patients Descriptive observational cohort study, 51 pediatric patients with a confirmed diagnosis of de novo AML (non‐promyelocitic) by convenience were included, with prior informed consent between March 2015 and June 2021, from HOMI Fundación Hospital Pediátrico La Misericordia and Clínica Infantil Colsubsidio Bogotá D.C., Colombia. Patients with Down syndrome or secondary AML were not included. This project was approved by the ethics committee of each institution, CEI 125‐18 and 243‐1, respectively, and the ethics committee of Universidad Nacional de Colombia (007‐091‐18). Patients received two cycles of induction chemotherapy including cytarabine 100 mg/m 2 /day for 7 days and daunorubycin 60 mg/m 2 /day for 3 days “7 × 3 cycle.” Consolidation chemotherapy was based on cytarabine high dose 3 mg/m 3 /day for 3 days, 2 or 3 cycles maximum. Minimal residual disease (MRD) evaluation was made based on multiparametric flow cytometry on bone marrow samples. Residual disease was detected using the leukemia‐associated immunophenotype at diagnosis and at follow‐up samples. A cutoff value of 0.1% was used as the threshold to distinguish MRD‐positive from MRD‐negative patients, acquiring around 2.5 million events (excluding all CD45‐negative cells and debris). 2.2 Pharmacogenetic testing We used a SNaPShot™ (Thermo Fisher Scientific) panel to simultaneously test for SNV (3435G > A, rs1045642), (1236C > T, rs1128503), (2677C > A/T, rs2032582) in ABCB1 gene; (79A > C, rs2072671), (‐451C > T, rs532545) CDA gene, and separately, (‐201C > A, rs2306744) in DCK gene. Primers and probes used for the SNaPshot™ assay are listed in Table . Primers were designed in order to perform a multiplex PCR reaction using Qiagen 2X PCR multiplex master mix (Qiagen). PCR amplification conditions were an initial denaturing step at 95°C for 15 min, followed by 32 cycles of denaturation at 94°C for 30 s, annealing at 60°C for 90 s followed by extension at 72°C for 60 s, with a final extension step at 72°C for 10 min. For DCK rs2306744, PCR amplification as described previously. Amplified products were purified according to the SNaPshot™ protocol, following manufacturers' recommendations. An aliquot of the purified sample was hybridized with probes designed to align to the specified genetic variant with an additional tail of nonhuman DNA sequences to obtain better separation of each variant. The purified PCR products were analyzed in a ABI3500 Genetic Analyzer using Liz120 as sizing standard and analysis software GeneMapper 4.2 (Applied Biosystems, Thermo Fisher Scientific). The GSTT1 and GSTM1 deletions were analyzed by conventional PCR (Table ). Each reaction also contained a control gene ABCB1 (rs1045642) for amplification control. PCR amplification conditions were an initial denaturing step at 95°C for 3 min, followed by 32 cycles of denaturation at 94°C for 30 s, annealing at 60°C for 30 s followed by extension at 72°C for 60 s, with a final extension step at 72°C for 10 min. The PCR products were resolved by 2% Nusieve gel electrophoresis in 1X TBE buffer. 2.3 Genotype, allele frequencies, and Hardy Weinberg equilibrium Genotype and allele frequencies were determined by direct counting method. Hardy Weinberg equilibrium was calculated based on observed and expected genotype frequencies. Genotypes for each gene variant obtained for each sample tested are listed in Table . 2.4 Statistical analysis Quantitative variables were reported as means or medians with dispersion measures given in standard deviation and ranges, according to the nature and distribution of the variables, based on Shapiro Wilks normality test to establish the use of parametric tests or non‐parametric. We evaluated the association of genotypes from each gene variant, as well as the co‐occurrence of genotypic variants in the ABCB1, DCK , and CDA with clinical outcomes and toxicity using Chi‐square. Qualitative variables were analyzed with Pearson's Chi‐square test and Fisher's exact test. Statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS) for Windows, version 25.0. A p ‐value <.05 was considered significant. Logistic regression analysis was performed to analyze the relationship of the different variables with outcomes, MRD, relapse, event‐free survival, and overall survival. 2.5 Outcomes and definitions Organ toxicity was evaluated according to the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) Version 5.0, grading Scales 3–4 included: colitis, mucositis, cardiotoxicity, transaminitis, and aspergilosis (Table ). In addition, we also evaluated the efficacy of induction therapy, after 2 cycles 7 × 3, with complete remission (less than 5% of morphological blast count in bone marrow smear and hematological recovery in peripheral blood with platelet count >50 000/μl, >1000/μl leukocytes and absolute neutrophil count >500/μl) and induction failure (>5% of morphological blast count in bone marrow smear and without hematological recovery in peripheral blood counts), relapse and HSCT related toxicity. Other outcomes were overall survival, defined as the time between diagnosis and last contact alive or dead; event‐free survival was defined as the time between diagnosis and death, induction failure, relapse, abandonment, change of treatment institution, or last contact alive. Patients Descriptive observational cohort study, 51 pediatric patients with a confirmed diagnosis of de novo AML (non‐promyelocitic) by convenience were included, with prior informed consent between March 2015 and June 2021, from HOMI Fundación Hospital Pediátrico La Misericordia and Clínica Infantil Colsubsidio Bogotá D.C., Colombia. Patients with Down syndrome or secondary AML were not included. This project was approved by the ethics committee of each institution, CEI 125‐18 and 243‐1, respectively, and the ethics committee of Universidad Nacional de Colombia (007‐091‐18). Patients received two cycles of induction chemotherapy including cytarabine 100 mg/m 2 /day for 7 days and daunorubycin 60 mg/m 2 /day for 3 days “7 × 3 cycle.” Consolidation chemotherapy was based on cytarabine high dose 3 mg/m 3 /day for 3 days, 2 or 3 cycles maximum. Minimal residual disease (MRD) evaluation was made based on multiparametric flow cytometry on bone marrow samples. Residual disease was detected using the leukemia‐associated immunophenotype at diagnosis and at follow‐up samples. A cutoff value of 0.1% was used as the threshold to distinguish MRD‐positive from MRD‐negative patients, acquiring around 2.5 million events (excluding all CD45‐negative cells and debris). Pharmacogenetic testing We used a SNaPShot™ (Thermo Fisher Scientific) panel to simultaneously test for SNV (3435G > A, rs1045642), (1236C > T, rs1128503), (2677C > A/T, rs2032582) in ABCB1 gene; (79A > C, rs2072671), (‐451C > T, rs532545) CDA gene, and separately, (‐201C > A, rs2306744) in DCK gene. Primers and probes used for the SNaPshot™ assay are listed in Table . Primers were designed in order to perform a multiplex PCR reaction using Qiagen 2X PCR multiplex master mix (Qiagen). PCR amplification conditions were an initial denaturing step at 95°C for 15 min, followed by 32 cycles of denaturation at 94°C for 30 s, annealing at 60°C for 90 s followed by extension at 72°C for 60 s, with a final extension step at 72°C for 10 min. For DCK rs2306744, PCR amplification as described previously. Amplified products were purified according to the SNaPshot™ protocol, following manufacturers' recommendations. An aliquot of the purified sample was hybridized with probes designed to align to the specified genetic variant with an additional tail of nonhuman DNA sequences to obtain better separation of each variant. The purified PCR products were analyzed in a ABI3500 Genetic Analyzer using Liz120 as sizing standard and analysis software GeneMapper 4.2 (Applied Biosystems, Thermo Fisher Scientific). The GSTT1 and GSTM1 deletions were analyzed by conventional PCR (Table ). Each reaction also contained a control gene ABCB1 (rs1045642) for amplification control. PCR amplification conditions were an initial denaturing step at 95°C for 3 min, followed by 32 cycles of denaturation at 94°C for 30 s, annealing at 60°C for 30 s followed by extension at 72°C for 60 s, with a final extension step at 72°C for 10 min. The PCR products were resolved by 2% Nusieve gel electrophoresis in 1X TBE buffer. Genotype, allele frequencies, and Hardy Weinberg equilibrium Genotype and allele frequencies were determined by direct counting method. Hardy Weinberg equilibrium was calculated based on observed and expected genotype frequencies. Genotypes for each gene variant obtained for each sample tested are listed in Table . Statistical analysis Quantitative variables were reported as means or medians with dispersion measures given in standard deviation and ranges, according to the nature and distribution of the variables, based on Shapiro Wilks normality test to establish the use of parametric tests or non‐parametric. We evaluated the association of genotypes from each gene variant, as well as the co‐occurrence of genotypic variants in the ABCB1, DCK , and CDA with clinical outcomes and toxicity using Chi‐square. Qualitative variables were analyzed with Pearson's Chi‐square test and Fisher's exact test. Statistical analysis was performed using the Statistical Package for the Social Sciences (SPSS) for Windows, version 25.0. A p ‐value <.05 was considered significant. Logistic regression analysis was performed to analyze the relationship of the different variables with outcomes, MRD, relapse, event‐free survival, and overall survival. Outcomes and definitions Organ toxicity was evaluated according to the National Cancer Institute Common Terminology Criteria for Adverse Events (CTCAE) Version 5.0, grading Scales 3–4 included: colitis, mucositis, cardiotoxicity, transaminitis, and aspergilosis (Table ). In addition, we also evaluated the efficacy of induction therapy, after 2 cycles 7 × 3, with complete remission (less than 5% of morphological blast count in bone marrow smear and hematological recovery in peripheral blood with platelet count >50 000/μl, >1000/μl leukocytes and absolute neutrophil count >500/μl) and induction failure (>5% of morphological blast count in bone marrow smear and without hematological recovery in peripheral blood counts), relapse and HSCT related toxicity. Other outcomes were overall survival, defined as the time between diagnosis and last contact alive or dead; event‐free survival was defined as the time between diagnosis and death, induction failure, relapse, abandonment, change of treatment institution, or last contact alive. RESULTS Fifty‐one patients were included, demographic and clinical characteristics are shown in Table . Thirty patients (58%) were male, M:F ratio 1.4:1, the median age was 10 years (0.15–18 years), median leukocytes at diagnosis was 25 580 × 10 9 /L (1.190–1.896.000), IQR 87020, and CNS involvement in 16 patients (31%). One patient died before starting treatment. Thirty‐three (65%) patients required HSCT. Demographic and clinical data are shown in Table . Genotypes and allele frequencies for each of the gene variants analyzed for ABCB1, CDA, DCK, GSTT1 , and GSTM1 are shown in Table . All genotypes tested were found to be in Hardy–Weinberg equilibrium (data not shown). Several associations between genotypic variants and toxicity outcomes were found. First, we analyzed each genotypic variant independently and later on, genotypic associations within each gene or between genes. Odds ratios, CI 95%, and p ‐values are shown in Table . We found that patients carrying ABCB1 (1236C > T, rs1128503) GG genotype in had a 6.8 OR (CI 95% 1.08–42.73, p = .044) for cardiotoxicity at the end of induction, compared to patients carrying either AA or GA genotypes 0.14 OR (CI 95% 0.023–0.92, p = .044). Patients carrying ABCB1 (3435C > T, rs1045642) GG genotype had a 4.51 OR (CI 95% 1.15–17.75, p = .032) for transaminitis, as opposed to those carrying either AA or GA genotypes 0.22 OR (CI 95% 0.05–0.87, p = .032). Also, ABCB1 (3435C > T, rs1045642) AA genotype was identified as a protective factor for relapse 0.69 OR (CI 95% 0.56–0.85, p = .025), compared to those patients with either GG or GA genotype 1.44 OR (CI 95% 1.17–1.78, p = .025). For ABCB1 (1236G > A rs1128503/2677C > A/T rs2032582/3435G > A rs1045642) AA/AA/AA combined genotypes, a strong association was found with death after HSTC OR 13.73 (CI 95% 1.94–97.17, p = .009). In addition, these genotypes were protective factors against relapse 0.632 OR (CI 95% 0.495–0.805, p = .040). Measurable residual disease (MRD) >0.1% after first cycle of chemotherapy was associated with ABCB1 (1236G > A rs1128503/2677C > A/T rs2032582/3435G > A rs1045642) genotypes GG/CC/GG in addition to CDA (79A > C, rs2072671) CA genotype with 4.11 OR (CI 95% 2.32–725, p = .007), and CDA (−451G > A rs532545) CT genotype also was associated with 3.8 OR (CI 95% 2.23–6.47, p = .027). ABCB1 (1236G > A rs1128503/2677C > A/T rs2032582/3435G > A rs1045642) genotypes GG/CC/GG in addition to CDA (79A > C, rs2072671) CA genotype, showed a risk association with MRD >0.1% after first chemotherapy cycle 4.11 OR (CI 95% 2.32–725, p = .007), and CDA (−451G > A rs532545) CT genotype also was associated with MRD >0.1% after first chemotherapy cycle 3.8 OR (CI 95% 2.23–6.47, p = .027). Genotype GA in DCK (‐201C > A, rs2306744) is a protective factor to develop toxicity related to HSCT 0.8 OR (CI 95% 0.36–0.68, p = .046). CDA (−451G > A, rs532545) genotype CC was found to be a protective factor for colitis 0.2 OR (CI 95% 0.048–0.828, p = .019) in our cohort. Combined genotypes for CDA (−451G > A rs532545) and (79A > C, rs2072671) CC/AA were associated with increased risk for mucositis and liver toxicity after the first 7 × 3 cycle and after consolidation, while genotypes CT/CA were a protective factor. We did not find any association between GSTT1 and GSTM1 null alleles with clinical or toxicity events. A logistic regression model was performed to evaluate the presence of independent predictors associated with relapse, finding a positive association for event‐free survival (relapse or death) and for overall survival with the presence of ABCB1 1236/2677/3435 GA/CA/GA 9.086 OR (CI 95% 1.669–49.466, p = .011). No associated predictors were found for overall survival or MRD. DISCUSSION Few studies have analyzed pharmacogenetic risk associations in pediatric AML patients. Most patients with AML are classified, using conventional cytogenetics, recurrent mutations, and response at the end of induction using morphological or MRD counts at different timelines, there are no prognostic genomic or molecular criteria routinely used to identify patients at risk of chemotherapy failure. Genetic background for genes involved in pharmacological response represents an additional factor in treatment response. , , Identification of pharmacogenetic determinants are potential predictive markers for treatment‐related adverse events and toxicity and in establishing differences in treatment schemes or intensification of therapy in post‐induction phase. Green et al, informed association between ABCB1 (1236C > T, rs1128503), GG genotype with decreased survival when treated with cytarabine in AML patients as compared with AA + AG genotypes, in our cohort we found for ABCB1 (1236C > T, rs1128503) GG genotype an increased risk for cardiotoxicity 6.8 OR (CI 95% 1.08–42.73, p = .044), while AA or GA genotype were a protective factor against cardiotoxicity OR 0.14 (CI 95% 0.023–0.92, p = .044). In the same study, ABCB1 (2677C > A/T, rs2032582) CC genotype was associated with decreased survival when patients with AML were treated with cytarabine compared to genotypes AA + AC, we did not find any association between ABCB1 (2677C > A/T rs2032582) CC genotype and survival or toxicity. ABCB1 (3435G > A, rs1045642) GG genotype was associated with increased likelihood of complete remission when treated with cytarabine and idarubicin in AML patients as compared to AA + AG genotypes, other study showed for ABCB1 (3435G > A, rs1045642) AA + AG genotypes were associated with increased overall survival when treated with cytarabine in AML patients as compared to genotype GG. In our patients for ABCB1 (3435G > A, rs1045642) GG genotype we found an association with toxicity, and not with the response at the end of induction. While with AA genotype we have a protective factor for relapse 0.69 OR (CI 95% 0.56–0.85, p = .025) in contrast with the study mentioned previously. In a previous study, the ABCB1 triple variant haplotype ABCB1 (1236G > A rs1128503/2677C > A/T rs2032582/3435G > A rs1045642) TT/TT/TT was related to increased nephrotoxicity than other genotypes. For the combined genotypes ABCB1 (1236G > A rs1128503/2677C > A/T rs2032582/3435G > A rs1045642) in our patients, we found that the haplotype AA/AA/AA was a protective factor against relapse 0.632 OR (CI 95% 0.495–0.805, p = .040). However, the same genotypes were a risk factor for death after HSCT 13.73 OR (CI 95% 1.94–97.17, p = .009). On the other hand, ABCB1 (1236G > A rs1128503/2677C > A/T rs2032582/3435G > A rs1045642) combined genotypes GA/CA/GA was a risk factor for relapse 4.3 OR (CI 95% 1.09–17.10, p = .037) in our cohort. In a study of adult AML patients, ABCB1 (1236G > A rs1128503/2677C > A/T rs2032582/3435G > A rs1045642) AA/AA/AA genotypes analyzed with other SNV genes showed an increased risk for nephrotoxicity and liver toxicity. In a previous study, DCK gene (‐360C > T, rs377182313) and (‐201C > A, rs2306744) were evaluated among AML adult patients. They found that patients with (‐360C > T, rs377182313) CG and (‐201C > A, rs2306744) CT and (‐360C > T, rs377182313) GG and (‐201C > A, rs2306744) TT compound genotypes displayed a favorable response to chemotherapy and increased expression of dCK mRNA, whereas those with (‐360C > T, rs377182313) CC and (‐201C > A, rs2306744) CC tended to have a poor response and lower expression of mRNA ( p = .025 and p = .0034, respectively). Although (‐360C > T, rs377182313) was not included, no association was found for (‐201C > A, rs2306744) in this cohort. Previously, CDA (79A > C, rs2072671) CC and DCK (‐201C > A, rs2306744) CC genotypes were associated with increased risk of death and DCK (‐201C > A, rs2306744) CC genotype as a risk factor for toxicity grades >3 in a cohort of 27 Mexican patients. We did not find any of these associations, instead, DCK (‐201C > A, rs2306744) GA genotype was found to be a protective factor for toxicity after HSCT. In another study, the CDA (79A > C, rs2072671) CC genotype was associated with increased cytotoxicity when exposed to cytarabine in AML patients as compared with AA genotype. There was no association between this genotype and any toxicity in our cohort. Megias‐Vericat et al, found that CDA (79A > C, rs2072671) AC genotype was associated with overall survival at 5 years 2.2 OR (CI 95% 1.2–4.5, p = .015), event free survival 1.9 OR (CI 95% 1.01–3.4, p = .045) and relapse free survival 9.1 OR (CI 95% 1.2–68.6, p = .032). There was no association between this genotype and better clinical outcomes in our cohort. Also, CDA (‐451C > T, rs532545) TT genotype was associated with increased cytotoxicity when exposed to cytarabine in AML patients as compared to CC genotype. In our cohort, we found that CDA (‐451C > T, rs532545) CC genotype was a protective factor against colitis 0.2 OR (CI 95% 0.048–0.828, p = .019). Parmar et al. reported CDA 79A > C rs2072671 AC + CC genotypes associated with increased drug toxicity when treated with cytarabine as compared to AA genotype in an in vitro assay on healthy volunteers. In our cohort, no effects were found for any combined CDA genotypes. In our cohort, no association was found between ABCB1 (1236G > A rs1128503/2677C > A/T rs2032582/3435G > A rs1045642) AA/AA/AA genotypes and toxicity outcomes when combined with CDA (79A > C, rs2072671) or (‐451C > T, rs532545) and DCK (‐201C > A, rs2306744). However, ABCB1 (1236G > A rs1128503/2677C > A/T rs2032582/3435G > A rs1045642) GG/CC/GG genotype in addition to CDA (79A > C, rs2072671) CA or CDA (‐451C > T, rs532545) CT genotypes showed a higher risk for MRD >0.1% at the end of the first cycle of induction, an important prognostic factor for overall survival and event‐free survival. , Previously, GSTT1 null genotype was associated with an increased rate of early death after the initiation of chemotherapy in Japanese AML patients treated with cytarabine, mercaptopurine, prednisone, and daunorubicin. In our cohort, no association between GSTT1 and GSTM1 and toxicity effects were found, probably due to sample size (data not shown). There is scanty information published on pharmacogenetic associations in pediatric AML evaluating CDA, DCK , and ABCB1 genes. In our cohort, the majority of genotypic associations between these gene variants and toxicity or clinical outcomes were different from previous studies. However, most reports analyzing these gene variants have been reported mainly in adult patients, and in other admixed populations different from our cohort. Further studies are needed to evaluate the associations found here, since this is the first study made in colombian population, which is a highly structured population due to admixture between European derived, Amerindians, and African descent populations. For example, the reported genotype frequencies for CDA (79A > C, rs2072671) in a Mexican cohort are quite different to those reported for a Spanish cohort and our sample. For DCK (‐201C > A, rs2306744), our genotype frequencies were similar to those reported previously, but quite different from those reported in the Mexican cohort. CONCLUSIONS This is the first study of AML pharmacogenetics in Colombia, a country with a highly admix population structure. Here, we have used a SNaPShot™ assay to simultaneously analyze SNV in the CDA, DCK , and ABCB1 genes. With this assay, we can run one sample at a time at very low cost, and results could be obtained at the same time as the cytogenetics studies. Although some genetic associations were found, the low number of pediatric AML cases analyzed could be a limitation and further studies will be required to validate the associations found in an independent cohort. Pharmacogenomics might be useful as a future tool for patient stratification for treatment with different chemotherapy regimens. Luz K. Yunis: Conceptualization (equal); data curation (equal); formal analysis (equal); funding acquisition (equal); investigation (equal); methodology (equal); writing – original draft (equal); writing – review and editing (equal). Adriana Linares‐Ballesteros: Conceptualization (lead); data curation (lead); formal analysis (lead); investigation (lead); methodology (lead); supervision (lead); writing – original draft (lead); writing – review and editing (lead). Nelson Aponte: Formal analysis (supporting); methodology (supporting). Gisela Barros: Data curation (supporting); investigation (supporting). Johnny García: Data curation (supporting); investigation (supporting). Laura Niño: Data curation (supporting); investigation (supporting). Gloria Uribe: Methodology (supporting). Edna Quintero: Methodology (supporting). Juan J. Yunis: Conceptualization (lead); data curation (lead); formal analysis (lead); funding acquisition (lead); investigation (lead); methodology (lead); supervision (lead); writing – original draft (lead); writing – review and editing (lead). The authors have stated explicitly that there are no conflicts of interest in connection with this article. Informed consent was obtained from each patient or guardian. This protocol was approved by the institutional ethics committee of the participant institutions. This project was approved by the ethics committee of each institution, CEI 125‐18 and 243‐1 (HOMI Fundación Hospital Pediátrico La Misericordia and Clínica Infantil Colsubsidio Bogotá D.C., Colombia, respectively) and the ethics committee of Universidad Nacional de Colombia (007‐091‐18). Supplementary Figure S1. Schematic representation of genes analyzed in the anthracyclines (idarrubicine) and cytarabine pathway. Click here for additional data file. Supplemental Table S1 ABCB1, CDA, DCK, GSTT1 and GSTM1 genotypes of pediatric AML patients. Click here for additional data file. Supplemental Table S2 Toxicity definitions according to CTCAE version 5. Click here for additional data file. |
Knowledge, Judgment, and Skills in Reproductive Health Care and Abortion Are Essential to the Practice of Obstetrics and Gynecology | 359f4780-111d-42d6-8307-caab1fa58b33 | 10026965 | Gynaecology[mh] | |
Using Test Question Templates to teach physiology core concepts | 4303292e-9928-48b7-b76e-d53a372f1417 | 10026985 | Physiology[mh] | Your undergraduate physiology students are starting a midterm exam. The first question is this: “How is the control of plasma glucose levels similar to the control of plasma calcium levels?” You are eager to see how students respond, since when you covered these two examples of homeostasis in lecture you were careful to use consistent terminology, referring in both cases to set points, sensors, control centers, effectors, and negative feedback. Student A thinks, “This is pretty straightforward. Both plasma calcium levels and plasma glucose levels are homeostatically regulated variables, meaning that they require set points, chemoreceptor sensors to sense the current concentrations, control centers to compare current levels to the set point and send out endocrine responses, and effectors to reduce discrepancies between the current levels and the set point. This reduction of discrepancies is called negative feedback in both cases.” Student B , however, is not quite as clear on the point of the question. “Well,” they think, “we have hormones to control both plasma calcium and plasma glucose. But the specific hormones are different and are made by different glands, and the responses to the hormones are different, too. Maybe I should just say that the endocrine system handles both?” Meanwhile, student C is drawing a blank until eventually coming up with the word “feedback.” “They both involve feedback, maybe?” they think. “There were two types of feedback, weren’t there? Like, um, afferent and efferent? No, that was something else. I think it was positive and negative. Negative sounds bad, so it’s probably positive. Let’s go with ‘both involve positive feedback to keep levels balanced.’” We offer this hypothetical scenario to acknowledge that we have all been in similar situations, ones where we thought we had taught core concept ideas with admirable clarity and emphasis, yet many students’ test performances suggest otherwise. How can we all do better? In teaching a subject like biology, certain “big ideas,” or core concepts, are considered central to the discipline. For undergraduate biology, the Vision and Change report proposed five core concepts: evolution; structure and function; information flow, exchange, and storage; pathways and transformations of energy and matter; and systems. These five core concepts were subsequently unpacked into component principles and specific statements in the BioCore Guide . The five core concepts were distinguished from six core competencies, or skills, also outlined initially by Vision and Change and later unpacked in the BioSkills Guide . Meanwhile, a team of physiology educators led by Joel Michael, Jenny McFarland, and Harold Modell developed a set of physiology-specific core concepts and have unpacked several of these into conceptual frameworks of component statements and terms . In both the general biology- and physiology-specific projects, there were multiple rounds of input from diverse faculty, offering assurance that the final core concept definitions are not simply the opinions of the authors but instead reflect broad consensus within the biology and physiology education communities. Any instructor who focuses on core concepts must consider the question of how to use assessments to determine whether students understand the core concepts being taught. Since defining the core concepts and unpacking them are herculean tasks , it is understandable that these projects did not initially grapple with practical assessment issues in detail (though a homeostasis concept inventory was eventually created; Ref. ). Moreover, thoughtful assessment in any cognitively rich biology course is inherently complicated. A straightforward one-to-one mapping of core concepts to assessment items is not necessarily feasible or desirable, since an ideal summative assessment would presumably cover core concepts, core competencies, and “noncore” material. However, in our view, any practical department- or course-level initiative to help students learn core concepts (and/or core competencies) must carefully consider assessment strategies, because, from the students’ perspective, “assessment drives learning” , i.e., students mostly study what they think will be on the summative assessments. This assertion, if accepted, arguably has at least two important implications: Instructors should assess students on what we really want them to learn. In other words, we need strong alignment between learning objectives (LOs) and assessments. Instructors should help students learn, practice, and review the material in ways that connect transparently to assessments. In other words, we also need strong alignment between learning activities and assessments (what some might call “teaching to the test”). Our broad impression of undergraduate biology education in the United States is that implication A is more widely accepted and practiced than implication B . As an example, consider the now-common practice of “Blooming” assessment questions according to the cognitive hierarchy originally proposed by Benjamin Bloom and colleagues . Many instructors “Bloom” their assessments because they want to see whether they are assessing the higher-order cognition that is reflected in their LOs. We applaud this practice but consider it insufficient since, in and of itself, it does not address implication B . In other words, using test questions that require mastery of important LOs is important but does not necessarily mean that students are able to practice those LOs in the ways required by the tests. Regarding implication B , diverse opinions on teaching to the test suggest that not all educators share our enthusiasm for closely aligning activities and assessments. However, there is evidence that adding active-learning practice on short-answer questions styled in the format of actual exam questions improves introductory biology students’ test performance and reduces the so-called achievement gap between historically underserved students and other students . Similar approaches have been implemented with apparent success in undergraduate anatomy and physiology courses . We believe that the recently introduced framework of Test Question Templates, or TQTs , is especially well suited for simultaneously addressing implication A and implication B , both in general and in regard to core concepts in particular. In the rest of this article, we first review TQTs and then illustrate how students’ understanding of physiology core concepts may be enhanced via the careful coupling of conceptual frameworks and TQTs. A TQT is a student-facing resource (i.e., a resource that is visible to and written for students) that explicitly connects a lesson learning objective (LLO) with multiple specific examples of how that LLO might be assessed on a test. A TQT begins with a statement of a LLO (more specific than a course-wide learning objective), sometimes called an input-output statement because it is of the form of “given X, students will do Y” . The TQT then provides multiple example questions showing how the LLO can be examined, practiced, and met/achieved, as well as how it could conceivably be assessed on the subsequent summative assessment (test). The examples are fairly different from each other to hint at the range of possible questions while still conforming to the input-output statement. In , for instance, example A involves interpreting an image whereas example B is entirely text based, yet both examples are rooted in the LLO of inferring transport mechanisms from the information given. Perhaps the single most important feature of a good TQT is that its LLO is specific enough to focus attention on particular concepts or skills yet broad enough to allow for a large number of example questions. Its structure should allow instructors to readily create novel questions (i.e., questions that students have not seen before) that, via their novelty, probe students’ true understanding of the LLO (rather than their memorization ability or test-taking savvy). Instructors can thus use TQTs to give students valuable guidance on the scope and format of upcoming test questions without giving away the exact details of the questions. Since students are often very focused on how they will be assessed—a rational, reasonable stance—we have found (unpublished observations) that students are most invested in and appreciative of TQTs when instructors use TQTs consistently throughout a course to show students what will be expected of them on tests. For example, one of us (G.J.C.) tells students that short-answer questions will comprise a majority of the points on every test and that every short-answer question will directly reflect a previously practiced TQT. This messaging clearly conveys that TQTs cover the most important, most testworthy, most studyworthy material. Moreover, it encourages some students to think of additional example questions (see , example C ) as further exploration of the various ways in which an LLO may be understood and assessed. BENEFIT #1 : DESIGNING AND REVISING CORE-CONCEPT COVERAGE The leaders of the core concept movements have consistently acknowledged that individual departments and instructors must choose which core concepts should be covered in which individual courses. Once those decisions are made, how might instructors determine whether they are focusing on their chosen core concepts effectively? We propose two basic goals for core concept coverage in a course: first, that a core concept be explored via examination of specific subconcepts of that core concept and, second, that these examinations occur repeatedly during a course. The first goal hinges on the fact that, as summarized in , each core concept is made up of subconcepts that can be arranged in a multilevel hierarchical structure . Therefore the first goal is that a course should cover the specific subconcepts that are essential to the chosen core concept in the context of the course. The second goal is that, for optimal reinforcement of each chosen core concept and its subconcepts, these should be taught at multiple times during the course (e.g., when exploring different organ systems). One could imagine various audits by which educators could check whether they are meeting these two goals. For example, an instructor could review a course’s lecture slides and notes and check off the points at which a given core concept and its key subconcepts were taught. This exercise could certainly be useful, but the fact that an instructor has presented a core concept and discussed it with appropriate context does not guarantee that students understand it well enough to apply it to a novel assessment question. If instead we embed TQTs throughout a course, encourage the students to actively practice identified core concepts via these TQTs, and then use these TQTs to write test questions (as recommended above), an audit of the TQTs may be especially informative and meaningful. That is, if a core concept’s key subconcepts appear in TQTs in multiple units of the course, the core concept is likely being covered in a manner that is thorough, transparent, and memorable to students. To illustrate the potential value of this approach, we present a self-audit of a 200-level human physiology course taught by one of us (G.J.C.). This survey course, which marches through the body’s organ systems in one 10-wk quarter, is taken mostly by pre-nursing and other pre-allied health students after they have fulfilled prerequisites of one-quarter courses in general chemistry, cell and molecular biology, and human anatomy. The course is designed to emphasize the core concepts of Cell-Cell Communication, the Cell Membrane, Flow Down Gradients, and Homeostasis. An audit of the 130 TQTs included in the Spring 2021 course suggested that certain core concepts and subconcepts were included successfully . For example, the core concept of Flow Down Gradients was highlighted in TQTs for the integumentary, nervous, cardiovascular, respiratory, and reproductive units, and subconcepts F2, F3, F4, and F5 were all addressed with TQTs in multiple organ systems. The self-audit also revealed gaps or missed opportunities for teaching these core concepts. Here is one example: the high-priority core concept of Cell-Cell Communication includes seven critical components, CC1–CC7 , but CC2 (“Transport of messenger molecules is determined by the chemical nature of the messenger”) and CC6 (“Termination of a messenger signal is accomplished in several ways”) were not directly addressed by any TQT in the course . The audit thus helped the instructor notice this gap in his coverage of Cell-Cell Communication and consider possible remedies. A second example of a gap was a failure to reinforce core concepts and/or subconcepts in additional units where they would have fit in well; e.g., the respiratory unit did not include any TQTs on Homeostasis, despite the respiratory system’s importance in keeping plasma CO 2 , O 2 , and pH levels near their set points. This observation led the instructor to write a new respiratory system TQT to cover this gap. One might also note that the muscular unit did not include TQTs addressing any of the four emphasized core concepts (though TQTs addressing the core concept of Structure-Function were included) and that perhaps one or more should be added (e.g., on cell-cell communication at the neuromuscular junction). Thus, a TQT-focused audit may suggest possible strengths and limitations in one’s current coverage of chosen core concepts. We stress that we have no specific beliefs about how many core concepts should be covered how many times during a course; we present this particular audit only as an example of how pedagogical intentions could be checked against pedagogical execution. Beyond the basic goal of iterative exploration of each targeted core concept’s key subconcepts, instructors should consider the additional goal of helping students understand how the subconcepts contribute to the core concept as a whole. Put another way, how might we reconcile core concepts’ inherent broadness with our need for more granular formative and summative assessment items? As we see it, the broad concepts and narrower assessment items must be arranged into clear, explicit hierarchies that will probably resemble the conceptual frameworks themselves. An example of this approach in a physiology-related course can be seen in Physiological Ecology (Biology 403) as taught by Deborah Donovan at Western Washington University (personal communication). In student-facing materials, Donovan connects each of her course’s success criteria (similar to the LLOs of TQTs) to a broader learning target and each learning target to a still-broader core concept. For example, success criteria like “I can analyze a graph comparing the metabolic rate of a homeothermic endotherm to ambient temperature” and “I can predict how and explain why the metabolism-temperature relationship of a homeothermic endotherm will change during different seasons” are shown to serve a learning target about homeothermic endotherms, which in turn is one component of a core concept about the control of body temperature. In our own similar efforts to more deliberately align our teaching with core concepts (and subconcepts) and help students appreciate the interconnectivity, we have forged connections between key subconcepts and TQTs, thus translating these subconcepts into concrete student-facing questions for active learning and assessment. To illustrate our alignment and examples of specific questions, we present examples of TQTs aligned with conceptual framework subconcepts for Homeostasis, Flow Down Gradients, Cell Membrane, and Cell-Cell Communication ( , , , and , respectively). As we developed TQTs, the direct inspiration for most came from lecture material rather than from published core concepts and conceptual frameworks, with TQTs subsequently linked to core concepts when auditing and designing subsequent courses (e.g., ). An advantage of this approach is that the TQTs fit naturally into the lecture material and could be seen by students as a review and direct extension of the lecture/text material. This approach has had two additional consequences. First, since TQTs were not initially written to cover specific subconcepts, most TQTs relate to more than one subconcept, with some connecting to as many as four or five [ – ; a given TQT LLO may receive connections from multiple subconcepts (solid lines)]. Second, each set of TQTs need not encompass the entirety of a core concept, as is evident in – (e.g., see , where subconcepts that connect directly to a TQT are shown; other subconcepts, like F1, are omitted). Collectively, these two consequences reflect the fact that we have not sought to create a TQT for every single subconcept; we have instead aimed to create TQTs that are biologically interesting and closely aligned with existing lecture material, while accepting that not all important subconcepts lend themselves to TQTs. We suggest that departments and instructors consider a similar approach based on what they believe their students need most in the time available (i.e., not all subconcepts need to be explicitly explored in a given unit or course). BENEFIT #2 : CORRECTING CORE-CONCEPT MISCONCEPTIONS Good teaching is, of course, not just a matter of delivering well-organized information; it also involves helping students identify, reflect on, and correct their misconceptions. Misconceptions can be defined as “inaccurate ideas that can predate or emerge from instruction” (p. 5 of Ref. ). By that definition, misconceptions can be important, unimportant, or anywhere in between. In considering the numerous core-concept misconceptions that students may harbor , the conceptual frameworks can help us judge their relative importance. For example, for the conceptual framework of the Cell Membrane , an instructor might deem a misconception about tight junctions (CM2.4.1) less important than a misconception about the structure of the cell membrane as a whole (CM1), since the latter subconcept is at a higher level of the framework (i.e., in the language of Ref. and , CM1 is a Critical Component whereas CM2.4.1 is a lower-tier Elaboration). Indeed, when Michael and Modell asked physiology faculty to rate the relative importance of the Cell Membrane subconcepts, the Critical Components tended to be rated as somewhat more important than their constituent Elaborations (Table 1 of Ref. ). By this kind of hierarchy-based reasoning, one might also deem a misconception about tight junctions (Elaboration CM2.4.1) less important than a misconception about protein-mediated membrane transport (Elaboration CM2.2.2) based on the fact that the latter Elaboration encompasses six substituents (CM2.2.2.1 through CM2.2.2.6). We believe that linking these conceptual frameworks to TQTs provides an additional advantage in revealing and dispelling misconceptions. This belief is based on Ambrose and colleagues’ practical recommendations for addressing misconceptions (p. 37–38), which include “ask students to make and test predictions” (since finding that predictions are wrong can be a wake-up call), “ask students to justify their reasoning” (since this may reveal internal contradictions that can then be addressed), and “provide multiple opportunities for students to use accurate knowledge” (since persistent or nuanced misconceptions may need to be counteracted multiple times during exploration of varying topics). These tips could perhaps be condensed into the following general directive: make lots of related practice problems available to students so that they can notice incorrect predictions/explanations, identify their weak points amidst other valid knowledge, and try again on other related problems. This directive is entirely compatible with the TQT framework, in which related problems are explicitly grouped, facilitating the practice of repetition with variation. We thus propose that TQTs can be a useful tool for unmasking misconceptions and helping students overcome them. To this end, we provide examples that extend the connections between conceptual framework TQTs to more directly address misconceptions, one for each of the core concepts of Homeostasis, Flow Down Gradients, the Cell Membrane, and Cell-Cell Communication ( , , , and , respectively). Although these examples are not prescriptive, they illustrate the general strategy of employing TQTs that are broad enough to be reused across multiple organ systems yet specific enough to focus attention on the misconception and closely related issues. Let us take as an example, which tackles the common misconception that homeostatic mechanisms are active only when the regulated variable diverges from its set point. The four TQT questions in focus on sensors involved in homeostatic feedback, examining the regulation of plasma calcium (usually covered in skeletal and/or endocrine units), plasma pH (usually covered in cardiovascular, respiratory, and/or urinary units), and plasma glucose (usually covered in endocrine and/or digestive units). The first of the four TQT questions addresses the core concept (and Critical Component H3) in general, whereas the following three TQT questions (see , bottom row) directly poke at the misconception (see “H3 Misconception” and “Correct Concept” in box), challenging students’ understanding so that any misconceptions can be revealed, defined, and revised into a more correct understanding. In this figure and the others , all four bottom-row questions (and alternatives not shown) align with a single TQT LLO that in turn is aligned with the conceptual framework, yet the LLO is general enough to be used across these example questions, across different organ systems, and even across courses (e.g., A&P I to A&P II). As we have seen, the repetition with variation recommended by Ambrose and colleagues is intentionally built into – . Is such repetition truly necessary, since it leaves less time for completely new material? Our response is rooted in Vision and Change’s observation that biology courses have historically been tilted too far toward breadth at the expense of in-depth understanding of the core concepts, and this repetition with variation offers a partial corrective. Among other benefits, repetition (especially with different specifics from different topics/organ systems) increases the likelihood that an understanding of these core concepts will stay with students well beyond the end of a particular course. In addition to providing needed iterative practice and assessment, TQTs’ explorations of the same core concept and subconcepts in different contexts (e.g., different organ systems) help students appreciate the pervasiveness of the core concepts throughout biology and physiology. To put it another way, if students only learn sensor function in thermoregulation, they might retain misconceptions about sensors that hinder a full, general understanding of sensors in homeostatic systems. The previous paragraph can also be viewed through the lens of equitable teaching, i.e., structuring classes to put success within the reach of as many students as possible . As suggested by our opening vignette, most classes include individuals who vary widely in initial content knowledge, study skills, self-efficacy, time available for studying, and so on. When TQTs are implemented as suggested above, they give students multiple opportunities to dispel misconceptions and acquire and demonstrate competence, so that initial struggles and occasional setbacks do not preclude eventual acquisition and demonstration of mastery . Our opening vignette also implies that physiology instructors typically assess students via traditional high-stakes exams, and, indeed, TQTs were originally invented to help students prepare for such exams . However, in light of well-reasoned critiques of and improvements upon traditional grading systems , TQTs can also support alternative grading systems that include tests, such as specifications (specs) grading and standards-based grading . In these approaches, student work is graded according to whether it meets an expected standard or specification (“spec”) of competence or mastery. Importantly, students whose work initially falls short of the standard are generally given additional chances to meet the standard, which generally necessitates creating additional versions of each test, which can be a significant burden on instructors. However, this burden can be greatly eased if tests are based directly on well-constructed TQTs. Imagine, for example, that students need to pass a “homeostasis standard” by correctly answering four out of five questions about homeostasis on a test. If each test question is based on a separate well-structured TQT, it should be feasible to spin off numerous examples from each TQT to create several versions of the test, all roughly equal in difficulty . Overall, then, by facilitating multiple rounds of core concept usage both in studying and on tests, TQTs can help make the teaching of physiology core concepts more equitable. Even if the promised benefits of our approach are within reach, implementation of conceptual framework-linked TQTs comes with several possible complications. Here we look at a few such complications and possible remedies. Variations in Learning Goals For educators considering the possible use of conceptual framework-linked TQTs, the biggest challenge may be that their goals for their students are different from those represented by – . [As one example, students’ misconceptions about hydrophilic and hydrophobic molecules are a higher priority for us than they are for many other instructors.] Different priorities would then necessitate exploration of different facets of the conceptual frameworks, and/or different conceptual frameworks altogether, as well as different TQTs. However, resources are available to support this work. Those wanting to access the full range of available conceptual frameworks can consult the compilation in Table 3 of Ref. , as well as subsequently published conceptual frameworks for Mass Balance , Structure-Function Relationships , and Biology Systems Thinking . On the TQT side, we are happy to share our existing TQTs (e.g., as archived at tinyurl.com/AnP-TQTs for 200-level courses in human A&P), help “TQT-ize” others’ existing assessment questions, and/or help synthesize new TQTs de novo. Inadequate Pretest Practice with TQTs Collections of TQTs are a kind of study guide but are rather unlike most study guides to which students are accustomed. Therefore, most students will benefit from multiple rounds of explanation and scaffolding. For example, before a test, we sometimes ask students to examine the corresponding test from the previous term and determine the TQT (if any) on which each question was based. On other pretest occasions, we ask students to write their own example questions based on TQTs. The latter assignment typically yields some off-target responses; for example, a student might look at the TQT in and produce an example question such as “What is the set point of the plasma pH?” This would suggest that the student does not yet understand how the existing examples relate to the LLO. With gentle feedback and further practice, most of our students eventually are successful in writing their own examples before the end of the term . TQTs Not Directly Linked to Tests If instructors find value in the TQT format and/or specific existing TQTs, they should, of course, use them according to their best judgment. However, we would caution that if TQTs are simply heaped upon students alongside many other assignments, and if there is not a direct link between TQTs and tests, students may perceive TQTs as low-priority “busy work” and may not give them the attention and investment necessary to yield strong benefits. In contrast, if TQTs are used to directly link active learning practices with summative assessments, students should feel better prepared for, and less anxious about, these assessments. Imperfect Matching of LLOs and Examples TQTs are most useful to students if the example questions always match the LLO perfectly . Although strict adherence to the TQT structure limits the range of questions that can be asked, that narrowing of scope is one of the main goals of TQTs. Students deserve the assurance that actual test questions will conform to the patterns they see in practice. Instructors’ consistent attentiveness to this issue should lower students’ test anxiety and increase their trust in their instructors. Overly Broad LLOs Related to the previous issue is the pitfall of overly broad TQT LLOs. Here is an example of an overly broad LLO: “Given an alteration of the nervous system, predict the functional consequences.” This LLO could potentially spawn example questions about any of the core concepts emphasized here (Homeostasis, Flow Down Gradients, the Cell Membrane, and Cell-Cell Communication), as well as others such as Structure-Function , potentially overwhelming students with possibilities and thus sacrificing depth for breadth. If a TQT does not focus the student’s attention on a somewhat specific kind of problem, it does not guide the student toward any particular concepts or skills, and we are back to the unfortunate default situation of students thinking that anything mentioned in lecture or homework is equally likely to appear on the test. Overly Narrow TQTs One might ask whether it is also problematic for a TQT to be too narrow. In general, this is less of a concern for us; very narrow TQTs can be useful for defining very specific expectations that may arise from very specific conceptual framework subconcepts. For the core concept of mass balance, for example, one could write a simple TQT LLO of “Given a substance’s rates of filtration, reabsorption, and secretion in the kidney, calculate the rate of excretion.” The main concern here is that because the range of problems is so narrow, students may become locked into automatically doing the problem a certain way and thus may learn little about the core concept aside from how to solve this very specific type of problem. The TQT above could be made a bit more interesting by making the LLO a bit broader: “Given three of the following rates for a substance in the kidney—filtration rate, reabsorption rate, secretion rate, and excretion rate—solve for the fourth rate.” The LLO is still clear, but it now allows for more permutations of the question (e.g., across lecture, active learning, and exams) and thus demands a better overall understanding. For educators considering the possible use of conceptual framework-linked TQTs, the biggest challenge may be that their goals for their students are different from those represented by – . [As one example, students’ misconceptions about hydrophilic and hydrophobic molecules are a higher priority for us than they are for many other instructors.] Different priorities would then necessitate exploration of different facets of the conceptual frameworks, and/or different conceptual frameworks altogether, as well as different TQTs. However, resources are available to support this work. Those wanting to access the full range of available conceptual frameworks can consult the compilation in Table 3 of Ref. , as well as subsequently published conceptual frameworks for Mass Balance , Structure-Function Relationships , and Biology Systems Thinking . On the TQT side, we are happy to share our existing TQTs (e.g., as archived at tinyurl.com/AnP-TQTs for 200-level courses in human A&P), help “TQT-ize” others’ existing assessment questions, and/or help synthesize new TQTs de novo. Collections of TQTs are a kind of study guide but are rather unlike most study guides to which students are accustomed. Therefore, most students will benefit from multiple rounds of explanation and scaffolding. For example, before a test, we sometimes ask students to examine the corresponding test from the previous term and determine the TQT (if any) on which each question was based. On other pretest occasions, we ask students to write their own example questions based on TQTs. The latter assignment typically yields some off-target responses; for example, a student might look at the TQT in and produce an example question such as “What is the set point of the plasma pH?” This would suggest that the student does not yet understand how the existing examples relate to the LLO. With gentle feedback and further practice, most of our students eventually are successful in writing their own examples before the end of the term . If instructors find value in the TQT format and/or specific existing TQTs, they should, of course, use them according to their best judgment. However, we would caution that if TQTs are simply heaped upon students alongside many other assignments, and if there is not a direct link between TQTs and tests, students may perceive TQTs as low-priority “busy work” and may not give them the attention and investment necessary to yield strong benefits. In contrast, if TQTs are used to directly link active learning practices with summative assessments, students should feel better prepared for, and less anxious about, these assessments. TQTs are most useful to students if the example questions always match the LLO perfectly . Although strict adherence to the TQT structure limits the range of questions that can be asked, that narrowing of scope is one of the main goals of TQTs. Students deserve the assurance that actual test questions will conform to the patterns they see in practice. Instructors’ consistent attentiveness to this issue should lower students’ test anxiety and increase their trust in their instructors. Related to the previous issue is the pitfall of overly broad TQT LLOs. Here is an example of an overly broad LLO: “Given an alteration of the nervous system, predict the functional consequences.” This LLO could potentially spawn example questions about any of the core concepts emphasized here (Homeostasis, Flow Down Gradients, the Cell Membrane, and Cell-Cell Communication), as well as others such as Structure-Function , potentially overwhelming students with possibilities and thus sacrificing depth for breadth. If a TQT does not focus the student’s attention on a somewhat specific kind of problem, it does not guide the student toward any particular concepts or skills, and we are back to the unfortunate default situation of students thinking that anything mentioned in lecture or homework is equally likely to appear on the test. One might ask whether it is also problematic for a TQT to be too narrow. In general, this is less of a concern for us; very narrow TQTs can be useful for defining very specific expectations that may arise from very specific conceptual framework subconcepts. For the core concept of mass balance, for example, one could write a simple TQT LLO of “Given a substance’s rates of filtration, reabsorption, and secretion in the kidney, calculate the rate of excretion.” The main concern here is that because the range of problems is so narrow, students may become locked into automatically doing the problem a certain way and thus may learn little about the core concept aside from how to solve this very specific type of problem. The TQT above could be made a bit more interesting by making the LLO a bit broader: “Given three of the following rates for a substance in the kidney—filtration rate, reabsorption rate, secretion rate, and excretion rate—solve for the fourth rate.” The LLO is still clear, but it now allows for more permutations of the question (e.g., across lecture, active learning, and exams) and thus demands a better overall understanding. In their 2020 report on physiology core concepts progress over the last ∼15 years, Michael and McFarland proposed that further progress is needed in five specific fronts, including the following: 4. the development of learning resources that specifically require students to use the core concepts and promote active learning; and 5. the development of learning outcomes that address student mastery of core concepts and the ability to apply core concepts to novel physiology problems. TQTs, although not a panacea for all core concept-related challenges, are designed to directly address points 4 and 5 by connecting abstract conceptual frameworks to practical, specific, student-facing questions. The structure of a TQT facilitates the creation of as many examples as are needed for in-class practice, homework, and exams, and thus for both formative and summative assessments. It is important to note that although TQTs have been constructed according to principles (e.g., backward design; Ref. ) for which there is considerable support, our claims regarding TQTs themselves still await empirical validation. We acknowledge that such empirical evidence must be gathered, and our team has begun this work . In the meantime, as a final illustration of the potential value of linking core concepts to TQTs, consider how the opening scenario of this paper might play out in a TQT-infused course. Imagine the following: Wanting to see whether TQTs help your students with physiology core concepts, you reinvigorate your class activities and tests so that TQTs give rise to more short-answer questions reaching higher levels of Bloom’s taxonomy, including a new question comparing calcium homeostasis with glucose homeostasis. So how are the students doing with this revised approach? Student A offers an excellent answer, as usual. (Thank you, student A , for serving as a positive control.) Student B , in active-learning exercises and discussions, initially struggled with the TQT on which this test question is based. However, having worked through their confusion across several examples, and having discovered and eventually rejected a subtle, hindering misconception, student B now recognizes the test question as another variation of this same TQT, understands what is expected, and provides a very good answer, too. And student C ? Well, to be honest, student C is still challenged by the differences between negative feedback and positive feedback. But with several weeks left in the course, and with more examples of this TQT to be explored in the organ systems still ahead, there is a reasonable chance that student C will eventually master it, too. Data will be made available upon reasonable request. G.J.C.’s work on relating TQTs to core concepts was supported by a Teaching Career Enhancement Award (TCEA) from the American Physiological Society and by a Research On STEM Education (ROSE) Fellowship under National Science Foundation (NSF) Grant No. 1826988 (PI Jeff Morris, University of Alabama at Birmingham). G.J.C. also acknowledges valuable advice from ROSE Network project leaders Christina Morra and Penny Carroll. No conflicts of interest, financial or otherwise, are declared by the authors. G.J.C. conceived and designed research; G.J.C. analyzed data; T.A.K. prepared figures; G.J.C. and T.A.K. drafted manuscript; T.A.K. edited and revised manuscript; T.A.K. approved final version of manuscript. |
A paradigm for fostering patient-centered research in liver disease: The liver transplant patient-engagement program | f5e092c6-3ea9-4ff9-9fbe-c3fb99c9c06c | 10027036 | Patient-Centered Care[mh] | As metrics for high-quality care encompass more than just clinical endpoints, it is increasingly important to incorporate patient perspectives into research efforts. In the past decade, patient engagement has become a major cornerstone of research endeavors for funding agencies such as The Patient-centered Outcomes Research Institute and regulatory agencies including the Food and Drug Administration’s patient-focused drug development program. , Despite the enormous value patient perspectives bring to clinical research, patient engagement has been featured in only a handful of studies on liver disease. , Herein, we offer guidance on how to engage patients in grassroots, patient-driven, clinical outcomes research. We adapted previously successful patient-engagement methodologies , to develop a liver transplantation patient-engagement program (LT-PEP) at the University of Texas Southwestern Medical Center. We use liver transplantation (LT) as a model, as LT survivorship goes beyond patient and graft survival to comprise the lived experiences of patients that includes symptoms, functioning, quality of life, and other patient-reported outcomes. , The goal of LT-PEP was to engage the patient community, establish trustworthy bi-directional communications between transplant providers and patients, and foster research collaborations that will directly inform clinical care and improve health outcomes that matter most to LT survivors.
LT-PEP was founded on principles of patient engagement developed by the Patient-centered Outcomes Research Institute and modeled after a Patient Engagement Group established in 2013 at the University of North Carolina for patients with chronic hepatitis C (HCV-Patient Engagement Group). , The goals of LT-PEP were to inform a transplant survivorship research agenda. Our first 5 sessions covered the following topics: Identifying key concepts important to study in LT. Determining the best patient-reported measures to assess these concepts. Identifying barriers and facilitators to conducting clinical research with patients. Reviewing research workflow, incentives, and materials (eg, flyers, consents, and interview guides). We identified 3 essential “P’s” to devising a successful LT-PEP including consideration of the: (1) population (ie, what members of the community should be represented in an advisory panel); (2) principles that should ground patient training and discussions about roles as research advisors versus participant; and (3) participation of advisors in various activities from research planning to dissemination to the stakeholder community (Figure ). Future sessions with advisors will contextualize preliminary data and brainstorm future research projects including interventional needs and development. IRB approval was not required to involve patients as research advisors, given they were hired as consultants with a contract. However, IRB approval was obtained to collect and publish survey and qualitative data describing LT-PEP experiences.
A purposeful sampling strategy was used to select a diverse group of 6 LT survivors—33% were female patients and ages ranged from 30 to 69 years. Patient advisors included white non-Hispanic (66%), Hispanic (17%), and African American (17%) patients; indications for LT included nonalcoholic fatty liver disease (33%), hepatocellular carcinoma (33%), Hepatitis C/alcohol-associated cirrhosis (17%), and acute alcohol-associated hepatitis (17%). Advisors were compensated $50 per hour and signed a memorandum of understanding defining roles and expectations as an advisor. Advisors completed online training sessions available from Patient-centered Outcomes Research Institute (Supplemental Materials, http://links.lww.com/HC9/A135 ). Five, 2-hour virtual sessions were conducted by Sarah R. Lieber and Alvaro Noriega Ramirez over a 3-month period June–September 2022 covering 4 topics outlined. LT-PEP advisors preferred virtual sessions to reduce travel requirements and infectious risk of in-person meetings. After each session, members completed an evaluation assessing quality metrics of participating in LT-PEP (Supplemental Materials, http://links.lww.com/HC9/A135 ). Patients identified several concepts as the most important study outcomes to evaluate in transplant-related research including uncertainty about health, cognition, positivity, financial burden, and caregiver distress (Supplemental Figure 1, http://links.lww.com/HC9/A135 ). Advisors then selected survey instruments that best measured these concepts from the patient's perspective. LT-PEP advisors identified several motivations to participate in research consistent with previously published perspectives: (1) an altruist drive to help future survivors through knowledge gained from research—“desire to see a change in the transplant community” and “giving back and making the transplant experience potentially more successful for future recipients”; (2) learning more about their own progress in recovery by tracking experiences—“help with understanding the situation I am in.” Appealing to these motivations and compensating study participants for their time were recruitment and retention strategies proposed by the LT-PEP. Advisors suggested $25–$50 per 1 hour of research activity plus parking vouchers. Advisors felt that the maximum survey burden should be ~90 minutes (ideally 30-minute sections divided into 2–3 parts). They were willing to stay an additional 60–-90 minutes after a clinic visit to participate in research. Most advisors (83%) preferred answering surveys electronically alone in a private setting, as compared with in-person. Integrating research participation seamlessly into clinical care was deemed essential. Barriers to research participation included: (1) time requirements, (2) illness/hospitalizations, (3) fatigue or other physical symptoms, and (4) obligations to work or family. To overcome barriers to research participation, LT-PEP members recommended frequent phone/text reminders, coordination around clinic/lab appointments, deadlines for data collection (eg, surveys), and reminders about the necessity of their participation to enhance knowledge about LT for the larger transplant community. Postsession assessments demonstrated a high degree of satisfaction with participating in LT-PEP. All “strongly agreed” or “agreed” that the sessions were worth their time, their feedback will improve LT research, and they would participate in future LT-PEP sessions. Advisors were excited to participate, wanted their voices to be heard, and enjoyed connecting with other LT survivors to share stories of recovery. The use of virtual videoconferencing technology facilitated the gathering of LT-PEP advisors in a safe and efficient manner. Extra time was needed to overcome connectivity and technical issues, as well as time for storytelling and socializing, which were critical to fostering connections and trust among members. Future directions include developing patient-engagement programs for other populations including monolingual Spanish speakers, elderly populations, and caregivers.
Standardizing processes for forming patient advisory panels will enhance patient-centered research initiatives across a broad range of gastrointestinal and liver diseases. The HCV-Patient Engagement Group and LT-PEP provide a paradigm for other liver and transplant programs to integrate patient perspectives into research endeavors (Figure ). As highlighted here, patient-engagement programs can prioritize important areas for clinical research, identify key knowledge gaps, inform research methods, solicit patient perspectives on interventions to evaluate in future research, and help disseminate findings to the community. These patient-engagement methodologies have proven effective and can be easily adapted to other patient populations. , Research teams should collaborate with patients as advisors to consult on all stages of the research cycle. Patient engagement is not simply a checkbox but requires meaningful bi-directional patient involvement. As research partners, patient stakeholders add more nuanced perspectives that directly impact patient-centeredness and clinical applicability of the research being performed.
|
The future of | a5792e82-c956-4d9d-a24f-20f56721b9ea | 10027037 | Internal Medicine[mh] | Hepatology Communications is home for all researchers invested in advancing their knowledge of liver disease. As such the journal will be led by co-EiC as partners capable of bridging varied forms of expertise in basic, translational, and clinical science that enrich our research community. Dr Carmen Wong is a basic scientist focused on the molecular pathogenesis of liver cancer aiming to identify new molecular targets for liver cancer treatments. Dr Elliot Tapper is a clinical researcher focused on epidemiology, clinical trials, and quality improvement with the purpose of defining and alleviating the burden of chronic liver disease. Together, our interests and networks are complementary and should provide authors with confidence that their work will be appreciated and treated with the utmost care by relevant experts.
HEPATOLOGY COMMUNICATIONS Mission Hepatology Communications is an open access journal that must define its mission accounting for the roles played by H epatology , Liver Transplantation , and Clinical Liver Disease , in addition to journals from other societies. Given the growing strength and volume of submissions to H epatology , it is likely that the journal could thrive as a source for excellent transfers alone. Our goal, however, is to develop the identity of the journal further. On the clinical side, we aim to provide a destination for hepatology health services research, attracting excellent work showcasing epidemiology, quality improvement, pilot trials, qualitative studies, and health economic inputs. On the basic research side, we aim to provide a home for work revealing the novel mechanisms that affect liver functions and contribute to different liver diseases. We also aim to attract translational work involving preclinical models that provide insights about the new biomarkers and therapeutic targets and treatments. We welcome work with integrated studies involving human specimens and different experimental models to complementarily reveal important processes in the development of liver diseases. We hope to provide a platform for researchers to communicate by sharing their newest discoveries and innovations about liver diseases, pushing the frontiers of knowledge that might potentially lead to new strategies to prevent, detect, and treat liver diseases. We wish for Hepatology Communications to be seen as a source for clear answers describing the state of liver disease as well as the key questions driving the field forward. Peer review The goal of peer review at Hepatology Communications will be to build the best possible work as silent collaborators through rigorous and constructive criticism. Authors should expect clear instruction for what is needed on revision. Our aim is to provide authors a fair, transparent, enjoyable—and an efficient—review process. Dissemination As an open access journal, exposure of your work to diverse readers represents a central strength. The ability to freely download your article means greater impact by reaching researchers, learners, patients, and other stakeholders the world over. You can also expect innovation in dissemination through AASLD networks and social media that has the potential to magnify your impact further. We strive to ensure rapid dissemination of high quality work.
Hepatology Communications is an open access journal that must define its mission accounting for the roles played by H epatology , Liver Transplantation , and Clinical Liver Disease , in addition to journals from other societies. Given the growing strength and volume of submissions to H epatology , it is likely that the journal could thrive as a source for excellent transfers alone. Our goal, however, is to develop the identity of the journal further. On the clinical side, we aim to provide a destination for hepatology health services research, attracting excellent work showcasing epidemiology, quality improvement, pilot trials, qualitative studies, and health economic inputs. On the basic research side, we aim to provide a home for work revealing the novel mechanisms that affect liver functions and contribute to different liver diseases. We also aim to attract translational work involving preclinical models that provide insights about the new biomarkers and therapeutic targets and treatments. We welcome work with integrated studies involving human specimens and different experimental models to complementarily reveal important processes in the development of liver diseases. We hope to provide a platform for researchers to communicate by sharing their newest discoveries and innovations about liver diseases, pushing the frontiers of knowledge that might potentially lead to new strategies to prevent, detect, and treat liver diseases. We wish for Hepatology Communications to be seen as a source for clear answers describing the state of liver disease as well as the key questions driving the field forward.
The goal of peer review at Hepatology Communications will be to build the best possible work as silent collaborators through rigorous and constructive criticism. Authors should expect clear instruction for what is needed on revision. Our aim is to provide authors a fair, transparent, enjoyable—and an efficient—review process.
As an open access journal, exposure of your work to diverse readers represents a central strength. The ability to freely download your article means greater impact by reaching researchers, learners, patients, and other stakeholders the world over. You can also expect innovation in dissemination through AASLD networks and social media that has the potential to magnify your impact further. We strive to ensure rapid dissemination of high quality work.
We expect that Hepatology Communications will grow and change in order to serve your needs. This will result in structural changes for the journal and the submission process. We expect to adapt to your needs over time. For now, we wish to highlight a few key features. Low stress submissions Formatting requirements will not result in the frustrating returns of your manuscript and will be specified for revised manuscripts only. Rapid reviews of revised manuscripts Authors are encouraged to submit revised manuscripts after rejection at other journals along with a point-by-point response to the reviewers. We aim to improve the rate of timely decisions and speed of publication for accepted work while reducing the burden of reviewing for our community. Seamless cascades For authors submitting to Hepatology first, transfers to Hepatology Communications avoid the need for reentering files and manuscript details into a submission website. If the decision is rejection after review, the editors will rapidly inform you of the decision. In most cases, further peer review after responding to Hepatology reviewers will be unnecessary. Research letters We are introducing a new article category for clinical research that focuses on studies with limited scope, limited to 1000 words, with markedly reduced open access fees.
Formatting requirements will not result in the frustrating returns of your manuscript and will be specified for revised manuscripts only.
Authors are encouraged to submit revised manuscripts after rejection at other journals along with a point-by-point response to the reviewers. We aim to improve the rate of timely decisions and speed of publication for accepted work while reducing the burden of reviewing for our community.
For authors submitting to Hepatology first, transfers to Hepatology Communications avoid the need for reentering files and manuscript details into a submission website. If the decision is rejection after review, the editors will rapidly inform you of the decision. In most cases, further peer review after responding to Hepatology reviewers will be unnecessary.
We are introducing a new article category for clinical research that focuses on studies with limited scope, limited to 1000 words, with markedly reduced open access fees.
Humbled to be entrusted with leading Hepatology Communications forward, we are excited to lead the journal to new heights. We cannot wait to read your work and partner with you to push our field forward.
|
Microfocus computed tomography for fetal postmortem imaging: an overview | a92eba54-e41b-461b-92b6-b0422d06e996 | 10027643 | Forensic Medicine[mh] | Fetal autopsy is the modality of choice for diagnosing causes of stillbirth and intrauterine fetal demise (IUFD) and for confirming congenital anomalies. It offers clinically significant findings in approximately 40–70% of cases . Knowing the cause of their loss provides consolation for bereaved parents and offers important information for management of future pregnancies . Over the last years parental consent for autopsy has dropped because parents often view the procedure as too invasive. It is also technically difficult in fetuses at early gestation . To overcome these issues, new and less invasive modalities are being actively proposed. A promising alternative to autopsy is microfocus computed tomography (micro-CT) , a technique that has already made its mark in non-medical industries, e.g., non-destructive precision engineering, ecology and geosciences . Like conventional CT, it is an X-ray-based technology, but instead of a rotating gantry, micro-CT scanners have a fixed radiation source, while the samples are mounted on a rotating platform. The radiation-source-to-sample distance, as well as sample-to-detector distance, can be altered to achieve much higher resolutions, up to sub-micron (<µm) level. In comparison, conventional CT scanners typically have maximum resolutions of 500–1,000 μm. Scan time and radiation dose are typically much higher in micro-CT, clearly making it less suited to imaging live patients, but ideally suited to imaging postmortem fetuses and specimens. Before imaging, the fetus is treated with staining agents to allow visualization of soft tissues, which otherwise offer very little contrast. Staining is frequently done using iodine compounds as a contrast agent, usually by submerging the fetus in the solution. This technique is referred to as diffusible iodine-based contrast-enhanced CT (diceCT) . Because staining time is directly related to diffusion speed, it takes longer in larger fetuses, ranging from hours to several weeks . Because this method is non-destructive, the fetus can be returned to the parents as soon as scanning is complete or after the discoloration caused by the iodine has been reversed. Imaging is preferred by parents as an alternative to autopsy because imaging is less invasive and does not leave disfiguration . The opportunities for fetal imaging and research provided by micro-CT imaging have recently been reviewed . Advances have been made with extensive scan and staining protocols , such as the use of buffered Lugol solution (B-Lugol), which limits the extent of tissue shrinkage when staining fetuses . In this review we provide an overview of the latest research concerning micro-CT imaging of human fetuses, with a special emphasis on diagnostic accuracy, endovascular staining approaches, placental studies and the reversibility of staining. We discuss new methods that could prove suitable for larger fetuses as well as other relevant techniques that might help to forward fetal postmortem imaging. This review covers data that were previously published, so no ethical approval was required.
The current reference standard for diagnosis in a postmortem setting is invasive autopsy. Through dissection and microscopic analyses, organs can be visualized to study the anatomy and pathology and to establish the cause of death. Recent studies comparing fetal micro-CT imaging to autopsy found 93% concordance for overall diagnosis among more than 250 cases; in only 1% of these cases did fetal autopsy provide a diagnosis that could not be established using micro-CT scanning . These findings led to a workflow in which micro-CT scanning was used to triage before autopsy, potentially avoiding autopsy in 87% of cases. Concordance between autopsy and micro-CT was met in most cases, with concordance of 97% across all body systems . The largest discordance was found in evaluating the cardiovascular system, which had a sensitivity of 67%; however, this number might be subject to bias because autopsy was only performed in cases where it was expected to add value . In other studies where the heart was extracted before imaging, the sensitivity of micro-CT in cardiac diagnosis has been higher (85–100%) [ , , ]. In fetuses younger than 16 weeks of gestation and in small or macerated fetuses, autopsy can sometimes be difficult or even impossible because of the technical challenges. For example, the fragile and semifluid fetal brain is generally removed from the skull by submersion in a water bath but remains hard to evaluate. Micro-CT imaging, however, can even be used in embryos (Fig. ) before 8 weeks of gestation and in macerated fetuses . Although at this early stage of gestation, not much is known about pathophysiology because there is a lack of data on these early stages, micro-CT has been found to be a good, if not better, alternative to autopsy for postmortem imaging of fetuses at early gestation [ – , ]. To ascertain a diagnosis, histology can be of added value. However, staining with iodine or other solutions and micro-CT scanning might hamper further histological analysis because of the potential disturbances in tissue integrity. Lupariello et al. used cardiac samples to determine the effects of staining and scanning samples prior to histology. Samples were divided into two groups, either stained with a combination of Lugol solution, methanol and Tween (Thermo Scientific, Waltham, MA) or perfused with Microfil (Flow Tech Inc., S Windsor, CT), an endovascular casting agent. After micro-CT scanning, samples were stained with immunohistochemistry, hematoxylin or Masson trichrome. All samples demonstrated histological staining in all tissues, and the Microfil samples showed a brown substance covering the endothelial layer of the arteries . Although the authors did not find that staining hampered further microscopic evaluation of the tissue, the micro-CT scanning itself appeared to alter proteins in these cardiac samples, possibly the result of thermal damage caused by scanning . The extent to which other proteins are denatured or affected remains unknown. This raises the questions of whether deoxyribonucleic acid (DNA) degradation occurs during micro-CT scanning, which needs be evaluated in future studies. One might consider taking microbiopsies for microscopic evaluation or DNA analysis prior to scanning to prevent iatrogenic thermal alteration of the sample, the extent of which is unknown.
Whole-body staining by submersion To enhance soft-tissue differentiation, staining is needed before scanning. For human fetuses, the most common method to enhance soft-tissue contrast is staining by submerging the fetus in a staining solution . Frequently used staining agents are based on iodine, phosphotungstic acid (PTA), phosphomolybdic acid (PMA) or osmium tetroxide [ , , ]. For human fetuses, the most frequently used staining solution is a water-based solution containing two parts potassium iodide (KI) for every one part iodine (I 2 ), or potassium triiodide (I 2 KI), also called Lugol solution. This solution is known for its rapid and deep penetration, provision of excellent contrast in all tissues and for being non-toxic and relatively cheap, all of which make it a versatile and robust staining agent. Nonetheless, successful staining by submersion in I 2 KI is dependent on several factors, including the size of the fetus, the concentration of the staining solution and the staining time. As fetuses become larger with increasing age, the triiodide must penetrate deeper, requiring longer incubation periods to reach adequate staining of all structures (Fig. ) . Moreover, with increasing gestational age the skin of the fetus becomes less penetrable to iodine, prolonging the staining time. Staining concentration and time are interdependent factors in the staining process. A shorter staining time can be reached with higher concentrations because faster diffusion occurs, although higher concentrations can result in overstaining, causing loss of tissue differentiation . Although some institutions have used unbuffered Lugol solution for fetal staining without encountering significant tissue shrinkage , users should be aware that Lugol solution can cause tissue shrinkage of up to 30% [ , , ]. This negative effect on the tissue was long thought to be the result of an osmotic imbalance between the staining solution and the fetus; however, the use of an isotonic staining solution did not prevent tissue shrinkage, thus demonstrating that osmotic imbalance is not the driving factor. Dawood et al. demonstrated that acidification of the solution is the key factor in soft-tissue shrinkage rather than the osmolarity of the staining solution . Moreover, the authors found that staining in Lugol solution prepared with a buffer (B-Lugol) stabilizes the pH and almost completely prevents soft-tissue shrinkage, without affecting the staining time . Endovascular diffusion staining Whole-body staining by submersion has proved effective for excised tissues and small whole-body fetuses (< 20 weeks of gestation). However, as mentioned, complete penetration of the staining agent takes longer when there is increased distance between surface and center of the fetus and requires large volumes of submersion fluid in bigger, i.e. older fetuses. Moreover, the increased skin density and ossified structures (e.g., the skull) may form barriers and hamper diffusion in older fetuses. This is particularly evident when aiming at complete staining of the brain in older fetuses in cases where the skull is formed. Depending on the size of the fetus, staining time ranges from a couple of days to several weeks. When providing a postmortem scanning service as an alternative for conventional autopsy, this delay is undesirable. Endovascular infusion of the contrast agent offers a possible alternative for staining by submersion to accelerate uptake and direct delivery to peripheral tissues. In theory, this would create a larger surface area and bypass hard-to-penetrate structures such as the skin and skull. To the best of our knowledge, this technique has not been used for whole-body staining of human fetuses to date. However, it has been tried in animal experiments using mice [ – ] and newborn piglets . In mice, contrast agents were either administered into the left ventricle of the heart or retrograde through the aorta . Before administration of the contrast agent, the mice were heparinized to prevent blood clotting . After infusing the staining agent, the fetuses were fixed . Staining agents PTA and Lugol were tested, and Lugol was found to have a faster uptake (15 min vs. 30 min stain time) and showed more homogeneous staining throughout the body compared to PTA. This may be because triiodide is a smaller molecule than PTA. Lugol also resulted in better contrast in the myocardium and bronchial walls than PTA. On the other hand, PTA appeared to have more uptake in the liver, renal parenchyma and vessel walls. So, if the aim of the staining is to visualize the vasculature, then PTA would be the preferred contrast agent . Schweitzer and coworkers recently presented the first results using diluted barium sulfate solution in combination with a high-pressure (max. 1.4 bar, max. 22 L/min) angiography pump for venous and arterial filling in postmortem newborn piglets. This procedure results in staining of the entire fetus without macroscopically visible discoloration of the tissues . In their studies, adequate staining could be achieved even several days after the animals’ demise while they were stored in a cool environment. These studies approach the needs of a clinical setting with delayed staining, for example if parents wanted some time with their child before diagnostics or there were logistical delays to staining. Time between demise and staining, however, should be kept short to prevent too much clotting of the blood and decay of the body. Only one study reported on the comparison and combined administration of Lugol through submersion and endovascular infusion. Fixed mice were either injected with Lugol in the right ventricle or submerged in Lugol after their skin was incised, in both cases using a similar concentration of Lugol solution; the endovascular approach failed to deliver contrast because the staining agent remained within the heart and was not distributed to the body . This is likely the consequence of specimen fixation prior to infusion of the agent and the lack of perfusion with heparin to prevent clotting. Endovascular delivery should be further explored because it might dramatically reduce staining time . However, it becomes increasingly technically difficult to administer a staining agent in smaller vessels, and for this reason submersion is likely to remain the staining method of choice for the smallest fetuses. Intravascular administration of various staining agents and concentrations might provide new ways of visualizing anatomy of large fetuses and thereby form a basis for future studies, although any interference with the body including needling or injection might be unacceptable to parents.
To enhance soft-tissue differentiation, staining is needed before scanning. For human fetuses, the most common method to enhance soft-tissue contrast is staining by submerging the fetus in a staining solution . Frequently used staining agents are based on iodine, phosphotungstic acid (PTA), phosphomolybdic acid (PMA) or osmium tetroxide [ , , ]. For human fetuses, the most frequently used staining solution is a water-based solution containing two parts potassium iodide (KI) for every one part iodine (I 2 ), or potassium triiodide (I 2 KI), also called Lugol solution. This solution is known for its rapid and deep penetration, provision of excellent contrast in all tissues and for being non-toxic and relatively cheap, all of which make it a versatile and robust staining agent. Nonetheless, successful staining by submersion in I 2 KI is dependent on several factors, including the size of the fetus, the concentration of the staining solution and the staining time. As fetuses become larger with increasing age, the triiodide must penetrate deeper, requiring longer incubation periods to reach adequate staining of all structures (Fig. ) . Moreover, with increasing gestational age the skin of the fetus becomes less penetrable to iodine, prolonging the staining time. Staining concentration and time are interdependent factors in the staining process. A shorter staining time can be reached with higher concentrations because faster diffusion occurs, although higher concentrations can result in overstaining, causing loss of tissue differentiation . Although some institutions have used unbuffered Lugol solution for fetal staining without encountering significant tissue shrinkage , users should be aware that Lugol solution can cause tissue shrinkage of up to 30% [ , , ]. This negative effect on the tissue was long thought to be the result of an osmotic imbalance between the staining solution and the fetus; however, the use of an isotonic staining solution did not prevent tissue shrinkage, thus demonstrating that osmotic imbalance is not the driving factor. Dawood et al. demonstrated that acidification of the solution is the key factor in soft-tissue shrinkage rather than the osmolarity of the staining solution . Moreover, the authors found that staining in Lugol solution prepared with a buffer (B-Lugol) stabilizes the pH and almost completely prevents soft-tissue shrinkage, without affecting the staining time .
Whole-body staining by submersion has proved effective for excised tissues and small whole-body fetuses (< 20 weeks of gestation). However, as mentioned, complete penetration of the staining agent takes longer when there is increased distance between surface and center of the fetus and requires large volumes of submersion fluid in bigger, i.e. older fetuses. Moreover, the increased skin density and ossified structures (e.g., the skull) may form barriers and hamper diffusion in older fetuses. This is particularly evident when aiming at complete staining of the brain in older fetuses in cases where the skull is formed. Depending on the size of the fetus, staining time ranges from a couple of days to several weeks. When providing a postmortem scanning service as an alternative for conventional autopsy, this delay is undesirable. Endovascular infusion of the contrast agent offers a possible alternative for staining by submersion to accelerate uptake and direct delivery to peripheral tissues. In theory, this would create a larger surface area and bypass hard-to-penetrate structures such as the skin and skull. To the best of our knowledge, this technique has not been used for whole-body staining of human fetuses to date. However, it has been tried in animal experiments using mice [ – ] and newborn piglets . In mice, contrast agents were either administered into the left ventricle of the heart or retrograde through the aorta . Before administration of the contrast agent, the mice were heparinized to prevent blood clotting . After infusing the staining agent, the fetuses were fixed . Staining agents PTA and Lugol were tested, and Lugol was found to have a faster uptake (15 min vs. 30 min stain time) and showed more homogeneous staining throughout the body compared to PTA. This may be because triiodide is a smaller molecule than PTA. Lugol also resulted in better contrast in the myocardium and bronchial walls than PTA. On the other hand, PTA appeared to have more uptake in the liver, renal parenchyma and vessel walls. So, if the aim of the staining is to visualize the vasculature, then PTA would be the preferred contrast agent . Schweitzer and coworkers recently presented the first results using diluted barium sulfate solution in combination with a high-pressure (max. 1.4 bar, max. 22 L/min) angiography pump for venous and arterial filling in postmortem newborn piglets. This procedure results in staining of the entire fetus without macroscopically visible discoloration of the tissues . In their studies, adequate staining could be achieved even several days after the animals’ demise while they were stored in a cool environment. These studies approach the needs of a clinical setting with delayed staining, for example if parents wanted some time with their child before diagnostics or there were logistical delays to staining. Time between demise and staining, however, should be kept short to prevent too much clotting of the blood and decay of the body. Only one study reported on the comparison and combined administration of Lugol through submersion and endovascular infusion. Fixed mice were either injected with Lugol in the right ventricle or submerged in Lugol after their skin was incised, in both cases using a similar concentration of Lugol solution; the endovascular approach failed to deliver contrast because the staining agent remained within the heart and was not distributed to the body . This is likely the consequence of specimen fixation prior to infusion of the agent and the lack of perfusion with heparin to prevent clotting. Endovascular delivery should be further explored because it might dramatically reduce staining time . However, it becomes increasingly technically difficult to administer a staining agent in smaller vessels, and for this reason submersion is likely to remain the staining method of choice for the smallest fetuses. Intravascular administration of various staining agents and concentrations might provide new ways of visualizing anatomy of large fetuses and thereby form a basis for future studies, although any interference with the body including needling or injection might be unacceptable to parents.
The placenta is the chief regulator of nutrient supply for the fetus, and placental disorders such as pre-eclampsia can impact the health of both mother and fetus. Pathogenesis of pre-eclampsia is not fully understood, and by visualizing the smallest vasculature of the placenta, new light could be shed on this problem and advance our understanding of such diseases. Two umbilical arteries and one umbilical vein provide the blood circulation of the fetus. They branch out in the placenta to form a complex network that is anatomically variable and is, even in healthy placentas, not fully understood. It has been investigated using micro-CT . For intravascular and placental research, Microfil and BriteVu (Scarlet Imaging, Murray, UT) are two frequently used casting agents . These agents are infused into capillaries while in liquid form and are allowed to solidify into a highly dense intravascular cast that can be visualized using various medical imaging techniques, micro-CT among them. The contrast agent does not leave the vessel and does not cause staining of soft tissue and therefore provides visualization exclusively of the vascular anatomy . In general, care must be taken that the tissue is not perfused with a pressure higher than the physiological blood pressure because this leads to damage of the vessels and overestimation of vascular volume from vessel expansion. Microfil is a lead-based substance designed for visualizing microcirculation of vessels larger than 100 microns . Before administration, the vessels must be flushed with sodium chloride and perfused with heparin to prevent clotting for maximal perfusion of the substance. Aughwane and coworkers infused Microfil into term placentas and found that only half of the vessels larger than 200 µm 2 were perfused adequately (more than 75% filling), which might indicate that Microfil is less suitable than anticipated to visualize the microcirculation. BriteVu is a barium sulfate–based substance that is less viscous than Microfil . Using BriteVu, James and coworkers were able to visualize capillary vessels down to 10 μm wide. From their study, it is not clear whether BriteVu reached the vasculature through the entire fetus because only specific areas are shown. In detailed images (Fig. ), the vessels appear to be well perfused, supporting their findings, and one might anticipate that the entire vasculature of the fetus can be reached . These casting agents have been used to study the microvasculature of the placenta, but no researchers have been able to scan a complete placenta at the resolution needed to image the 10-µm wide micro-vessels . For vascular research it remains questionable whether it is necessary to visualize the smallest vessels, considering their anatomical variability. Depending on the expected size of the studied entity, intravascular casting agents could be an asset in future studies. Note, however, that endovascular casts might interfere with (subsequent) soft-tissue staining, especially in endovascular approaches.
After iodine staining of fetuses, there is potential skin discoloration to a brown color, which might be undesirable to parents if they are not forewarned that this can happen, or if scanning is done on rare museum specimens as an alternative for the inevitable destructive effects of dissection. Lanzetti and coworkers found that in animal museum specimens that were preserved in 70% ethanol and stained using 1% iodine in 70% ethanol, the discoloration could be reversed using 3% sodium thiosulfate (STS) in 70% ethanol within hours to a few days (depending on fetal size) . Figure shows an example of a very rare humpback whale embryo that was stained and de-stained according to this protocol. Although specimens are de-stained on the outside, they might still be stained internally . When scanning postmortem human fetuses, superficial de-staining might be sufficient for parents if this accelerates the process of returning the body to the family. For rare museum specimens this might not be sufficient and requires future studies. Moreover, it should be noted that STS does not restore a fetus to its earlier chemical state because STS reacts with the brownish triiodide and reduces it to the colorless iodide. To remove the iodide from the iodine stained specimen, further incubation steps are required, though it should be considered that these additional incubation steps might alter the fetus because of an osmotic imbalance between the incubation solution and the tissue. This process, called leaching, may easily take several weeks or months, depending on the size and volume of the fetus, and requires regular handling to refresh the solution as it saturates with iodide . Longer de-staining times are, however, less of an issue in the setting of museum specimens. The notion that ethanol-preserved specimens can be stained, scanned and de-stained without causing damage might provide an opportunity to scan even rare pathological human fetal specimens and thus give more insight in human fetal development by utilizing rare collections.
Over the last decade a considerable number of studies have been published regarding micro-CT imaging of human fetuses. The use of micro-CT offers high diagnostic accuracy in the clinical setting comparable to that of a classic autopsy. It should be noted that even after autopsy, 55% of intrauterine fetal demise cases are still left unexplained , but this information can help counsel parents appropriately following pregnancy loss. It is to be expected that the diagnostic accuracy of fetal micro-CT imaging will increase in the coming years because of advancing techniques and protocols, improved scan quality and the growing experience of developmental biologists, radiologists and radiographers. With further research, we anticipate that contrast-enhanced micro-CT will become a suitable alternative to fetal autopsy at early gestation and for rare museum specimens. Moreover, the acquired micro-CT images can be made available worldwide to be studied and rendered into three-dimensional models for both research and wider teaching.
|
Family physicians' involvement in palliative cancer care | fbd081e1-b096-41df-944e-226472ea74e9 | 10028020 | Internal Medicine[mh] | INTRODUCTION Early outpatient palliative care improves quality of life, symptom control, and satisfaction with care for patients with advanced cancer. , , While clinical trials of early palliative care have predominantly examined specialized palliative care interventions, it is increasingly recognized that a sustainable model requires both generalist and specialist palliative care. Family physicians (FPs) are ideally positioned to provide basic palliative care, with specialized palliative care physicians providing care in more complex situations. , , , FPs have long‐term relationships with patients and their families, which enable continuity of care and smooth transitions to end‐of‐life care and facilitate provision of psychosocial support and eventual bereavement care. , , , , However, receiving oncological care at a cancer center may supplant FPs' care provision. In addition, an increasing number of cancer centers have outpatient palliative care clinics, where patients receive specialized symptom control, advance care planning, and emotional support. , Although the intent of these clinics is to provide specialized palliative care in collaboration with FPs and oncologists, it is possible that longitudinal follow‐up in the outpatient palliative care clinic could compromise the receipt of palliative care by FPs. Factors that may facilitate or impede FP involvement in palliative care for patients with advanced cancer can be conceptualized using Andersen's Behavioral Model of Health Services Use, which describes four main groups of factors determining access to and use of healthcare services. , Predisposing factors include patient‐related factors, such as demographic characteristics, that affect the use of healthcare services. Enabling factors are conditions, such as transportation and travel time, that facilitate or impede use of services. Need factors include conditions that reflect a requirement for medical treatment, such as the diagnosis of a medical illness. Outcome factors include patient‐reported outcomes, such as symptom control and satisfaction with care; these outcomes are influenced by the predisposing, enabling and need factors and, in turn, affect subsequent healthcare use. Although the perspective of FPs on their role in providing cancer palliative care has been well described, , , , , , , , research on patients' perspectives of FP involvement in their palliative care consists mostly of small qualitative studies. The aims of the present study were to describe patient‐reported involvement of FPs in their palliative care and to identify factors associated with such involvement among patients with advanced cancer referred to an outpatient palliative care clinic.
METHODS 2.1 Participants Participants were patients with cancer who were attending the Oncology Palliative Care Clinic (OPCC) of the Princess Margaret Cancer Centre, a large integrated cancer treatment, teaching, and research center in Toronto, Canada. The OPCC offers specialized palliative care services to outpatients with cancer, receiving approximately 1500 new referrals annually. To avoid selection bias, eligible patients were identified through screening of daily clinic patient lists for the OPCC. Inclusion criteria included attending the OPCC; age ≥18 years; ability to understand English sufficiently to provide informed consent and complete the study questionnaires; currently have, or previously had, an FP in the 5 years prior to recruitment; and physical and cognitive capacities to participate, according to their palliative care physician. 2.2 Materials Participants' sociodemographic and medical characteristics were extracted from their medical charts using a chart review form developed for the study. A survey was developed to assess patients' perspectives of their FPs' involvement in their care. The survey comprised 25 multi‐component questions with a completion time of approximately 20 minutes. Two categorical survey items formed the two primary study outcomes: (1) having seen the FP for palliative care services (symptom control, advance care planning, and emotional care for cancer) in the last 6 months (versus ≥7 months or never); and (2) having scheduled or planning to schedule an appointment with the FP (versus not having scheduled and not planning to schedule an appointment). Two questions addressed from whom patients currently received and would prefer to receive various cancer and general medical services. The remaining questions included items regarding predisposing factors (e.g. patient demographics), enabling factors (e.g. length of time with FP, travel time to FP's office, provision of home visits), and outcome factors (e.g. perception of medical care). For perceptions of medical care by the FP, seven survey items were rated from 0 (strongly disagree) to 4 (strongly agree): satisfaction with the FP's care; confidence in receiving the best care from the FP; feeling that the FP's role in their care was clear; feeling that the FP provided sufficient time to address their problems; feeling that the FP knew them as a person; receipt of prompt care from their FP; and availability of the FP for desired services. Given the mostly high intercorrelations among these seven items (Spearman's ρ = 0.46–0.87) and high internal consistency (Cronbach's α = 0.92), their ratings were summed to create a total score reflecting patients' perception of medical care for use in analyses. An additional item—feeling that it was easy to reach the FP or their team after hours—was included separately to reflect a specific, important aspect of palliative care and was rated on the same 0 to 4 scale. , This item demonstrated lower correlations with the seven perceptions of medical care items (Spearman's ρ = 0.19–0.39). Additional outcome factors, which could also be conceptualized as need factors, were the Patient‐Reported Functional Status (PRFS) measure, a validated adaptation of the clinician‐rated ECOG and the Revised Edmonton Symptom Assessment System‐CS (ESASr‐CS) scale. The latter scale measures the severity of 11 common cancer symptoms: pain, tiredness, drowsiness, nausea, appetite, shortness of breath, depression, anxiety, well‐being, constipation, and sleep. Average severity of each symptom over the last 24 h was rated from 0 (best) to 10 (worst). A total summed score was calculated, as well as a physical subscale (pain, tiredness, drowsiness, nausea, appetite, shortness of breath, constipation, and sleep) and an emotional subscale (depression and anxiety). A chart review was conducted to abstract additional need factors including cancer diagnosis and number of previous visits to the OPCC. 2.3 Procedure The study received approval from the University Health Network Research Ethics Board (REB #16‐5061‐CE). Participant recruitment in the OPCC was conducted between May 2016 and August 2018. Study staff approached the attending physicians or nurses to confirm eligible patients' physical and cognitive abilities to participate and then approached patients about participating in the study. After receiving information about the study, patients willing to participate provided written informed consent. They either completed the survey immediately or took it home to complete and return by mail using a provided addressed, stamped envelope. 2.4 Statistical methodology The target sample size was 250 participants. Based on our clinical experience with the study population, we estimated that 25% of patients would have seen a FP for palliative care in the last 6 months; thus, with 250 patients, we would estimate that about 63 participants (250 × 25%) would have seen their family physician for palliative care in the last 6 months. Given the minimal number of events per predictor variable (EPV) of ≥10 that would result in valid regression coefficients, we would be able to accommodate five to six variables in a multivariate logistic regression analysis. Descriptive statistics were used for patients' demographic and medical characteristics and the characteristics of their FP and their care services. Comparisons of groups within each outcome on patients' demographic and medical characteristics were conducted using chi‐square tests, Cochran‐Armitage trend test, and t ‐tests. Univariable binary logistic regression was used to determine the relationship between positive perception of medical care and each outcome; this was done for the seven perception items individually and for the single score reflecting overall perception of the FP's medical care. Binary logistic regression analyses were conducted to identify factors associated with the two outcomes. Candidate factors included the following variables from the survey, classified according to Andersen's model , : (1) predisposing patient factors: age, sex, marital status, ethnic background, first language, and education level; (2) enabling factors: duration of relationship with the FP, travel time to FP's office, provision of home visits during or after office hours, provision of 24‐h telephone support, and ease of reaching the FP after hours; (3) need factors: time since diagnosis, current oncology treatment, and number of previous OPCC visits; and (4) outcome factors: symptom severity (ESASr‐CS total and physical and emotional subscales), perception of medical care total score, and PRFS rating. Factors that were associated with each outcome at p < 0.25 were entered into the respective stepwise selection procedures; factors with p < 0.05 were retained in the multivariable logistic regression model. Odds ratios (OR) and 95% confidence intervals (CI) for significant factors were reported. Analyses were performed on all available data and conducted using SPSS version 25 and SAS version 9.4. The significance level was set to 0.05.
Participants Participants were patients with cancer who were attending the Oncology Palliative Care Clinic (OPCC) of the Princess Margaret Cancer Centre, a large integrated cancer treatment, teaching, and research center in Toronto, Canada. The OPCC offers specialized palliative care services to outpatients with cancer, receiving approximately 1500 new referrals annually. To avoid selection bias, eligible patients were identified through screening of daily clinic patient lists for the OPCC. Inclusion criteria included attending the OPCC; age ≥18 years; ability to understand English sufficiently to provide informed consent and complete the study questionnaires; currently have, or previously had, an FP in the 5 years prior to recruitment; and physical and cognitive capacities to participate, according to their palliative care physician.
Materials Participants' sociodemographic and medical characteristics were extracted from their medical charts using a chart review form developed for the study. A survey was developed to assess patients' perspectives of their FPs' involvement in their care. The survey comprised 25 multi‐component questions with a completion time of approximately 20 minutes. Two categorical survey items formed the two primary study outcomes: (1) having seen the FP for palliative care services (symptom control, advance care planning, and emotional care for cancer) in the last 6 months (versus ≥7 months or never); and (2) having scheduled or planning to schedule an appointment with the FP (versus not having scheduled and not planning to schedule an appointment). Two questions addressed from whom patients currently received and would prefer to receive various cancer and general medical services. The remaining questions included items regarding predisposing factors (e.g. patient demographics), enabling factors (e.g. length of time with FP, travel time to FP's office, provision of home visits), and outcome factors (e.g. perception of medical care). For perceptions of medical care by the FP, seven survey items were rated from 0 (strongly disagree) to 4 (strongly agree): satisfaction with the FP's care; confidence in receiving the best care from the FP; feeling that the FP's role in their care was clear; feeling that the FP provided sufficient time to address their problems; feeling that the FP knew them as a person; receipt of prompt care from their FP; and availability of the FP for desired services. Given the mostly high intercorrelations among these seven items (Spearman's ρ = 0.46–0.87) and high internal consistency (Cronbach's α = 0.92), their ratings were summed to create a total score reflecting patients' perception of medical care for use in analyses. An additional item—feeling that it was easy to reach the FP or their team after hours—was included separately to reflect a specific, important aspect of palliative care and was rated on the same 0 to 4 scale. , This item demonstrated lower correlations with the seven perceptions of medical care items (Spearman's ρ = 0.19–0.39). Additional outcome factors, which could also be conceptualized as need factors, were the Patient‐Reported Functional Status (PRFS) measure, a validated adaptation of the clinician‐rated ECOG and the Revised Edmonton Symptom Assessment System‐CS (ESASr‐CS) scale. The latter scale measures the severity of 11 common cancer symptoms: pain, tiredness, drowsiness, nausea, appetite, shortness of breath, depression, anxiety, well‐being, constipation, and sleep. Average severity of each symptom over the last 24 h was rated from 0 (best) to 10 (worst). A total summed score was calculated, as well as a physical subscale (pain, tiredness, drowsiness, nausea, appetite, shortness of breath, constipation, and sleep) and an emotional subscale (depression and anxiety). A chart review was conducted to abstract additional need factors including cancer diagnosis and number of previous visits to the OPCC.
Procedure The study received approval from the University Health Network Research Ethics Board (REB #16‐5061‐CE). Participant recruitment in the OPCC was conducted between May 2016 and August 2018. Study staff approached the attending physicians or nurses to confirm eligible patients' physical and cognitive abilities to participate and then approached patients about participating in the study. After receiving information about the study, patients willing to participate provided written informed consent. They either completed the survey immediately or took it home to complete and return by mail using a provided addressed, stamped envelope.
Statistical methodology The target sample size was 250 participants. Based on our clinical experience with the study population, we estimated that 25% of patients would have seen a FP for palliative care in the last 6 months; thus, with 250 patients, we would estimate that about 63 participants (250 × 25%) would have seen their family physician for palliative care in the last 6 months. Given the minimal number of events per predictor variable (EPV) of ≥10 that would result in valid regression coefficients, we would be able to accommodate five to six variables in a multivariate logistic regression analysis. Descriptive statistics were used for patients' demographic and medical characteristics and the characteristics of their FP and their care services. Comparisons of groups within each outcome on patients' demographic and medical characteristics were conducted using chi‐square tests, Cochran‐Armitage trend test, and t ‐tests. Univariable binary logistic regression was used to determine the relationship between positive perception of medical care and each outcome; this was done for the seven perception items individually and for the single score reflecting overall perception of the FP's medical care. Binary logistic regression analyses were conducted to identify factors associated with the two outcomes. Candidate factors included the following variables from the survey, classified according to Andersen's model , : (1) predisposing patient factors: age, sex, marital status, ethnic background, first language, and education level; (2) enabling factors: duration of relationship with the FP, travel time to FP's office, provision of home visits during or after office hours, provision of 24‐h telephone support, and ease of reaching the FP after hours; (3) need factors: time since diagnosis, current oncology treatment, and number of previous OPCC visits; and (4) outcome factors: symptom severity (ESASr‐CS total and physical and emotional subscales), perception of medical care total score, and PRFS rating. Factors that were associated with each outcome at p < 0.25 were entered into the respective stepwise selection procedures; factors with p < 0.05 were retained in the multivariable logistic regression model. Odds ratios (OR) and 95% confidence intervals (CI) for significant factors were reported. Analyses were performed on all available data and conducted using SPSS version 25 and SAS version 9.4. The significance level was set to 0.05.
RESULTS 3.1 Participant characteristics A total of 832 patients were screened. Of these, 289 (34.7%) were ineligible (169 [20.3%] had a language barrier, 55 [6.6%] were too ill to approach, 23 [2.8%] had a cognitive deficit, 17 [2.0%] had no FP, and, for 25 [3.0%], clinicians asked that patients not be approached). The remaining 543 eligible patients (65.3%) were approached at least once. Of these, 280 (51.6%) declined participation (150 [27.6%] were not interested, 80 [14.7%] felt the study would take too much time or be burdensome, 44 [8.1%] were not feeling well, and 6 [1.1%] expressed dissatisfaction with their FP). The remaining 263 (48.4%) consented to the study, and 258 (47.5%) completed the survey. Of those who responded, 89/253 (35.2%; five missing responses) reported having seen their FP for palliative care in the last 6 months, whereas 130/254 (51.2%; four missing responses) had scheduled or planned to schedule an appointment with their FP. Table summarizes the demographic and medical characteristics of the 258 participants by each of these two outcomes. There were no significant differences in demographic or medical characteristics between those who had or had not seen their FP for palliative care within the last 6 months. Participants who had a scheduled or planned appointment with the FP were less likely to be female ( p = 0.02) and more likely to speak English as a first language ( p = 0.02) than those who did not have a scheduled or planned appointment. Table summarizes enabling factors for patients' visits to the FP. More than 60% of patients indicated that they had been with their FP for more than 5 years, and almost half for more than 10 years. Few patients indicated that their FP offered home visits during (23/256, 9.0%) or after office hours (21/255, 8.2%) or 24‐h telephone support services (54/255, 21.2%). The majority (147/246, 59.8%) disagreed or strongly disagreed that it was easy to reach their FP or the FP's team after hours. 3.2 Current and preferred providers of medical care Table identifies the different clinicians currently involved in provision of non‐cancer and cancer‐related medical care and patients' preferences for providers of such care. FPs were the main providers of non‐cancer related acute care (193/254, 76.0%) and non‐cancer chronic medical management (176/254, 69.3%) and were also the preferred healthcare providers for both types of care (218/252, 86.5%, and 221/253, 87.4%, respectively). In contrast, few FPs were involved in cancer‐related care, including coordination of cancer care and pain and symptom management for cancer and cancer treatments (0.8%–2.0%), nor were they the preferred providers for such care (2.0%–4.1%). Rather, palliative care clinicians tended to be the actual and preferred providers of cancer‐related pain and symptom management, whereas oncologists were most commonly the actual and preferred providers of care for cancer treatment‐related symptoms. Few patients indicated that FPs provided psychosocial palliative care services, including emotional care, advance care directives, support for caregivers and family members, and arrangement of home‐care services (2.8%–6.0%), although a somewhat larger proportion of patients preferred FPs to provide such care (12.6%–27.6%). Compared to FPs and oncologists, the largest proportions of patients reported that palliative care clinicians were their actual (10.7%–29.6%) and preferred (57.8%–68.9%) providers of psychosocial care. 3.3 Perception of FP 's medical care by previous and planned visits to the FP Table summarizes the number of participants who agreed or strongly agreed with statements of positive perceptions of care received from their FP. Patients were most likely to endorse the item “knows me as a person” and least likely to endorse the item “able to provide the time I need to address all of my problems.” Almost all statements were more likely to be endorsed by those who had visited their FP for palliative care in the past 6 months than by those who had not. Feeling that the FP provided sufficient time to address problems was the only statement that was significantly more likely to be endorsed by those who had a scheduled or planned appointment with their FP than by those who did not (OR = 1.29, 95% CI = 1.07–1.55, p = 0.01). The total score, reflecting positive perceptions of overall medical care, was significantly associated with both having visited the FP for palliative care in the last 6 months (OR = 1.06, 95% CI = 1.02–1.11, p = 0.002) and having a scheduled or planned appointment with the FP (OR = 1.03, 95% CI = 1.00–1.07, p = 0.049). 3.4 Multivariable factors associated with previous and planned visits to the FP Results of the multivariable analyses are shown in Table . Three FP factors remained negatively or positively associated with having visited the FP for palliative care in the past 6 months: travel time to the FP (OR = 0.67, 95% CI = 0.48–0.93, p = 0.02), indicating that the FP offers 24‐h telephone support (OR = 1.96, 95% CI = 1.02–3.76, p = 0.04), and better perceived care by the FP (OR = 1.05, 95% CI = 1.01–1.09, p = 0.01). The following variables were negatively or positively associated with having a scheduled appointment with the FP: female sex (OR = 0.51, 95% CI = 0.30–0.87, p = 0.01), English as first language (OR = 2.90, 95% CI = 1.04–8.11, p = 0.04), travel time to the FP (OR = 0.66, 95% CI = 0.47–0.93, p = 0.02), and ease of reaching the FP after hours (OR = 1.33, 95% CI = 1.08–1.64, p = 0.008).
Participant characteristics A total of 832 patients were screened. Of these, 289 (34.7%) were ineligible (169 [20.3%] had a language barrier, 55 [6.6%] were too ill to approach, 23 [2.8%] had a cognitive deficit, 17 [2.0%] had no FP, and, for 25 [3.0%], clinicians asked that patients not be approached). The remaining 543 eligible patients (65.3%) were approached at least once. Of these, 280 (51.6%) declined participation (150 [27.6%] were not interested, 80 [14.7%] felt the study would take too much time or be burdensome, 44 [8.1%] were not feeling well, and 6 [1.1%] expressed dissatisfaction with their FP). The remaining 263 (48.4%) consented to the study, and 258 (47.5%) completed the survey. Of those who responded, 89/253 (35.2%; five missing responses) reported having seen their FP for palliative care in the last 6 months, whereas 130/254 (51.2%; four missing responses) had scheduled or planned to schedule an appointment with their FP. Table summarizes the demographic and medical characteristics of the 258 participants by each of these two outcomes. There were no significant differences in demographic or medical characteristics between those who had or had not seen their FP for palliative care within the last 6 months. Participants who had a scheduled or planned appointment with the FP were less likely to be female ( p = 0.02) and more likely to speak English as a first language ( p = 0.02) than those who did not have a scheduled or planned appointment. Table summarizes enabling factors for patients' visits to the FP. More than 60% of patients indicated that they had been with their FP for more than 5 years, and almost half for more than 10 years. Few patients indicated that their FP offered home visits during (23/256, 9.0%) or after office hours (21/255, 8.2%) or 24‐h telephone support services (54/255, 21.2%). The majority (147/246, 59.8%) disagreed or strongly disagreed that it was easy to reach their FP or the FP's team after hours.
Current and preferred providers of medical care Table identifies the different clinicians currently involved in provision of non‐cancer and cancer‐related medical care and patients' preferences for providers of such care. FPs were the main providers of non‐cancer related acute care (193/254, 76.0%) and non‐cancer chronic medical management (176/254, 69.3%) and were also the preferred healthcare providers for both types of care (218/252, 86.5%, and 221/253, 87.4%, respectively). In contrast, few FPs were involved in cancer‐related care, including coordination of cancer care and pain and symptom management for cancer and cancer treatments (0.8%–2.0%), nor were they the preferred providers for such care (2.0%–4.1%). Rather, palliative care clinicians tended to be the actual and preferred providers of cancer‐related pain and symptom management, whereas oncologists were most commonly the actual and preferred providers of care for cancer treatment‐related symptoms. Few patients indicated that FPs provided psychosocial palliative care services, including emotional care, advance care directives, support for caregivers and family members, and arrangement of home‐care services (2.8%–6.0%), although a somewhat larger proportion of patients preferred FPs to provide such care (12.6%–27.6%). Compared to FPs and oncologists, the largest proportions of patients reported that palliative care clinicians were their actual (10.7%–29.6%) and preferred (57.8%–68.9%) providers of psychosocial care.
Perception of FP 's medical care by previous and planned visits to the FP Table summarizes the number of participants who agreed or strongly agreed with statements of positive perceptions of care received from their FP. Patients were most likely to endorse the item “knows me as a person” and least likely to endorse the item “able to provide the time I need to address all of my problems.” Almost all statements were more likely to be endorsed by those who had visited their FP for palliative care in the past 6 months than by those who had not. Feeling that the FP provided sufficient time to address problems was the only statement that was significantly more likely to be endorsed by those who had a scheduled or planned appointment with their FP than by those who did not (OR = 1.29, 95% CI = 1.07–1.55, p = 0.01). The total score, reflecting positive perceptions of overall medical care, was significantly associated with both having visited the FP for palliative care in the last 6 months (OR = 1.06, 95% CI = 1.02–1.11, p = 0.002) and having a scheduled or planned appointment with the FP (OR = 1.03, 95% CI = 1.00–1.07, p = 0.049).
Multivariable factors associated with previous and planned visits to the FP Results of the multivariable analyses are shown in Table . Three FP factors remained negatively or positively associated with having visited the FP for palliative care in the past 6 months: travel time to the FP (OR = 0.67, 95% CI = 0.48–0.93, p = 0.02), indicating that the FP offers 24‐h telephone support (OR = 1.96, 95% CI = 1.02–3.76, p = 0.04), and better perceived care by the FP (OR = 1.05, 95% CI = 1.01–1.09, p = 0.01). The following variables were negatively or positively associated with having a scheduled appointment with the FP: female sex (OR = 0.51, 95% CI = 0.30–0.87, p = 0.01), English as first language (OR = 2.90, 95% CI = 1.04–8.11, p = 0.04), travel time to the FP (OR = 0.66, 95% CI = 0.47–0.93, p = 0.02), and ease of reaching the FP after hours (OR = 1.33, 95% CI = 1.08–1.64, p = 0.008).
DISCUSSION In our study, approximately one‐third of patients had seen their FP for palliative care in the last 6 months, and half had a scheduled or planned FP's appointment. Enabling factors associated with having visited FPs for palliative care or with having a scheduled or planned appointment included shorter travel time to FPs, 24‐h telephone support services, and ease of reaching FPs after hours. In addition, a positive perception of the FP's care was associated with having seen the FP for palliative care in the last 6 months. Female patient sex was negatively associated, and English as a first language was positively associated, with having scheduled or planning to schedule a FP visit. The number of previous visits to the palliative care clinic was not associated with either outcome. The majority of participants indicated that their FPs knew them well and expressed general satisfaction with the FP's care. However, there was less endorsement of adequate time to address problems and of confidence in receiving the best care possible. Moreover, few patients indicated that FPs provided palliative care services such as cancer‐related symptom management, emotional care, discussions of advance care directives, support for caregivers and family, and provision or arrangement of home care. , , , , FPs have reported barriers to providing palliative care, including insufficient time , ; lack of resources including education and training in palliative care , ; poor integration and communication with other healthcare providers; and ambiguity about their role in end‐of‐life care. , , , , , Given that FPs' involvement in palliative care may enable holistic care, decrease emergency room visits, and increase the likelihood of dying at home, , additional supports and resources are needed to enable this involvement. These could include training and education, better integration and communication with hospital‐based services, and compensation for care provision. , , Enabling factors, including shorter travel time to the FP's office, availability of 24‐h telephone support, and after‐hours services, were associated with previous visits to the FP for palliative care and/or with planned FP visits. Ease of access to FPs is particularly important for patients with advanced cancer, for whom fatigue is a prominent symptom , , and whose condition may worsen unexpectedly. FPs' provision of after‐hours care was similarly deemed important in a survey of patients attending a radiation oncology clinic. Group practices, out‐of‐hours cooperatives, and better remuneration for after‐hours care may facilitate FPs' provision of 24‐h care. , Additional factors were associated with both outcomes. Female patients were less likely to have a planned appointment with FPs, which suggests that they were more likely to seek care exclusively from the palliative care team. Indeed, previous research demonstrated that female patients are more likely to know about and receive palliative care. , English as a first language was also positively associated with having a scheduled or planned FP appointment. Canadian patients have reported lower rates of same‐day response from their family physicians if their first language is neither English nor French (the official Canadian languages). Extending telephone translation services that are available in many hospitals (including Princess Margaret Cancer Centre) to primary care settings would promote equitable access to primary care. Lastly, having seen FPs for palliative care services in the last 6 months was associated with a positive perception of FPs' care, with sufficient time to address patients' problems and clarity of the FP's role being deemed particularly important. Although providing sufficient time for consultations may be challenging in a busy family practice, patients with advanced cancer value patient‐led, unhurried palliative care visits. , Role clarity of the FP in this regard may be increased by patient education to explain the relevance and value of their FP as an active provider of palliative care. The observed importance of enabling and predisposing factors is relevant for models of integration of primary care into models of comprehensive, coordinated palliative care. , , , Patients with advanced disease desire continuous care that is accessible and close to home, and FPs are well placed to meet such needs. Integrated care pathways and ways to redress challenges FPs face in providing palliative care have been advanced but must be enacted concurrently with patient education about the role of FPs in palliative care. A limitation of this study is that all participants were attending an outpatient palliative care clinic and that only 10% of participants were attending the clinic for the first time. Although the number of previous visits to the palliative care clinic was not associated with either outcome in the regression analyses, comparative studies with patients who have not been referred to a specialty palliative care clinic care would be informative. However, that most patients were acquainted with a specialty palliative care service was also a strength, as their understanding of palliative care may have helped them to determine whether they would like to receive these services from their FP. Although patients had advanced disease and were receiving palliative care, they reported relatively low symptom burden; findings may not generalize to patients with worse symptom severity. Generalizability may also be limited due to the inclusion criterion of English fluency. In conclusion, the present study identified enabling factors reflecting ease of access to FPs, predisposing factors of sex and English as the first language, and positive perceptions of FPs' care as correlates of having received palliative care services from FPs or having a planned or scheduled FP appointment. Further research should investigate interventions to promote FP‐provided palliative care.
Christine C. Moon: Conceptualization (equal); data curation (lead); investigation (equal); methodology (equal); writing – original draft (equal); writing – review and editing (equal). Kenneth Mah: Formal analysis (equal); writing – original draft (equal); writing – review and editing (equal). Ashley Pope: Data curation (supporting); project administration (lead); writing – review and editing (equal). Nadia Swami: Investigation (equal); project administration (supporting); writing – review and editing (equal). Breffni Hannon: Investigation (equal); writing – review and editing (equal). Jenny Lau: Investigation (equal); writing – review and editing (equal). Ernie Mak: Investigation (equal); writing – review and editing (equal). Ahmed Al‐Awamer: Investigation (equal); writing – review and editing (equal). Subrata Banerjee: Investigation (equal); writing – review and editing (equal). Laura A. Dawson: Investigation (equal); methodology (equal); writing – review and editing (equal). Amna Husain: Investigation (equal); methodology (equal); writing – review and editing (equal). Gary Rodin: Investigation (equal); methodology (equal); writing – review and editing (equal). Lisa W. Le: Conceptualization (equal); formal analysis (equal); methodology (equal); writing – review and editing (equal). Camilla Zimmermann: Conceptualization (equal); investigation (equal); methodology (equal); writing ‐ original draft (equal); writing ‐ review and editing (equal).
This study was funded by the Canadian Institutes of Health Research (Grant No. 152996 to Dr. Zimmermann). The funding body had no role in the design of the study, collection, analysis, and interpretation of data or in writing the manuscript.
We declare that we have no competing conflicts of interest.
This study was conducted in line with the principles of the Declaration of Helsinki. Approval was granted by the University Health Network Research Ethics Board (REB #16‐5061‐CE).
|
Mobile application to support oncology patients during treatment on patient outcomes: Evidence from a randomized controlled trial | 0fb5a939-9565-4c3d-b457-b2d407bf29d0 | 10028030 | Internal Medicine[mh] | INTRODUCTION Cancer treatments, especially chemotherapy, often require substantial demands on patients and their informal caregivers, including managing an overwhelming amount of information to coordinate treatment and related activities. Comprehensive coordinating efforts are needed to manage chemotherapy treatments that usually include a complex schedule and are associated with significant or even disabling side effects. Additionally, diagnostic testing and other therapeutic modalities (e.g., surgery and radiation) may require multiple visits with different providers at various locations. The increasing penetration of smartphones across socioeconomic groups in the United States can be leveraged by individuals dealing with cancer to support their efforts to manage their health, coordinate appointments, and keep track of treatment notes and questions. , , Leveraging mobile health technologies to improve patient health care access and engagement has been proposed by the Institute of Medicine and the American Society of Clinical Oncology as a method of decreasing medical errors and increasing health care quality. , , Although thousands of apps are available to support patients during treatment, rigorous evidence is needed to understand the advantages and limitations of this technology. Prior research found that although hundreds of apps are available for cancer patients to download, few provide support for self‐management activities, and even fewer have been rigorously evaluated. , , , We designed this randomized controlled trial to evaluate the usability of a patient‐facing application (app) and test its impact on patients' quality of life and health care utilization. The app, LivingWith®, was developed by Pfizer to support individuals undergoing cancer treatment. Changes in these endpoints were evaluated from baseline to 3 months after the intervention relative to the control group.
METHODS The study protocol was approved by the University of Tennessee Health Science Center Institutional Review Board and was registered with ClinicalTrials.gov (NCT04331678). 2.1 Participants This randomized controlled trial was conducted at West Cancer Center and Research Institute (WCCRI) between June 2020 and December 2021. WCCRI is a large comprehensive cancer center with nine clinics and over 70 physicians serving 60% of all patients in the tri‐state area of west Tennessee, north Mississippi, and east Arkansas. The study site is an industry leader in implementing innovative technology to support patient engagement activities aimed at improving patient outcomes. To be eligible, WCCRI patients had to be ≥18 years old, with a confirmed cancer diagnosis within 1 month of chemotherapy initiation, had a valid email address, a smartphone device with a data plan, and were willing to download and use the app. Patients unable to communicate in English were excluded, as only English versions of the surveys and app were available at the time. 2.2 Recruitment and randomization Potentially eligible patients with a diagnosis of active solid tumor or hematologic malignancy were identified for recruitment by a research staff member using the WCCRI's electronic health record system. After confirming eligibility and completing informed consent, participants completed the enrollment survey and were randomly assigned in a 1:1 ratio to one of the two study groups: intervention and control. 2.3 Intervention Patients randomized to the intervention arm were asked to download the LivingWith® app, which is available free of cost in the Apple App store and the Google Play store. Participants were asked to enter a referral code to verify their use of the app and use it weekly for 3 months. The LivingWith® app was designed to help individuals manage life with cancer. Key app functions include: “Dashboard” to manage and view appointments and important dates; “Well‐being” to track health data from connected wearables and other health apps, and track symptoms like pain, fatigue, and sleep; “Me” to store personal health‐related documents, such as insurance cards, prescriptions, and other documents; “Health notes” to take notes and recordings from visits and pictures to support treatment; “My circle” to maintain a network of friends and families to share health updates and request emotional and logistical support; and “Resources” which provides positive affirmations, educational resources, and local event information. Select screenshots of the LivingWith® app are provided in Figure . 2.4 Study measures All consented participants completed a survey at enrollment and at 3 months. The key study endpoints were self‐reported quality of life, health care utilization, and usability of the app (intervention group only). 2.4.1 Quality of life Two validated instruments were used to assess quality of life: the Functional Assessment of Cancer Therapy‐General (FACT‐G) and the Short‐Form Health Survey (SF‐12). Composite scores were calculated based on the user manual, which included: a composite score (27 items, range 0–108) from the FACT‐G instrument; and physical and mental component scores (each with a mean of 50 and standard deviation of 10 in the general United States population) from the SF‐12 instrument. In general, a higher score is more favorable. 2.4.2 Health care utilization Survey questions about health care utilization were adapted from validated items available through the National Health Interview Survey. They included questions about the number of times an individual received the following services in the preceding 3 months: (1) office visit, (2) visit with a mental health professional, (3) emergency room, urgent care, same‐day appointment, or walk‐in clinic visit, and (4) virtual care (phone calls, emails, or web portal messages). 2.4.3 App use We measured the numbers of logins among intervention participants using Pfizer analytics tools. In the follow‐up survey, intervention group participants were asked to rate the usefulness of the app overall, and of individual function on a 1 to 5 scale (1 = Never used, 2 = Not at all useful, to 5 = Very Useful). Structured questions about specific reasons they used and did not use the app were asked using a 1 to 4 scale (1 = Strongly Disagree, to 4 = Strongly Agree). The survey also allowed for free‐text feedback from participants regarding the reasons they used the app and concerns with using the app. 2.4.4 Patient characteristics We collected patients' age at enrollment, gender, race, education, income, marital status, smoking status, cancer types, and cancer stage at diagnosis. 2.5 Statistical analyses Baseline characteristics were summarized by study arms using frequency and percentage for categorical measures and mean and standard deviation for continuous measures. Among the intervention arm, we also compared demographic and clinical characteristics stratified by whether the individual reported having used the app. Changes from baseline to the 3 month follow‐up for each outcome were constructed and compared between study arms. Student t ‐test and Chi‐square test were used to compare continuous measures and categorical measures respectively by subgroups. Multivariable linear regression was used to examine changes in study outcomes by study arm adjusting for their baseline value. Thematic analyses were used to evaluate free‐text survey questions on app usability. We used SAS Software (SAS Institute Inc.) for our statistical analyses.
Participants This randomized controlled trial was conducted at West Cancer Center and Research Institute (WCCRI) between June 2020 and December 2021. WCCRI is a large comprehensive cancer center with nine clinics and over 70 physicians serving 60% of all patients in the tri‐state area of west Tennessee, north Mississippi, and east Arkansas. The study site is an industry leader in implementing innovative technology to support patient engagement activities aimed at improving patient outcomes. To be eligible, WCCRI patients had to be ≥18 years old, with a confirmed cancer diagnosis within 1 month of chemotherapy initiation, had a valid email address, a smartphone device with a data plan, and were willing to download and use the app. Patients unable to communicate in English were excluded, as only English versions of the surveys and app were available at the time.
Recruitment and randomization Potentially eligible patients with a diagnosis of active solid tumor or hematologic malignancy were identified for recruitment by a research staff member using the WCCRI's electronic health record system. After confirming eligibility and completing informed consent, participants completed the enrollment survey and were randomly assigned in a 1:1 ratio to one of the two study groups: intervention and control.
Intervention Patients randomized to the intervention arm were asked to download the LivingWith® app, which is available free of cost in the Apple App store and the Google Play store. Participants were asked to enter a referral code to verify their use of the app and use it weekly for 3 months. The LivingWith® app was designed to help individuals manage life with cancer. Key app functions include: “Dashboard” to manage and view appointments and important dates; “Well‐being” to track health data from connected wearables and other health apps, and track symptoms like pain, fatigue, and sleep; “Me” to store personal health‐related documents, such as insurance cards, prescriptions, and other documents; “Health notes” to take notes and recordings from visits and pictures to support treatment; “My circle” to maintain a network of friends and families to share health updates and request emotional and logistical support; and “Resources” which provides positive affirmations, educational resources, and local event information. Select screenshots of the LivingWith® app are provided in Figure .
Study measures All consented participants completed a survey at enrollment and at 3 months. The key study endpoints were self‐reported quality of life, health care utilization, and usability of the app (intervention group only). 2.4.1 Quality of life Two validated instruments were used to assess quality of life: the Functional Assessment of Cancer Therapy‐General (FACT‐G) and the Short‐Form Health Survey (SF‐12). Composite scores were calculated based on the user manual, which included: a composite score (27 items, range 0–108) from the FACT‐G instrument; and physical and mental component scores (each with a mean of 50 and standard deviation of 10 in the general United States population) from the SF‐12 instrument. In general, a higher score is more favorable. 2.4.2 Health care utilization Survey questions about health care utilization were adapted from validated items available through the National Health Interview Survey. They included questions about the number of times an individual received the following services in the preceding 3 months: (1) office visit, (2) visit with a mental health professional, (3) emergency room, urgent care, same‐day appointment, or walk‐in clinic visit, and (4) virtual care (phone calls, emails, or web portal messages). 2.4.3 App use We measured the numbers of logins among intervention participants using Pfizer analytics tools. In the follow‐up survey, intervention group participants were asked to rate the usefulness of the app overall, and of individual function on a 1 to 5 scale (1 = Never used, 2 = Not at all useful, to 5 = Very Useful). Structured questions about specific reasons they used and did not use the app were asked using a 1 to 4 scale (1 = Strongly Disagree, to 4 = Strongly Agree). The survey also allowed for free‐text feedback from participants regarding the reasons they used the app and concerns with using the app. 2.4.4 Patient characteristics We collected patients' age at enrollment, gender, race, education, income, marital status, smoking status, cancer types, and cancer stage at diagnosis.
Quality of life Two validated instruments were used to assess quality of life: the Functional Assessment of Cancer Therapy‐General (FACT‐G) and the Short‐Form Health Survey (SF‐12). Composite scores were calculated based on the user manual, which included: a composite score (27 items, range 0–108) from the FACT‐G instrument; and physical and mental component scores (each with a mean of 50 and standard deviation of 10 in the general United States population) from the SF‐12 instrument. In general, a higher score is more favorable.
Health care utilization Survey questions about health care utilization were adapted from validated items available through the National Health Interview Survey. They included questions about the number of times an individual received the following services in the preceding 3 months: (1) office visit, (2) visit with a mental health professional, (3) emergency room, urgent care, same‐day appointment, or walk‐in clinic visit, and (4) virtual care (phone calls, emails, or web portal messages).
App use We measured the numbers of logins among intervention participants using Pfizer analytics tools. In the follow‐up survey, intervention group participants were asked to rate the usefulness of the app overall, and of individual function on a 1 to 5 scale (1 = Never used, 2 = Not at all useful, to 5 = Very Useful). Structured questions about specific reasons they used and did not use the app were asked using a 1 to 4 scale (1 = Strongly Disagree, to 4 = Strongly Agree). The survey also allowed for free‐text feedback from participants regarding the reasons they used the app and concerns with using the app.
Patient characteristics We collected patients' age at enrollment, gender, race, education, income, marital status, smoking status, cancer types, and cancer stage at diagnosis.
Statistical analyses Baseline characteristics were summarized by study arms using frequency and percentage for categorical measures and mean and standard deviation for continuous measures. Among the intervention arm, we also compared demographic and clinical characteristics stratified by whether the individual reported having used the app. Changes from baseline to the 3 month follow‐up for each outcome were constructed and compared between study arms. Student t ‐test and Chi‐square test were used to compare continuous measures and categorical measures respectively by subgroups. Multivariable linear regression was used to examine changes in study outcomes by study arm adjusting for their baseline value. Thematic analyses were used to evaluate free‐text survey questions on app usability. We used SAS Software (SAS Institute Inc.) for our statistical analyses.
RESULTS Overall, 240 patients initiating chemotherapy treatment at three WCCRI locations consented to participate in this trial from June 2020 to June 2021. A total of 224 participants completed the baseline survey and were randomized into the intervention ( n = 113) and control arms ( n = 111). The attrition rate was 28.3% for the intervention group and 20.7% for the control group. The final analysis included 169 patients who completed both the baseline and 3‐month follow‐up surveys (Figure ). The mean age was 56.3 years (SD = 12.4), 68.6% were female, 60.9% were White, and 35.5% were Black. The most common cancer among our cohort was breast cancer ( n = 65, 38.5%), followed by hematologic cancer ( n = 24, 14.2%), and gastrointestinal cancer ( n = 21, 12.4%). Most participants had more than high school education ( n = 135, 79.9%) and about half of the participants had a household income of $60,000 or more ( n = 85, 50.3%). See Table for details. 3.1 App use Among intervention group participants, most (75.3%) reported using the app (Figure ). Demographic and clinical characteristics were statistically similar between intervention arm participants who reported having used or never using the app (eTable ). Notably, 80.0% of Black participants versus 70.8% of White participants, and 81.8% of those with household income less than $60,000 versus 72.7% for those with income over $60,000 reported using the app. The mean age among app users was 55.0 versus 59.0 for nonusers. The mean number of logins was 11.7 times (SD = 15.2) during the 3‐month intervention (range from 0 to 79). Among participants who had any app use, the mean number of logins was 15.6 times (SD = 15.7). Patients' reported usefulness varied by distinct functions of the app. Overall, among those who used the app, 45.7% reported the app useful. The most used app function was tracking emotion, pain, and sleep (65.8%), and among those who used this function, more than half found it useful (51.9%). The least used app function was creating a personal profile for their documents, health insurance information and prescription (63.3% reported that they never used it), and only 27.8% of those who used this function found it useful. Other app functions, such as dashboard and upcoming appointment reminders, were more commonly used and rated as useful by 46.3% and 43.8% of users (Figure ). Among specific reasons for using the app, over 40% of the participants agreed that the app helped them better understand their treatment and overall health, and to organize health care including appointments, medications, and information related to the treatment. On the other hand, over 75% of the participants reported that a barrier to using the app was a personal preference for communicating with friends and family directly in person or over the phone instead of with an app. About a third of the participants cited the complexity of the app (36.4%) and privacy concerns (30.8%) as barriers to using the app (Figure ). Among a total of 42 free‐text responses for questions about the reasons and impacts of using the app, about a quarter mentioned that using it to track appointments and notes (26.8%) and track their daily symptoms and moods (24.4%) was beneficial. For example, one participant noted that: “The app helped me to keep track of my symptoms and allowed me to share with my doctor on my visits.” Another participant responded that the app: “…helped me to see how many bad days against good days.” Respondents (11.9%) specifically pointed out that the app provided them with motivation. One participant noted that: “The diary and the appointment notes are my favorite features in the app and the daily affirmations.” Some (9.5%) reported that the app was convenient to use. One participant wrote: “My husband could get my phone and pull up all my meds when I wasn't feeling like taking them during my chemo.” Several participants (39.9%) reported not using the app due to their busy schedules, use of other similar apps, not being an app user in general, and some concerns with the usability of the app. For example, a participant noted “It would be helpful for me and would help my friend and family understand and stay updated. I was unable to utilize the app in that manner because some people were concerned about downloading the app.” Another participant wrote: “I work full time from home. The additional effort required for inputting all that data into the app was not something that I wanted to do. Just didn't have the energy.” 3.2 Health care utilization At baseline, participants reported a similar level of utilization across two study arms. On average, they had about one visit to the emergency room, urgent care, same‐day appointment, or walk‐in clinic (0.8 vs. 1.1, p = 0.22), and less than one visit with a mental health professional (0.2 vs. 0.3, p = 0.61) in intervention and control groups in the prior 3 months, respectively. Participants had over four scheduled office visits (4.5 vs. 4.3, p = 0.72) and over three virtual care encounters in the prior 3 months (3.7 vs. 3.8, p = 0.95, Figure , eTable ). At 3‐month follow‐up survey, overall utilization was lower among intervention group participants than at baseline. Patients assigned to the intervention group reported having fewer office visits (3.5 vs. 4.3, p = 0.046) and visits with a mental health professional (0.1 vs. 0.3, p = 0.064) at 3‐month follow‐up compared with those in the control group. Adjusted results show 0.74 fewer office visits ( p = 0.043) and 0.24 fewer visits with a mental health professional ( p = 0.061) among intervention participants relative to controls (Table ). Utilization for other services was similar across the two arms. For example, the average number of emergency visits, urgent care, same‐day appointments, or walk‐in clinic visit was 0.3 vs. 0.6 ( p = 0.11), and the average number of phone calls, emails, or web portal messages with providers was 2.8 vs. 2.4 ( p = 0.42) for the intervention and control groups, respectively (Figure , eTable ). 3.3 Quality of life At baseline, participants in both arms reported similar quality of life across all scales (eTable ). As expected for patients undergoing chemotherapy treatment, participants experienced worsening physical health in both arms. The mental health composite scores slightly increased (i.e., improved) from baseline. There were no statistically significant differences in these changes by study arm (Table ).
App use Among intervention group participants, most (75.3%) reported using the app (Figure ). Demographic and clinical characteristics were statistically similar between intervention arm participants who reported having used or never using the app (eTable ). Notably, 80.0% of Black participants versus 70.8% of White participants, and 81.8% of those with household income less than $60,000 versus 72.7% for those with income over $60,000 reported using the app. The mean age among app users was 55.0 versus 59.0 for nonusers. The mean number of logins was 11.7 times (SD = 15.2) during the 3‐month intervention (range from 0 to 79). Among participants who had any app use, the mean number of logins was 15.6 times (SD = 15.7). Patients' reported usefulness varied by distinct functions of the app. Overall, among those who used the app, 45.7% reported the app useful. The most used app function was tracking emotion, pain, and sleep (65.8%), and among those who used this function, more than half found it useful (51.9%). The least used app function was creating a personal profile for their documents, health insurance information and prescription (63.3% reported that they never used it), and only 27.8% of those who used this function found it useful. Other app functions, such as dashboard and upcoming appointment reminders, were more commonly used and rated as useful by 46.3% and 43.8% of users (Figure ). Among specific reasons for using the app, over 40% of the participants agreed that the app helped them better understand their treatment and overall health, and to organize health care including appointments, medications, and information related to the treatment. On the other hand, over 75% of the participants reported that a barrier to using the app was a personal preference for communicating with friends and family directly in person or over the phone instead of with an app. About a third of the participants cited the complexity of the app (36.4%) and privacy concerns (30.8%) as barriers to using the app (Figure ). Among a total of 42 free‐text responses for questions about the reasons and impacts of using the app, about a quarter mentioned that using it to track appointments and notes (26.8%) and track their daily symptoms and moods (24.4%) was beneficial. For example, one participant noted that: “The app helped me to keep track of my symptoms and allowed me to share with my doctor on my visits.” Another participant responded that the app: “…helped me to see how many bad days against good days.” Respondents (11.9%) specifically pointed out that the app provided them with motivation. One participant noted that: “The diary and the appointment notes are my favorite features in the app and the daily affirmations.” Some (9.5%) reported that the app was convenient to use. One participant wrote: “My husband could get my phone and pull up all my meds when I wasn't feeling like taking them during my chemo.” Several participants (39.9%) reported not using the app due to their busy schedules, use of other similar apps, not being an app user in general, and some concerns with the usability of the app. For example, a participant noted “It would be helpful for me and would help my friend and family understand and stay updated. I was unable to utilize the app in that manner because some people were concerned about downloading the app.” Another participant wrote: “I work full time from home. The additional effort required for inputting all that data into the app was not something that I wanted to do. Just didn't have the energy.”
Health care utilization At baseline, participants reported a similar level of utilization across two study arms. On average, they had about one visit to the emergency room, urgent care, same‐day appointment, or walk‐in clinic (0.8 vs. 1.1, p = 0.22), and less than one visit with a mental health professional (0.2 vs. 0.3, p = 0.61) in intervention and control groups in the prior 3 months, respectively. Participants had over four scheduled office visits (4.5 vs. 4.3, p = 0.72) and over three virtual care encounters in the prior 3 months (3.7 vs. 3.8, p = 0.95, Figure , eTable ). At 3‐month follow‐up survey, overall utilization was lower among intervention group participants than at baseline. Patients assigned to the intervention group reported having fewer office visits (3.5 vs. 4.3, p = 0.046) and visits with a mental health professional (0.1 vs. 0.3, p = 0.064) at 3‐month follow‐up compared with those in the control group. Adjusted results show 0.74 fewer office visits ( p = 0.043) and 0.24 fewer visits with a mental health professional ( p = 0.061) among intervention participants relative to controls (Table ). Utilization for other services was similar across the two arms. For example, the average number of emergency visits, urgent care, same‐day appointments, or walk‐in clinic visit was 0.3 vs. 0.6 ( p = 0.11), and the average number of phone calls, emails, or web portal messages with providers was 2.8 vs. 2.4 ( p = 0.42) for the intervention and control groups, respectively (Figure , eTable ).
Quality of life At baseline, participants in both arms reported similar quality of life across all scales (eTable ). As expected for patients undergoing chemotherapy treatment, participants experienced worsening physical health in both arms. The mental health composite scores slightly increased (i.e., improved) from baseline. There were no statistically significant differences in these changes by study arm (Table ).
DISCUSSIONS This study used a randomized controlled trial to evaluate the efficacy and usability of an app (LivingWith®) that provides self‐management support for patients with cancer undergoing chemotherapy on health care utilization and quality of life. The findings demonstrate a positive impact of the intervention in reducing the number of office visits utilized and an encouraging trend in fewer visits with a mental health provider. There were no significant changes by study groups in quality of life, use of virtual care, or emergency room and urgent care visits. Patients with cancer undergoing chemotherapy must manage an overwhelming amount of information to coordinate their treatment and related activities. Comprehensive coordinating efforts are needed to manage complex schedules and side‐effects of chemotherapy treatments. Our findings showed similar reported app use across sociodemographic subgroups, with slightly higher use among participants who identified as Black or had lower incomes (eTable ). This suggests that the increasing availability of smartphones across socioeconomic groups and regions provides an opportunity to develop digital solutions to reduce health disparities through increasing access to educational resources and symptom monitoring and management. Apps can be leveraged by patients to support their efforts to manage their health, coordinate appointments, and keep track of treatment notes and questions. Although hundreds of health apps are available for patients with cancer, few provide support for self‐management activities, and even fewer have been rigorously evaluated. , , , , Emerging evidence suggests that health apps can be used by various population groups and may support improvements in pain and fatigue. In our study, the two study groups did not report differences in quality of life, physical health, or mental health. This may be due to the relatively short follow‐up period while patients are still in the midst of completing chemotherapy, which can impair daily function and worsen quality of life. Future studies should examine the impact of the intervention on patients in different phases of their treatment starting with the initial diagnosis, and on longer‐term quality of life after they complete their active cancer treatments. Most intervention participants reported using the app and used it weekly during the 3‐month intervention. Notably, nearly a quarter of intervention participants who consented to participating in this app‐based study reported not even trying the app. Common barriers to using the app included personal preference to communicate with friends and family directly, usability challenges, and privacy concerns. Similarly, previous studies have found that usability was a challenge, especially for a diverse group of patients. , Satisfaction with the app was mixed with fewer than half of those using it reporting it as useful. The functions most reported as useful included tracking health, reviewing the dashboard, and upcoming appointments. The functions least commonly reported as useful included setting up a personal profile, connecting with friends and family, and connecting with other health apps. The app was designed by Pfizer to provide comprehensive support for people impacted by cancer. Providing multiple functions to support an array of self‐management activities and education may have added to the complexity of using the app. Still, developers are continuously reviewing and updating the app to improve usability. Even with improved usability, it is possible that some individuals will be more open to using technology and apps to support their health than others. The variability in personal preference highlights the importance of not relying exclusively on app‐based interventions to support self‐management activities and health behavior changes. This is the first prospectively randomized trial of an app designed to support individuals dealing with cancer with tools to support self‐management activities that shows fewer office visits and no adverse impact on quality of life or urgent care among intervention participants relative to controls. Notably, adjusted results showed a larger reduction in health care utilization for all in‐person services (office visits, mental health, emergency or urgent care) among the intervention group relative to controls, however, the difference only reached statistical significance for office visits (Table ). The app provided several functions to support patients and their informal caregivers to coordinate activities and track health and information during treatment. Although more research is needed to understand the mechanism for app participants to require fewer visits, it is possible that using the app to manage and track information resulted in improved self‐management and less need for visits. In fact, among app users, many reported that the app helped them to better understand their treatment plan. Despite having fewer office visits, intervention group participants reported similar levels of quality of life and urgent and emergency care. This is encouraging and implies that fewer visits were not associated with adverse outcomes for patients. Despite its strengths, including the rigorous randomized study design, the novelty of the research, and the diverse sample of patients, this study has limitations. First, participants were recruited from a single cancer center located in the mid‐South region of the United States and the study was limited to English speakers. Findings may not be generalizable to other regions of the country, other cancer centers, and non‐English speakers. Second, intervention group participants were asked to download the study app and use it weekly when consenting to participate in the study, but app use was not enforced, and nearly 1 in 4 reported not using the app. Moreover, it's possible that participants in both groups used other apps to support their health during the study. Lastly, the study only followed participants for a short time, 3 months, and relied on self‐reported data for study outcomes.
CONCLUSION In conclusion, in a randomized controlled trial, patients undergoing chemotherapy randomized to use an app to support self‐management activities reported having fewer office visits during the 3‐month intervention compared to controls, with no adverse effects or changes in other study outcomes. Although there are hundreds of health apps available to support patients undergoing cancer treatment, few have been rigorously evaluated. This study is the first to show that an app designed to support self‐management functions helped patients better understand and coordinate their treatment plan resulting in fewer office visits. Still, nearly a quarter of patients randomized to the intervention arm reported never using the app and among app users, fewer than half found it useful. Barriers to using the app included personal preference, privacy concerns, and usability challenges. To address these challenges, future studies should focus efforts to improve user‐centered design and test strategies to increase awareness and willingness to use apps for heath management. Nonetheless, it is important to keep in mind that not all patients will want to use apps to support their care. More research is needed to understand the mechanism for app use to reduce the need for office visits and its impact on longer‐term health outcomes.
Ilana Graetz: Conceptualization (lead); formal analysis (equal); funding acquisition (equal); investigation (equal); methodology (equal); writing – original draft (lead); writing – review and editing (lead). Xin Hu: Formal analysis (lead); methodology (equal); writing – review and editing (equal). Andrea N. Curry: Data curation (lead); project administration (lead); writing – review and editing (equal). Andrew Robles: Data curation (equal); project administration (equal); writing – review and editing (equal). Gregory A. Vidal: Project administration (equal); supervision (equal); writing – review and editing (equal). Lee S. Schwartzberg: Conceptualization (equal); funding acquisition (equal); investigation (equal); methodology (equal); supervision (equal); writing – review and editing (equal).
Dr. Vidal reported receiving personal fees from Roche/Genentech, Novartis, Eli Lilly, Immunometric, Puma, Pfizer, AstraZeneca, Biotheranautics, Daiichi Sankyo, Vector Oncology and research funding from Roche/Genentech, Puma, Celcuity, Merck, BMS, Eli Lilly, GTx Inc., AstraZeneca, Pfizer, Immunomedics, Tesaro, Halozyme, and ownership of Oncodisc. Dr. Schwartzberg reported receiving personal fees from Amgen, Pfizer, Helsinn, Genentech, Genomic Health, BMS, Myriad, AstraZeneca, Bayer, Spectrum, Napo and research support from Amgen, Pfizer. Dr. Graetz received research support from Pfizer. Ms. Hu received a dissertation grant from PhRMA Foundation. No other disclosures were reported.
Appendix S1 Click here for additional data file.
|
Application of NGS molecular classification in the diagnosis of endometrial carcinoma: A supplement to traditional pathological diagnosis | c5f69f88-7306-4e4a-a7df-efa0a8838834 | 10028062 | Anatomy[mh] | INTRODUCTION Endometrial cancer (EC) is the most common gynecologic malignancy in China, with an increasing number of new diagnoses and a rapid mortality rate, which takes a heavy toll on women's health. , , The risk factors for EC include an excess of endogenous and exogenous estrogens, obesity, and others, and these factors can vary substantially between and within countries. , The gold standard treatment of EC is surgery, but for the patients with advanced disease, postoperative adjuvant treatment (chemotherapy, radiotherapy or targeted therapy) can be added to provide the best long‐term survival. , , The management of patients with EC largely relies on tumor histopathology. Traditional classification based on pathologic assessment includes Bokhman classification and WHO 2014 classification , , can be served as an important input for (adjuvant) treatment decisions, but has proved to be unreliable and has poor reproducibility. In addition, with considerable interobserver variation, high‐grade EC cannot be reliably classified according to histomorphologic criteria, resulting in misdiagnosis due to mixed high‐grade components in histology. , , Moreover, stage and lymphovascular space invasion (LVSI), parameters that are needed to define the risk group, are only available after definitive surgery, , such information comes too late for EC patients who may benefit from fertility‐sparing alternatives. This situation prompted research into developing biologically informative diagnostic tools to enhance the diagnosis and risk stratification of patients with EC. During the last decade, molecular characterization of EC is advancing rapidly. The Cancer Genome Atlas (TCGA) project for EC has shown how combinations of molecular features can be used to guide treatment decisions. According to different molecular profiles, TCGA divided EC into four distinct molecular subgroups: Polymerase‐epsilon ( POLE ) ultramutated, microsatellite instability hypermutated, a copy‐number low group with a low mutational burden, and a copy‐number high serous‐like group. Patients with POLE ‐mutated ECs showed an excellent prognosis while the copy‐number high group have the worst. Unfortunately, the high cost of whole‐genome sequencing has greatly limited its practical application. For improvement, Talhouk A proposed a pragmatic molecular categorization model (ProMisE) based on POLE mutational analysis, mismatch repair protein immunohistochemistry (IHC), and p53 IHC. Though IHC is easier to carry out clinically, the interpretation of IHC is greatly affected by different expert pathologists, turning to unreliable conclusions. The emergence of next generation sequencing (NGS) has enabled the high comprehensive analysis of genomes and transcriptomes. Compared with Sanger sequencing, NGS has revolutionized the ability to sequence nucleic acids, and made genome sequencing more efficient and more cost‐effective. Jutta et al assessed molecular subtype using the ProMisE classifier and an NGS‐based approach, and compared the concordance between the two methods, the result showed excellent agreement. Together, these significant findings all point to indicating the potential of NGS use in clinical molecular subgrouping of EC, leaving it wide open for further exploration. Assessment of preoperative tumor histologic subtype and grade using specimens taken from hysteroscopic biopsy, or dilatation and curettage is essential in determining the surgical staging. Studies revealed that there are limitations in preoperative sampling diagnosis because of inadequate tumor sampling or the observer variability, and the agreement rate for tumor grade between diagnostic sampling and post‐operative diagnosis varies from 32% to 97%. , Discordances may result in excessive or insufficient treatment, such as unnecessary surgical procedures like lymph node dissection. The emergence of molecular classification provides new insights into the shortcomings of traditional pathological diagnosis of diagnostic samples. The research illustrated that the concordance rate for pre‐ versus post‐surgery samples of TCGA classifier and ProMisE classifier are 92% and 89%, respectively, significantly higher than that of pathological diagnosis. At least to our knowledge, there is no documentation that NGS‐based testing can be performed on diagnostic endometrial specimens obtained prior to surgical staging and its concordance with subsequent hysterectomy specimens. A simplified NGS panel is convenient for operation, easy to generalize, and with more accuracy, which might overcome the shortcomings of TCGA classifier and ProMisE classifier. On this basis, we applied a simplified NGS panel to a large population‐based cohort of Chinese EC patients to validate this tool, and additionally, compared discriminatory concordance between curettage and subsequent hysterectomy specimens.
METHODS 2.1 Patient cohort, sample collection, and pathological information collection The study consists of a retrospective cohort of patients with EC from Sun Yat‐Sen Memorial Hospital, Sun Yat‐Sun University treated for EC between 2018 and 2021. Paraffin‐embedded (FFPE) tissue samples were selected from the biobank at the Department of Pathology. Of 120 patients, 40 patients were excluded owing to insufficient DNA quality. 35 of 80 patients with both curettage and hysterectomy endometrial specimens were picked out and used for further concordance analysis. The study was granted ethical approval SYSEC‐KY‐KS‐2021‐195. 2.2 MMR and P53 immunohistochemistry MMR/P53 IHC staining was performed on full sections. , MMR were classified into two categories according to the MMR protein expression status: dMMR (protein expression of MLH1/PMS2/MSH2/MSH6 is negative) and MMR intact (all MMR proteins positively expressed). P53 was interpreted as abnormal if there was complete negative staining or strong/diffuse staining in 70% of tumor cells (aberrant expression pattern). , 2.3 DNA extraction, library construction, and targeted sequencing Genomic DNA (Tumor cell content ≥30%) from FFPE samples was extracted, purified, and quantified using MagPure FFPE DNA LQ Kit (Magen) according to regulated processes. For NGS analysis, a commercially available targeted AmoyDx EC Panel covering POLE , TP53 , and MSI was used in this study (Amoy Diagnostics). DNA sequencing was performed on the NextSeq500 Illumina platform (Illumina). The average depth was 1000×, and the effective sequencing depth was greater than 300×. The proportion of Q30 bases was ≥75%. The sequencing data were analyzed by Sequencing Data Analysis Software (Amoy Diagnostics), the deleterious and suspected deleterious gene mutations were scrutinized and interpreted according to American College of Medical Genetics and Genomics (ACMG) guidelines and/or in ClinVar. The MSI phenotype detection method is based on the read count distribution of 55 specific microsatellite loci. A given threshold is set using the coverage ratio of a specific set of repeat lengths for each microsatellite locus, then the locus is categorized as unstable if the coverage ratio is less than the given threshold. MSS is categorized when the percentage of unstable loci is less than 15% in the given sample. 2.4 Statistical analysis Histotype, tumor grade, IHC results, and other pathological details were taken from the original pathology report system in our center. The level of concordance for a molecular profile in both specimens was determined using overall accuracy and Cohen's kappa estimates. 95% confidence intervals are computed using the bootstrap approach with 1000 bootstrap samples.
Patient cohort, sample collection, and pathological information collection The study consists of a retrospective cohort of patients with EC from Sun Yat‐Sen Memorial Hospital, Sun Yat‐Sun University treated for EC between 2018 and 2021. Paraffin‐embedded (FFPE) tissue samples were selected from the biobank at the Department of Pathology. Of 120 patients, 40 patients were excluded owing to insufficient DNA quality. 35 of 80 patients with both curettage and hysterectomy endometrial specimens were picked out and used for further concordance analysis. The study was granted ethical approval SYSEC‐KY‐KS‐2021‐195.
MMR and P53 immunohistochemistry MMR/P53 IHC staining was performed on full sections. , MMR were classified into two categories according to the MMR protein expression status: dMMR (protein expression of MLH1/PMS2/MSH2/MSH6 is negative) and MMR intact (all MMR proteins positively expressed). P53 was interpreted as abnormal if there was complete negative staining or strong/diffuse staining in 70% of tumor cells (aberrant expression pattern). ,
DNA extraction, library construction, and targeted sequencing Genomic DNA (Tumor cell content ≥30%) from FFPE samples was extracted, purified, and quantified using MagPure FFPE DNA LQ Kit (Magen) according to regulated processes. For NGS analysis, a commercially available targeted AmoyDx EC Panel covering POLE , TP53 , and MSI was used in this study (Amoy Diagnostics). DNA sequencing was performed on the NextSeq500 Illumina platform (Illumina). The average depth was 1000×, and the effective sequencing depth was greater than 300×. The proportion of Q30 bases was ≥75%. The sequencing data were analyzed by Sequencing Data Analysis Software (Amoy Diagnostics), the deleterious and suspected deleterious gene mutations were scrutinized and interpreted according to American College of Medical Genetics and Genomics (ACMG) guidelines and/or in ClinVar. The MSI phenotype detection method is based on the read count distribution of 55 specific microsatellite loci. A given threshold is set using the coverage ratio of a specific set of repeat lengths for each microsatellite locus, then the locus is categorized as unstable if the coverage ratio is less than the given threshold. MSS is categorized when the percentage of unstable loci is less than 15% in the given sample.
Statistical analysis Histotype, tumor grade, IHC results, and other pathological details were taken from the original pathology report system in our center. The level of concordance for a molecular profile in both specimens was determined using overall accuracy and Cohen's kappa estimates. 95% confidence intervals are computed using the bootstrap approach with 1000 bootstrap samples.
RESULTS 3.1 Patient cohort The study flow chart is presented in Figure . A total of 120 EC patients with post‐operative specimens from Sun Yat‐Sen Memorial Hospital between 2018 and 2021 were included. 40 patients were excluded for failed sequencing or insufficient tumor tissue for DNA extraction. The remaining 80 patients were compared for MMR/p53 IHC, grade, stages, histotype, and molecular subgroup assignment by a simplified NGS panel (Figure ). 35 of 80 cases that had paired curettage specimens were further analyzed of NGS‐based molecular classification and traditional pathological classification. 3.2 Description statistics of EC patients according to NGS molecular subgrouping Eighty patients with post‐operative specimens were qualified for analysis. The pathological characteristics and demographics of these patients were described in Table . The median patient age at diagnosis was 60 years (range from 35 to 75 years). The majority of cases (62, 77.5%) were endometrioid histotypes, 4 (5%) were serous, and the rest were clear cell and mixed histology. Grade distribution shown 17 (21.3%) grade 1, 34 (42.5%) grade 2, and 29 (36.2%) grade 3. Molecular classification using simplified NGS panel yielded 4 molecular subgroups: 8 (10%) POLE , 21 (26.2%) MSI‐H, 30 (37.5%) TP53 wt, and 21 (26.3%) TP53 abn. A small proportion (5, 6.25%) of patients demonstrated more than one molecular feature. ECs harbored POLE exonuclease domain mutations were mostly endometrioid (87.5%) and consistent with previous findings, p53abn patients were older, serous (61.9%), higher rate of myometrial invasion ≥50% (47.6%) and were correlated with high stage and grade (G3 [76.2%], stages II ~ IV [47.6%]). The distribution of the four molecular subgroups with EC patients was similar to the proportion reported in previous studies (Figure ), proving the accurate discriminatory ability of our NGS classifier. 3.3 Concordance between NGS‐based MSI status detection and MMR Immunohistochemistry on post‐operative specimens Immunohistochemistry (the protein expression of MLH1, MSH2, MSH6, and PMS2) and pentaplex PCR‐based assays are the two common methodologies for the assessment of MSI phenotype in EC classification. In this study, we analyzed the MSI status using NGS based on read‐count distribution and compared the agreement with MMR IHC. A total of 32 cases with post‐operative specimens that had MMR (MLH1/PMS2, MSH2/MSH6) IHC reports in system from our center were picked out and subsequently performed NGS‐based MSI detection. Finally, 32 cases have been compared. As shown in Table , 25 cases were MMR IHC intact and their DNA sequencing results showed microsatellite stable (MSS), while 6 cases were MSI‐H and MMR protein‐deficient (Figure ). However, one case with MSI‐H did not find expression loss for all four MMR genes. Thus, a high agreement rate (31/32, 96.9%) was observed between these two methods, inconsistency was only seen in one case. 3.4 Concordance of TP53 mutational analysis and p53 Immunohistochemistry on post‐operative specimens Scoring of p53 IHC sections between the three gynecologic pathologists was evaluated within the 65 hysterectomy samples. Complete negative staining or strongly positive (set to >70%, >80%, >85%, and > 90% of tumor cells for comparison in our study) was interpreted as abnormal. As detailed in Table , the concordance rate of p53 IHC and TP53 mutational analysis lay between 60% to 81.5%, which was highest when using >80% and >85% of tumor cells as criteria, suggesting TP53 mutational analysis could be utilized in helping p53 IHC define the right cutoff score of aberrant positive staining in clinical practice. Figure showed consistency between expression patterns of p53 IHC and TP53 mutation patterns. Information about 12 discordant cases were summarized in Table . Here we set an optimal threshold of tumor cell content ≥80%. Patients from 1 to 10 had high expression of TP53 in protein level which was considered abnormal, while no DNA mutation was found. Conversely, both patient 11 and patient 12 detected TP53 mutations. To further explore the correlation between clinicopathological features and p53 IHC, and TP53 mutation subgrouping, we analyzed the distribution of high‐risk features, including high grade, stage, LVSI, non‐endometrioid type, myometrial invasion (≥50%), in TP53 abn and p53 IHC abn subgroup. Intriguingly, a much more common high‐risk features distribution was observed in TP53 abn subgroup rather than p53 IHC abn subgroup (Table ). TP53 mutation can occur in POLE mut or MSI‐H EC, but is often considered a passenger mutation. When excluded EC with more than one molecular feature, the correlation between high‐risk features and TP53 abn subgroup turned out more significant. 3.5 Concordance of histotype and tumor grade between curettage and subsequent hysterectomy specimens The descriptive analysis, including patient demographics, tumor grade, histotype, and molecular subgroups for 35 patients with paired curettage and subsequent hysterectomy specimens were detailed in Table and Table . Table showed the concordance metrics for grade and histotype. The overall concordance rate for grade and histotype was only 74.29% (26/35) and 54.29% (19/35), confirming the lack of reproducibility of the results of the pathologic assessments between curettage and hysterectomy specimens. Downgrading was found in 8.57% (3/35) and upgrading was found in 14.29% (5/35) of the cases, one representative discordant case was illustrated in Figure . 3.6 Concordance of the molecular classification between curettage and subsequent hysterectomy specimens Concordance metrics for NGS‐based molecular classification comparing curettage and subsequent hysterectomy specimens was illustrated in Table . The overall accuracy and Cohen's kappa value were 0.88 and 0.84, respectively (Table ), demonstrating a great improvement over the conventional pathological assessment of grade and histotype. 3.7 Interrogation of patients discordant on NGS molecular classification Four patients (4/35) with discordant results between curettage and subsequent hysterectomy specimens as assessed by NGS classifier were summarized in Supplement table 6. In Cases 1, 3, and 4. The result was not changed after retesting and reassessment. For case 1, a POLE mutation (p.T278A) was detected in the curettage sample but not in the post‐staging sample. The p.T278A is a missense mutation with uncertain clinical significance. The discordance might be related to the spatial heterogeneity of the tumor. Case 2 showed discordant results in MSI status, with MSI‐H found in curettage but MSS in the final hysterectomy sample. Retesting of the sample using Sanger sequencing changed the curettage result to MSS. Cases 3 and 4 showed discordancy in TP53 mutation results and remained discordant after retesting. p.G245S and p.R175H are missense mutations. Reasons for discordance might also attribute to inadequate tumor sampling and spatial heterogeneity of the tumor.
Patient cohort The study flow chart is presented in Figure . A total of 120 EC patients with post‐operative specimens from Sun Yat‐Sen Memorial Hospital between 2018 and 2021 were included. 40 patients were excluded for failed sequencing or insufficient tumor tissue for DNA extraction. The remaining 80 patients were compared for MMR/p53 IHC, grade, stages, histotype, and molecular subgroup assignment by a simplified NGS panel (Figure ). 35 of 80 cases that had paired curettage specimens were further analyzed of NGS‐based molecular classification and traditional pathological classification.
Description statistics of EC patients according to NGS molecular subgrouping Eighty patients with post‐operative specimens were qualified for analysis. The pathological characteristics and demographics of these patients were described in Table . The median patient age at diagnosis was 60 years (range from 35 to 75 years). The majority of cases (62, 77.5%) were endometrioid histotypes, 4 (5%) were serous, and the rest were clear cell and mixed histology. Grade distribution shown 17 (21.3%) grade 1, 34 (42.5%) grade 2, and 29 (36.2%) grade 3. Molecular classification using simplified NGS panel yielded 4 molecular subgroups: 8 (10%) POLE , 21 (26.2%) MSI‐H, 30 (37.5%) TP53 wt, and 21 (26.3%) TP53 abn. A small proportion (5, 6.25%) of patients demonstrated more than one molecular feature. ECs harbored POLE exonuclease domain mutations were mostly endometrioid (87.5%) and consistent with previous findings, p53abn patients were older, serous (61.9%), higher rate of myometrial invasion ≥50% (47.6%) and were correlated with high stage and grade (G3 [76.2%], stages II ~ IV [47.6%]). The distribution of the four molecular subgroups with EC patients was similar to the proportion reported in previous studies (Figure ), proving the accurate discriminatory ability of our NGS classifier.
Concordance between NGS‐based MSI status detection and MMR Immunohistochemistry on post‐operative specimens Immunohistochemistry (the protein expression of MLH1, MSH2, MSH6, and PMS2) and pentaplex PCR‐based assays are the two common methodologies for the assessment of MSI phenotype in EC classification. In this study, we analyzed the MSI status using NGS based on read‐count distribution and compared the agreement with MMR IHC. A total of 32 cases with post‐operative specimens that had MMR (MLH1/PMS2, MSH2/MSH6) IHC reports in system from our center were picked out and subsequently performed NGS‐based MSI detection. Finally, 32 cases have been compared. As shown in Table , 25 cases were MMR IHC intact and their DNA sequencing results showed microsatellite stable (MSS), while 6 cases were MSI‐H and MMR protein‐deficient (Figure ). However, one case with MSI‐H did not find expression loss for all four MMR genes. Thus, a high agreement rate (31/32, 96.9%) was observed between these two methods, inconsistency was only seen in one case.
Concordance of TP53 mutational analysis and p53 Immunohistochemistry on post‐operative specimens Scoring of p53 IHC sections between the three gynecologic pathologists was evaluated within the 65 hysterectomy samples. Complete negative staining or strongly positive (set to >70%, >80%, >85%, and > 90% of tumor cells for comparison in our study) was interpreted as abnormal. As detailed in Table , the concordance rate of p53 IHC and TP53 mutational analysis lay between 60% to 81.5%, which was highest when using >80% and >85% of tumor cells as criteria, suggesting TP53 mutational analysis could be utilized in helping p53 IHC define the right cutoff score of aberrant positive staining in clinical practice. Figure showed consistency between expression patterns of p53 IHC and TP53 mutation patterns. Information about 12 discordant cases were summarized in Table . Here we set an optimal threshold of tumor cell content ≥80%. Patients from 1 to 10 had high expression of TP53 in protein level which was considered abnormal, while no DNA mutation was found. Conversely, both patient 11 and patient 12 detected TP53 mutations. To further explore the correlation between clinicopathological features and p53 IHC, and TP53 mutation subgrouping, we analyzed the distribution of high‐risk features, including high grade, stage, LVSI, non‐endometrioid type, myometrial invasion (≥50%), in TP53 abn and p53 IHC abn subgroup. Intriguingly, a much more common high‐risk features distribution was observed in TP53 abn subgroup rather than p53 IHC abn subgroup (Table ). TP53 mutation can occur in POLE mut or MSI‐H EC, but is often considered a passenger mutation. When excluded EC with more than one molecular feature, the correlation between high‐risk features and TP53 abn subgroup turned out more significant.
Concordance of histotype and tumor grade between curettage and subsequent hysterectomy specimens The descriptive analysis, including patient demographics, tumor grade, histotype, and molecular subgroups for 35 patients with paired curettage and subsequent hysterectomy specimens were detailed in Table and Table . Table showed the concordance metrics for grade and histotype. The overall concordance rate for grade and histotype was only 74.29% (26/35) and 54.29% (19/35), confirming the lack of reproducibility of the results of the pathologic assessments between curettage and hysterectomy specimens. Downgrading was found in 8.57% (3/35) and upgrading was found in 14.29% (5/35) of the cases, one representative discordant case was illustrated in Figure .
Concordance of the molecular classification between curettage and subsequent hysterectomy specimens Concordance metrics for NGS‐based molecular classification comparing curettage and subsequent hysterectomy specimens was illustrated in Table . The overall accuracy and Cohen's kappa value were 0.88 and 0.84, respectively (Table ), demonstrating a great improvement over the conventional pathological assessment of grade and histotype.
Interrogation of patients discordant on NGS molecular classification Four patients (4/35) with discordant results between curettage and subsequent hysterectomy specimens as assessed by NGS classifier were summarized in Supplement table 6. In Cases 1, 3, and 4. The result was not changed after retesting and reassessment. For case 1, a POLE mutation (p.T278A) was detected in the curettage sample but not in the post‐staging sample. The p.T278A is a missense mutation with uncertain clinical significance. The discordance might be related to the spatial heterogeneity of the tumor. Case 2 showed discordant results in MSI status, with MSI‐H found in curettage but MSS in the final hysterectomy sample. Retesting of the sample using Sanger sequencing changed the curettage result to MSS. Cases 3 and 4 showed discordancy in TP53 mutation results and remained discordant after retesting. p.G245S and p.R175H are missense mutations. Reasons for discordance might also attribute to inadequate tumor sampling and spatial heterogeneity of the tumor.
DISCUSSION Histological classification of EC provides important prognostic information to help determine appropriate surgery and adjuvant therapy. However, pathological classification by histological subtype and grade was considered highly subjective and has reproducibility challenges, and overlap between histological subtypes, and grade leads to complicating clinical decision‐making, and significant interobserver variability further hamper histological classification. Studies have shown interobserver disagreement is about 10%–20% and may reach 30% or higher in high‐grade endometrial carcinoma. Research from C Blake Gilks shows the agreement is only 62.5% for three pathologists. During the last decade, the EC molecular classification introduced by The Cancer Genome Atlas has provoked a transition toward molecular‐based classification with clear prognostic value. The TCGA project defined four distinct prognostic EC subtypes based on somatic copy number alterations (SCNA) and tumor mutational burden, raising the possibility of more precise guidance of adjuvant therapy, surgery, and disease surveillance. Although TCGA represented a meaningful step toward informative classification, it was impractical. Using a simplified molecular classifier that identifies four molecular subtypes that are analogous to TCGA, a highly successful rate (>95%) was shown in the PORTEC trials, demonstrating strong prognostic value in EC. , In this study, we applied a simplified NGS panel to categorize EC molecular type, and additionally, compared the level of concordance between endometrial biopsies and subsequent hysterectomy specimens. The distribution of the four molecular subgroups identified by our NGS classifier was similar to the proportion reported in TCGA, demonstrating the discriminatory ability of our NGS classifier. Consistent with previous findings, , , p53 abn patients were older, mostly serous, and diagnosed at a high stage and grade (G3 [76.2%], stages II ~ IV [47.6%]). They showed other aggressive characteristics such as myometrial invasion (47.6%) and LVSI (61.9%). Notably, the subtype with the second most aggressive features was POLE in our study, with myometrial invasion (37.5%) and LVSI (75.0%) comparable to p53 abn tumors, and much higher than observation in women with MSI‐H or p53 wt ECs. The association with aggressive features of POLE subtype was also observed in other studies and was not related to their exceptionally better prognosis. , Taken together, our NGS panel‐based EC classification proved to be simplified and pragmatic, the result demonstrated the molecular feature of Chinese EC patients. EC is a tumor type associated with MSI‐H and MMR. However, the result for immunohistochemical staining of MMR protein was not always consistent with DNA loci testing. In this study, we found one case with MSI‐H in DNA level but all four MMR protein expressions were intact. This may be caused by inactivating missense mutation which did not affect protein expression or inactivation of other repair genes such as MSH3 . Immunohistochemical staining of tumors for p53 has a long history and has been widely used in clinical. Unfortunately, the interpretation of IHC results varies among observers, which complicates clinical decision‐making. Moreover, p53 proteins have relatively short half‐lives and their detection is therefore dependent on prompt fixation. , Here we attempted to establish an approach to provide auxiliary judgments for IHC results by correlating IHC patterns with the underlying mutation status, which pathologists can use in daily practice. It has been shown that the concordance between TP53 mutation status and p53 overexpression, which was individually detected by NGS and immunohistochemistry, was 92% in EC. In gastric cancer, TP53 missense mutation, and p53 overexpression were highly consistent (90.9%). The concordance rate between p53 IHC and TP53 mutation we observed reached the highest level at 81.5% when the IHC score standard was set at 80% of tumor cells, which suggests that TP53 mutation status could be utilized in defining p53 IHC patterns. To further illustrate the correlation between clinicopathological features and p53 IHC, TP53 mutation, we compared the proportion of high‐risk features in TP53 abn subgroup and p53 IHC abn subgroup and intriguingly found that TP53 abn group had an even higher proportion of high‐risk features than p53 IHC abn group, no matter POLE mut and dMMR tumors were excluded or not. One possible explanation of this finding is that not all p53 are uniformly expressed in all tissues. In pancreatic ductal adenocarcinoma, TP53 missense mutation was confirmed to be associated with nuclear P53 overexpression in ≥25% of neoplastic cells. In diffuse large B‐cell lymphoma patients treated with R‐CHOP immunohistochemical analysis showing >50% cells expressing p53 protein was able to stratify patients with significantly different prognoses. The result also might be due to the smaller sample size in the study. Taken together, our results showed that NGS‐based TP53 mutation analysis was highly consistent with IHC patterns, but has advantages in identifying high‐risk clinicopathological features, and could be utilized in defining IHC score. Examination of curettage specimens is a common method for early diagnosis of EC and helps doctors obtain risk stratification in EC patients before operation. However, the histological types and grade of tumor differentiation of cancer diagnosed in endometrial curettage and post‐operative specimens were highly heterogenetic. , A study implicated that the highest concordance (127/148, 85.81%) between the histological diagnosis on curettage and post‐operative specimens were found for endometrioid carcinoma (21/44, 47.73%), and postoperative pathological examination is more accurate for diagnosis of atypical endometrial hyperplasia. In addition, in a study that compared the FIGO grades between preoperative endometrial samplings and hysterectomy specimens, the accuracy for grade 3 tumors (90.4%, N = 98) was significantly higher than for grade 1. In our research, the overall concordance rate of grade and histotype in curettage sample vs. final hysterectomy samples were 54.2% and 74.29%, with downgrading was found in 8.6% (3/35) and upgrading was found in 14.3% (5/35) of the endometrial samples. These data confirmed the previously reported lack of reproducibility when comparing curettage specimens with hysterectomy specimens. Unreliable histotype and grade assignment led to the inconsistent categorization of EC, which offered confusing information and brought barriers to clinical decision‐making. Conversely, reproducibility at a high level (88%) was seen with our NGS‐based molecular subclassification. Our study is the first to use NGS‐based molecular classification techniques for curettage specimens, and compare the molecular features of the curettage specimens with post‐operative specimens. In this way, we successfully classified all endometrial cancers using the NGS‐based classifier at initial diagnosis. Inconsistencies also arise, as in our study, which may be related to intratumor heterogeneity due to curettage of a small segment of tumor. As illustrated by Marta Brunetti, EC tumors are characterized by a high degree of tumoral heterogeneity. Several limitations should be acknowledged in this study. Firstly, the overall sample size of our research was limited, particularly for those cases with curettage and paired hysterectomy specimens. Secondly, prognostic data are required to determine the prognostic discriminatory ability of the four molecular subgroups identified by our NGS classifier. Thirdly, the p53 IHC and MMR IHC results of curettage reports collected in our system were insufficient for statistics. Molecular subgrouping has been fundamental in evolving the evaluation system and represents a new trend of EC, but there are still challenges that need to be solved. Using a small NGS panel, we have demonstrated the distribution of the four molecular subgroups in Chinese EC patients, and more importantly, be the first research that applied NGS method to curettage specimens and compared concordance between endometrial biopsies and subsequent hysterectomy specimens. We provided insight into the molecular classifier of EC, thus expanding its potential clinical application. Our next steps will focus on determining the optimal utilization of our NGS classifier in clinical trials, and explore whether our NGS classifier can guide fertility‐saving treatments, surgery, adjuvant therapy, and surveillance to improve outcomes for EC patients.
Qunxian Rao: Writing – original draft (equal). Jianwei Liao: Writing – original draft (equal); writing – review and editing (equal). Yangyang Li: Data curation (equal). Xin Zhang: Data curation (equal). Guocai Xu: Methodology (equal). Changbin Zhu: Software (equal). Shengya Tian: Methodology (equal). Qiuhong Chen: Data curation (equal). Hui Zhou: Data curation (equal). bingzhong zhang: Conceptualization (equal); writing – original draft (equal).
This study was supported by Grants from the National Natural Science Foundation of China (No. 81903043) and the Natural Science Foundation of Guangdong Province (No. 2018A030310086).
Shengya Tian, Changbin Zhu, and Qiuhong Chen were employed by Amoy Diagnostics Co., Ltd. Other authors declared no conflict of interest.
Figure S1 Click here for additional data file. Figure S2 Click here for additional data file. Figure S3 Click here for additional data file. Figure S4 Click here for additional data file. Table S1 Table S2 Table S3 Table S4 Table S5 Table S6 Click here for additional data file.
|
Cancer treatments touch a wide range of values that count for patients and other stakeholders: What are the implications for decision‐making? | d921e7ce-3567-4911-beb8-8676768289d2 | 10028089 | Internal Medicine[mh] | INTRODUCTION The worldwide cancer rate is expected to increase from 19 million in 2020 to 30 million in 2040. Costs of cancer care are high and continue to rise. In Europe, costs have almost doubled from €52 billion in 1995 to €103 billion in 2018. Over the years, patient‐centred care is gaining increasing interest in which the needs and desires of individual patients drive the force behind healthcare decisions and quality measurements. , , , For payors and policymakers, it is challenging to provide optimal healthcare needs for individual patients while managing budgets. Decision‐making on reimbursement and resource allocation is often aided by using Health Technology Assessment (HTA). , However, HTA frameworks do not always reflect all values that count for patients and other stakeholders, therefore, various different value frameworks have been published in recent years, , , , for instance, by the American Society of Clinical Oncology (ASCO) and European Society for Medical Oncology (ESMO). Previous studies compared such oncological value frameworks and revealed some inconsistencies which mainly derived from the differences in perspective. , , , Additionally, it is argued that they do not take the unique aspects of the evolving therapeutic landscape with targeted therapy, immunotherapy and more precision medicine into consideration, which has led to increasing uncertainty about the true value of new treatments. , The International Society for Pharmaceutical Outcomes Research (ISPOR) Value Assessment Frameworks Special Task Force emphasized gaps in value assessment in general as elements from a societal perspective are missing, for example, equity. , It is also argued that the commonly used quality of life (QoL) outcome measures do not always seem adequate for mapping all aspects in the social domain ; they contain a subset of relevant outcomes. , , , , Over the years, different disease‐specific QoL measurement tools have been developed. What exactly constitutes value to whom in the context of cancer treatments, limited financial budgets and patient‐centred care is still not clear and varies between different stakeholders. The aim of this study is therefore to explore values–elements regarding new oncological treatments and to assess their implications in decision‐making procedures to inform the oncological field and generate the maximum value with fixed national budgets.
METHOD 2.1 Study design We conducted semi‐structured interviews to explore values regarding new oncological treatments from different stakeholder perspectives. Values are defined as elements that warrant consideration in treatment assessment and decision‐making procedures. In addition, a focus group with an expert panel was performed to discuss the usefulness and potential implications of the values that were found in the interviews in different decision‐making procedures. The study was reported based on recommendations in the consolidated criteria for reporting qualitative studies (COREQ) checklist. 2.2 Study setting This study was performed in the Netherlands. See Box for a description of the Dutch healthcare system and see Kroneman et al. for a more extended description. 2.3 Study population To identify values within oncology, 12 stakeholder groups were defined for interviews: patients, partners of patients, oncologists, oncological nurses, occupational doctors, general practitioners, insurance advisors, the National Healthcare Institute (ZIN), Netherlands Comprehensive Cancer Center, the Ministry of Health (MOH), Health insurance companies and dedicated oncological care networks. From each group, participants for the interviews were purposively sampled to ensure a minimum of two interviews. We approached 41 potential participants by email through our investigator network, stakeholder websites and snowball sampling. To discuss the usefulness and potential implications of the values in decision‐making procedures, an expert panel was purposively sampled with experts on quality of life, ethics, HTA and spokesmen of patients, ZIN, health insurance companies and healthcare professionals. Eleven experts were approached. Stakeholder selection was based on screening of Dutch policy documents and literature and consultation with experts. The interviews and expert panel meetings were performed by video calls, and informed consent was obtained from all study participants. 2.4 Data collection For the interviews, we drafted a topic guide (Appendix ) on overarching themes found in a literature search (Appendix ). These themes were societal impact, quality of life, impact on daily life, family burden, costs and quality of care. We applied an iterative approach in which, depending on the stakeholder and after reflections, slight adjustments or more emphasis on specific topics were made during the interviews. One trained researcher (CV) conducted the interviews. The topic guide of the expert focus group (Appendix ) was constructed based on findings from the interviews and contained questions regarding the desire and current use of values in decision‐making procedures, within which context, it was desirable and with which methods values could be best measured. The expert panel was moderated by an experienced researcher (RH) and planned for 60 minutes. 2.5 Data processing and analysis Interviews and the focus group with the expert panel were audio‐recorded and transcribed verbatim. The transcripts were thematically analysed. Codes were assigned to relevant text passages and codes referring to the same underlying concept were divided into subcategories and themes. Twelve transcripts, one from each stakeholder group, were coded by two trained researchers individually (CV and ES). After comparing transcripts, the researchers reached a consensus on the codes, subcategories and themes. The remaining transcripts were assessed by one researcher (CV) and discussed with the second researcher (ES) until a consensus was reached. All generated themes and subcategories are reported and listed indiscriminatory in the results. Atlas.ti (version 8) was used for data processing and analyses.
Study design We conducted semi‐structured interviews to explore values regarding new oncological treatments from different stakeholder perspectives. Values are defined as elements that warrant consideration in treatment assessment and decision‐making procedures. In addition, a focus group with an expert panel was performed to discuss the usefulness and potential implications of the values that were found in the interviews in different decision‐making procedures. The study was reported based on recommendations in the consolidated criteria for reporting qualitative studies (COREQ) checklist.
Study setting This study was performed in the Netherlands. See Box for a description of the Dutch healthcare system and see Kroneman et al. for a more extended description.
Study population To identify values within oncology, 12 stakeholder groups were defined for interviews: patients, partners of patients, oncologists, oncological nurses, occupational doctors, general practitioners, insurance advisors, the National Healthcare Institute (ZIN), Netherlands Comprehensive Cancer Center, the Ministry of Health (MOH), Health insurance companies and dedicated oncological care networks. From each group, participants for the interviews were purposively sampled to ensure a minimum of two interviews. We approached 41 potential participants by email through our investigator network, stakeholder websites and snowball sampling. To discuss the usefulness and potential implications of the values in decision‐making procedures, an expert panel was purposively sampled with experts on quality of life, ethics, HTA and spokesmen of patients, ZIN, health insurance companies and healthcare professionals. Eleven experts were approached. Stakeholder selection was based on screening of Dutch policy documents and literature and consultation with experts. The interviews and expert panel meetings were performed by video calls, and informed consent was obtained from all study participants.
Data collection For the interviews, we drafted a topic guide (Appendix ) on overarching themes found in a literature search (Appendix ). These themes were societal impact, quality of life, impact on daily life, family burden, costs and quality of care. We applied an iterative approach in which, depending on the stakeholder and after reflections, slight adjustments or more emphasis on specific topics were made during the interviews. One trained researcher (CV) conducted the interviews. The topic guide of the expert focus group (Appendix ) was constructed based on findings from the interviews and contained questions regarding the desire and current use of values in decision‐making procedures, within which context, it was desirable and with which methods values could be best measured. The expert panel was moderated by an experienced researcher (RH) and planned for 60 minutes.
Data processing and analysis Interviews and the focus group with the expert panel were audio‐recorded and transcribed verbatim. The transcripts were thematically analysed. Codes were assigned to relevant text passages and codes referring to the same underlying concept were divided into subcategories and themes. Twelve transcripts, one from each stakeholder group, were coded by two trained researchers individually (CV and ES). After comparing transcripts, the researchers reached a consensus on the codes, subcategories and themes. The remaining transcripts were assessed by one researcher (CV) and discussed with the second researcher (ES) until a consensus was reached. All generated themes and subcategories are reported and listed indiscriminatory in the results. Atlas.ti (version 8) was used for data processing and analyses.
RESULTS For the interviews, 41 stakeholders were approached of which 32 participated in 31 separate interviews (Table ) with at least two interviews per stakeholder group. Of the non‐participators, one declined based on time constraints, six did not respond to email contact and two believed not to be the right person for this study. The interviews lasted between 20 and 64 minutes. For the focus group, seven experts participated: an expert on health‐related quality of life, ethics, health technology assessment and a spokesman of a patient organization, the ZIN, a health insurance company and an oncologist. Of the non‐participators, one was not available on the date of the expert panel and three did not feel able to contribute. 3.1 Interviews stakeholders The themes generated from the interviews were impact on daily life and future, costs for patients and loved ones, quality of life, impact on loved ones, societal impact and quality of treatments. The themes and subcategories generated from the interviews are presented in Table . Quotes from the interviews are presented in Table . 3.1.1 Impact on daily life and future of patients The interviewees mentioned several values that can be impacted by a cancer diagnosis or its treatment in daily or future life. Participants mentioned the ability to continue daily activities like sports, hobbies, doing one's own groceries or being independent. It was mentioned that patients can participate less in society and in activities such as working, volunteering, caring for children and being an informal caregiver. Participants stated that the importance of re‐integration or continuing work can differ per patient. Different participants mentioned the inability to make future plans and life choices. In addition, participants stated that mortgages are more difficult to get for (former) patients as they pay higher premiums or are rejected for life insurance, which they need to get a mortgage. 3.1.2 Costs for patients and loved ones The interviewees mentioned values regarding costs for patients and loved ones. A few participants stated that some additional healthcare or complementary care needs to be (partly) paid for by the patients themselves. Another cost aspect mentioned in the interviews was the deductible premium for the health insurance which is statutory in the Netherlands, even after treatments, because of a long‐term need for additional care. However, it was also stated that these problems differ per patient as it depends on a patient's personal financial situation. Additional indirect costs for patients that were mentioned were loss of work and income. This problem also concerns partners as their ability to work might be affected as well (partly), because of caring for children, accompanying patients to the hospital or delivering informal care. Additionally, costs for re‐integration after patients have lost their job or for those who are self‐employed. Additional costs are spent on higher premiums for life insurance or disability insurance in case of self‐employment. Other examples are travel costs for patients and family members, house renovations, new wardrobes after weight loss or gain, wigs and sometimes prosthetics or specialized bras and swimwear. 3.1.3 Quality of life The interviewees mentioned values regarding the quality of life. Quality of life values is sub‐divided into physical, psychological, social and spiritual. It was mentioned that patients might be in need of support or companionship for these domains. First, a cancer diagnosis can result in many physical consequences. It was mentioned that consequences differ depending on the treatment or tumour type. For instance, early menopause and infertility are common for patients with gynaecological cancers. For colon cancer, bowel dysfunction and stomas are common. Second, participants stated that the psychological consequences can be quite severe. An often‐mentioned psychological effect is fear of death. Also, even after being cured, there is fear of recurrence or suffering from damage caused by cancer treatments. It was mentioned that some of the psychological impacts occurs after the passage of a certain time interval, that is when treatments are already completed. Psychological consequences might result in higher healthcare costs. Conflicting opinions existed about the value of hope. Hope can benefit a patient's quality of life but can also interfere with acceptance of the situation. In addition, different participants emphasized to focus more on positivity, fighting spirit and empowerment of patients instead of negative psychological consequences. Third, participants state that social needs can be negatively influenced, for instance, by not attending social gatherings or not being energetic when a patient does attend a gathering. Additionally, the effect of these cancer treatments can affect patients' relationships. Last, some spiritual aspects of quality of life were mentioned. Patients can find support in religion or existential questioning about the meaning of life. 3.1.4 Impact on loved ones The interviewees mentioned values regarding the impact on loved ones like psychological impact, workability, the positive or negative impact on relationships, social activities or the impact on future plans. Loved ones are often informal caregivers to patients. This entails being attentive to patients, driving patients to the hospitals and also providing care for patients at home. In addition, they can experience a change in mentality during the course of the disease. Loved ones might feel the need for support and companionship. 3.1.5 Societal impact Interviewees mentioned values regarding societal impact. The most frequently mentioned was the loss of productivity and lack of ability to return to work. These costs are borne by the patient (by not receiving wages), the employer and by society. In addition, costs are made on benefits and allowances paid to people losing their job or that are working less. Patients often have a certain capacity to work during the illness which is not always utilized. One participant mentioned the possibility of increased healthcare costs because of lower work productivity as it can affect a patient's psychological well‐being. Other societal impacts mentioned were the loss of informal caregivers because the oncological patient cannot care for another family member or indirect healthcare costs on long‐term effects like heart diseases as a result of chemotherapy. Participants often mention the importance of societal balancing of resource allocation as premium money paid on health insurance by society must be handled with caution and the most value for money should be created. Additional values that were mentioned to take into consideration were the prevalence of the disease, the development of new knowledge, innovation, freedom of choice for patients, access to care, equality, the necessity of a new drug and implementation feasibility. 3.1.6 Quality of treatments The interviewees mentioned values regarding the quality of treatments. It was often mentioned that patients should always get the best possible care and treatments should comply with the established medical sciences and practice. Besides survival or extending life, quality of life outcomes were deemed important. In addition, oncological treatments can cause serious long‐term health problems, like heart failure or other cancers. It was mentioned that side effects and ease of use of treatments should be proportional to the benefit. The ease of use includes, among others, the possibility and choice for care at home or the need for problems with transportation. Finally, it was mentioned that therapy compliance should be optimized. 3.2 Focus group with an expert panel For patient‐level decision‐making, the extended exploration of values was deemed useful to ensure the inclusion of all values that count for patients in decision‐making that are important for a specific patient. However, the usefulness of many values also depends on the situation of the patient; and different aspects are important in different phases of the disease (i.e., palliative or curative), phase of treatment (i.e., when you start treatment or when you have a recurrence) and age of patients. It was mentioned that not all values are always relevant regarding choices of treatments (or not treating). The values were mentioned to be important to discuss with patients regarding long‐term effects and long‐term quality of life, for general information provision and empowerment of patients. In addition, the extended list of values was mentioned to be of importance in discussions on a patient's life after treatment or after cancer. It was mentioned that patient values are currently already being investigated to some extent during the treatment process, for instance, by means of questionnaires. For reimbursement decisions, the use of the extended exploration of values was less clear. It was mentioned that the Netherlands has a mechanism of decision‐making using the QALY framework which is already incorporated many values. Other values are also considered, like the ease of use, although participants questioned whether national solidarity stretches as far as to reimburse treatments from national budgets due to them being specifically easier to use. Regarding the measurement methods of values, the expert panel mentioned that some values have established methods, while others do not. Potential disadvantages of incorporating the extended exploration of values for reimbursement decisions were mentioned: (1) Risking an increase of the sustainability dilemma as more therapies might be reimbursed, (2) the uncertainty of how the value of, for example hope is proportional to benefits regarding survival and (3) the desire to use generic values that are applicable to all diseases instead of specific diseases. Table presents quotes from the expert panel.
Interviews stakeholders The themes generated from the interviews were impact on daily life and future, costs for patients and loved ones, quality of life, impact on loved ones, societal impact and quality of treatments. The themes and subcategories generated from the interviews are presented in Table . Quotes from the interviews are presented in Table . 3.1.1 Impact on daily life and future of patients The interviewees mentioned several values that can be impacted by a cancer diagnosis or its treatment in daily or future life. Participants mentioned the ability to continue daily activities like sports, hobbies, doing one's own groceries or being independent. It was mentioned that patients can participate less in society and in activities such as working, volunteering, caring for children and being an informal caregiver. Participants stated that the importance of re‐integration or continuing work can differ per patient. Different participants mentioned the inability to make future plans and life choices. In addition, participants stated that mortgages are more difficult to get for (former) patients as they pay higher premiums or are rejected for life insurance, which they need to get a mortgage. 3.1.2 Costs for patients and loved ones The interviewees mentioned values regarding costs for patients and loved ones. A few participants stated that some additional healthcare or complementary care needs to be (partly) paid for by the patients themselves. Another cost aspect mentioned in the interviews was the deductible premium for the health insurance which is statutory in the Netherlands, even after treatments, because of a long‐term need for additional care. However, it was also stated that these problems differ per patient as it depends on a patient's personal financial situation. Additional indirect costs for patients that were mentioned were loss of work and income. This problem also concerns partners as their ability to work might be affected as well (partly), because of caring for children, accompanying patients to the hospital or delivering informal care. Additionally, costs for re‐integration after patients have lost their job or for those who are self‐employed. Additional costs are spent on higher premiums for life insurance or disability insurance in case of self‐employment. Other examples are travel costs for patients and family members, house renovations, new wardrobes after weight loss or gain, wigs and sometimes prosthetics or specialized bras and swimwear. 3.1.3 Quality of life The interviewees mentioned values regarding the quality of life. Quality of life values is sub‐divided into physical, psychological, social and spiritual. It was mentioned that patients might be in need of support or companionship for these domains. First, a cancer diagnosis can result in many physical consequences. It was mentioned that consequences differ depending on the treatment or tumour type. For instance, early menopause and infertility are common for patients with gynaecological cancers. For colon cancer, bowel dysfunction and stomas are common. Second, participants stated that the psychological consequences can be quite severe. An often‐mentioned psychological effect is fear of death. Also, even after being cured, there is fear of recurrence or suffering from damage caused by cancer treatments. It was mentioned that some of the psychological impacts occurs after the passage of a certain time interval, that is when treatments are already completed. Psychological consequences might result in higher healthcare costs. Conflicting opinions existed about the value of hope. Hope can benefit a patient's quality of life but can also interfere with acceptance of the situation. In addition, different participants emphasized to focus more on positivity, fighting spirit and empowerment of patients instead of negative psychological consequences. Third, participants state that social needs can be negatively influenced, for instance, by not attending social gatherings or not being energetic when a patient does attend a gathering. Additionally, the effect of these cancer treatments can affect patients' relationships. Last, some spiritual aspects of quality of life were mentioned. Patients can find support in religion or existential questioning about the meaning of life. 3.1.4 Impact on loved ones The interviewees mentioned values regarding the impact on loved ones like psychological impact, workability, the positive or negative impact on relationships, social activities or the impact on future plans. Loved ones are often informal caregivers to patients. This entails being attentive to patients, driving patients to the hospitals and also providing care for patients at home. In addition, they can experience a change in mentality during the course of the disease. Loved ones might feel the need for support and companionship. 3.1.5 Societal impact Interviewees mentioned values regarding societal impact. The most frequently mentioned was the loss of productivity and lack of ability to return to work. These costs are borne by the patient (by not receiving wages), the employer and by society. In addition, costs are made on benefits and allowances paid to people losing their job or that are working less. Patients often have a certain capacity to work during the illness which is not always utilized. One participant mentioned the possibility of increased healthcare costs because of lower work productivity as it can affect a patient's psychological well‐being. Other societal impacts mentioned were the loss of informal caregivers because the oncological patient cannot care for another family member or indirect healthcare costs on long‐term effects like heart diseases as a result of chemotherapy. Participants often mention the importance of societal balancing of resource allocation as premium money paid on health insurance by society must be handled with caution and the most value for money should be created. Additional values that were mentioned to take into consideration were the prevalence of the disease, the development of new knowledge, innovation, freedom of choice for patients, access to care, equality, the necessity of a new drug and implementation feasibility. 3.1.6 Quality of treatments The interviewees mentioned values regarding the quality of treatments. It was often mentioned that patients should always get the best possible care and treatments should comply with the established medical sciences and practice. Besides survival or extending life, quality of life outcomes were deemed important. In addition, oncological treatments can cause serious long‐term health problems, like heart failure or other cancers. It was mentioned that side effects and ease of use of treatments should be proportional to the benefit. The ease of use includes, among others, the possibility and choice for care at home or the need for problems with transportation. Finally, it was mentioned that therapy compliance should be optimized.
Impact on daily life and future of patients The interviewees mentioned several values that can be impacted by a cancer diagnosis or its treatment in daily or future life. Participants mentioned the ability to continue daily activities like sports, hobbies, doing one's own groceries or being independent. It was mentioned that patients can participate less in society and in activities such as working, volunteering, caring for children and being an informal caregiver. Participants stated that the importance of re‐integration or continuing work can differ per patient. Different participants mentioned the inability to make future plans and life choices. In addition, participants stated that mortgages are more difficult to get for (former) patients as they pay higher premiums or are rejected for life insurance, which they need to get a mortgage.
Costs for patients and loved ones The interviewees mentioned values regarding costs for patients and loved ones. A few participants stated that some additional healthcare or complementary care needs to be (partly) paid for by the patients themselves. Another cost aspect mentioned in the interviews was the deductible premium for the health insurance which is statutory in the Netherlands, even after treatments, because of a long‐term need for additional care. However, it was also stated that these problems differ per patient as it depends on a patient's personal financial situation. Additional indirect costs for patients that were mentioned were loss of work and income. This problem also concerns partners as their ability to work might be affected as well (partly), because of caring for children, accompanying patients to the hospital or delivering informal care. Additionally, costs for re‐integration after patients have lost their job or for those who are self‐employed. Additional costs are spent on higher premiums for life insurance or disability insurance in case of self‐employment. Other examples are travel costs for patients and family members, house renovations, new wardrobes after weight loss or gain, wigs and sometimes prosthetics or specialized bras and swimwear.
Quality of life The interviewees mentioned values regarding the quality of life. Quality of life values is sub‐divided into physical, psychological, social and spiritual. It was mentioned that patients might be in need of support or companionship for these domains. First, a cancer diagnosis can result in many physical consequences. It was mentioned that consequences differ depending on the treatment or tumour type. For instance, early menopause and infertility are common for patients with gynaecological cancers. For colon cancer, bowel dysfunction and stomas are common. Second, participants stated that the psychological consequences can be quite severe. An often‐mentioned psychological effect is fear of death. Also, even after being cured, there is fear of recurrence or suffering from damage caused by cancer treatments. It was mentioned that some of the psychological impacts occurs after the passage of a certain time interval, that is when treatments are already completed. Psychological consequences might result in higher healthcare costs. Conflicting opinions existed about the value of hope. Hope can benefit a patient's quality of life but can also interfere with acceptance of the situation. In addition, different participants emphasized to focus more on positivity, fighting spirit and empowerment of patients instead of negative psychological consequences. Third, participants state that social needs can be negatively influenced, for instance, by not attending social gatherings or not being energetic when a patient does attend a gathering. Additionally, the effect of these cancer treatments can affect patients' relationships. Last, some spiritual aspects of quality of life were mentioned. Patients can find support in religion or existential questioning about the meaning of life.
Impact on loved ones The interviewees mentioned values regarding the impact on loved ones like psychological impact, workability, the positive or negative impact on relationships, social activities or the impact on future plans. Loved ones are often informal caregivers to patients. This entails being attentive to patients, driving patients to the hospitals and also providing care for patients at home. In addition, they can experience a change in mentality during the course of the disease. Loved ones might feel the need for support and companionship.
Societal impact Interviewees mentioned values regarding societal impact. The most frequently mentioned was the loss of productivity and lack of ability to return to work. These costs are borne by the patient (by not receiving wages), the employer and by society. In addition, costs are made on benefits and allowances paid to people losing their job or that are working less. Patients often have a certain capacity to work during the illness which is not always utilized. One participant mentioned the possibility of increased healthcare costs because of lower work productivity as it can affect a patient's psychological well‐being. Other societal impacts mentioned were the loss of informal caregivers because the oncological patient cannot care for another family member or indirect healthcare costs on long‐term effects like heart diseases as a result of chemotherapy. Participants often mention the importance of societal balancing of resource allocation as premium money paid on health insurance by society must be handled with caution and the most value for money should be created. Additional values that were mentioned to take into consideration were the prevalence of the disease, the development of new knowledge, innovation, freedom of choice for patients, access to care, equality, the necessity of a new drug and implementation feasibility.
Quality of treatments The interviewees mentioned values regarding the quality of treatments. It was often mentioned that patients should always get the best possible care and treatments should comply with the established medical sciences and practice. Besides survival or extending life, quality of life outcomes were deemed important. In addition, oncological treatments can cause serious long‐term health problems, like heart failure or other cancers. It was mentioned that side effects and ease of use of treatments should be proportional to the benefit. The ease of use includes, among others, the possibility and choice for care at home or the need for problems with transportation. Finally, it was mentioned that therapy compliance should be optimized.
Focus group with an expert panel For patient‐level decision‐making, the extended exploration of values was deemed useful to ensure the inclusion of all values that count for patients in decision‐making that are important for a specific patient. However, the usefulness of many values also depends on the situation of the patient; and different aspects are important in different phases of the disease (i.e., palliative or curative), phase of treatment (i.e., when you start treatment or when you have a recurrence) and age of patients. It was mentioned that not all values are always relevant regarding choices of treatments (or not treating). The values were mentioned to be important to discuss with patients regarding long‐term effects and long‐term quality of life, for general information provision and empowerment of patients. In addition, the extended list of values was mentioned to be of importance in discussions on a patient's life after treatment or after cancer. It was mentioned that patient values are currently already being investigated to some extent during the treatment process, for instance, by means of questionnaires. For reimbursement decisions, the use of the extended exploration of values was less clear. It was mentioned that the Netherlands has a mechanism of decision‐making using the QALY framework which is already incorporated many values. Other values are also considered, like the ease of use, although participants questioned whether national solidarity stretches as far as to reimburse treatments from national budgets due to them being specifically easier to use. Regarding the measurement methods of values, the expert panel mentioned that some values have established methods, while others do not. Potential disadvantages of incorporating the extended exploration of values for reimbursement decisions were mentioned: (1) Risking an increase of the sustainability dilemma as more therapies might be reimbursed, (2) the uncertainty of how the value of, for example hope is proportional to benefits regarding survival and (3) the desire to use generic values that are applicable to all diseases instead of specific diseases. Table presents quotes from the expert panel.
CONCLUSION AND DISCUSSION Our findings reveal a broad range of values that matter to patients and involved stakeholders regarding new oncological treatments and decision‐making in oncological care. They can be categorized in (1) impact on daily life and future, (2) costs for patients and loved ones, (3) quality of life (physical, psychological, social and spiritual), (4) impact on loved ones, (5) societal impact and (6) quality of treatment. Experts revealed the exploration of values was useful in patient‐level decision‐making. Not all values are always relevant to include in treatment choices, however, they can also be used for patient information and empowerment and support during and after treatment of cancer. The use of the extended exploration of values for reimbursement decisions was less clear. 4.1 Previous literature Previous studies support the broad range of values that are relevant within oncology, going beyond existing value frameworks. Our results suggest that decision‐making at the patient‐level should contain many values and patient preferences. Shared decision‐making (SDM) between physicians and patients can increase incorporation of patient preferences in treatment decisions, however, different studies reveal that SDM is not yet fully implemented and used to its full potential. , , Our results also suggest that not all values found in our study are included in reimbursement decisions. A review of US Value Frameworks by the ISPOR Value Special Task Force supports our findings as they state it is difficult for frameworks to represent values for all decision contexts. They mention that reimbursement decisions should be focused on efficiently allocating resources to maximize population health‐ and patient‐level decisions should include patients' values and preferences, within the larger constraints imposed by decisions on national levels. Different studies found differences across countries regarding reimbursement decisions. , , A study by Angelis et al. examined differences in eight European countries regarding the assessment of the value of new medicines in the context of reimbursement decisions. Most countries implement a type of economic evaluation in addition to the assessment of clinical benefit. The preferred health gain measure usually is the QALY. Additional values beyond the QALY concept are captured to a different extent between countries, explaining some of the heterogeneity in reimbursement decisions. Ease of use, nature of the treatment, public health benefit, social productivity, place in therapeutic strategy and ethical considerations are criteria considered in some countries (either implicitly or explicitly) and not considered in others. 4.2 Comparison to existing value frameworks Different frameworks exist for evaluating the value of health interventions. These frameworks can be generic or oncology specific and can inform reimbursement or patient‐level decisions. The added value of our findings to these existing frameworks is discussed below. In the United States, the Institute for Clinical and Economic Review (ICER) has been evaluating the clinical and economic values of health interventions to aid reimbursement decisions. Their value framework incorporates long‐term value and short‐term affordability. Long‐term value includes incremental cost‐effectiveness and provides the possibility of incorporating additional other benefits or disadvantages and contextual considerations through deliberative processes. Short‐term affordability includes budget impact. In Europe, the EUnetHTA published an HTA core model to aid reimbursement decisions for which different European countries collaborated. The ontology in this model covers the health problem and current use of technology, description and technical characteristics, safety, clinical effectiveness, costs and economic evaluation, ethical analysis, organizational aspect, patient and social aspects and legal aspects. The ICER value framework and the HTA core model are extensive frameworks used for the systematic assessment of new treatments for reimbursement decisions, and our study can add to these frameworks by more explicitly using the values that are specific for oncological care in the deliberative processes. Besides these models, two common oncological value frameworks are the American Society of Clinical Oncology (ASCO) and the European Society for Medical Oncology (ESMO) frameworks. These are used for facilitating shared decision‐making by patients and oncologists and could aid reimbursement decisions. The ASCO and ESMO incorporate clinical benefit, toxicity, QoL and improvement of (cancer‐related) symptoms beyond QoL. In addition, the ASCO adds treatment‐free interval, drug acquisition costs and patient co‐pay, and the ESMO adds daily well‐being, response rate and duration of response. The value frameworks are used to aid physicians in explaining treatment benefits to patients and facilitate shared decision‐making. Our study adds to these cancer‐specific value frameworks by presenting explicit values regarding the impact on daily life and future, costs for patients and loved ones, QoL and quality of treatment. In addition, the impact on loved ones and societal impact are (mostly) lacking in these frameworks. 4.3 Strengths and limitations The main strengths of this study are that we included a wide array of various stakeholders in the interviews and that multiple interviews were held per category, thus ensuring the capture of all the relevant perspectives. To our knowledge, this is the first study to explore a wider set of values in oncology. We acknowledge some limitations. First, possibilities for incorporation of the values found in current decision‐making are only explored qualitatively. Case studies on oncological treatments will provide further insights into the true extent of how these values are included in decision‐making procedures. Second, our study did not include patients from all tumour types. Different patients with different tumour types could reveal additional values mainly for the physical and psychological quality of life. Third, as this study is performed in the Netherlands the results are mainly generalizable to different countries with a comparable healthcare setting (see Box for a description). Finally, the aimal of the research was to explore value–elements for the assessment of new treatments. The themes represent those values, however, the subcategories within the main themes went beyond the scope of new treatment assessment based on the explorative and semi‐structured design of our interviews. Our approach ensures no mentioned values were missed and the focus group was used to assess the usefulness of this broad exploration of values. Despite these limitations, an overall idea of the broad range of values and their inclusion in decision‐making can be generated. 4.4 Conclusion and implications In conclusion, clinical values are not the only ones that matter to oncological patients and involved stakeholders regarding the evaluation of new treatments and decision‐making procedures in oncological care. We found a much broader range of values. The recognition and appreciation of those values might add to patient‐level decision‐making, but the usefulness for reimbursement decisions is less clear. The values add to existing value frameworks used for both patient‐level and reimbursement decisions. In addition, the values might improve patient information, empowerment and support. The results could be useful to guide clinicians and policymakers when it comes to decision‐making in oncology. We recommend exploring a more structural and explicit incorporation of values within oncology in patient‐level and reimbursement decision processes. At the patient level, the list of values can inform clinicians on which values to address in SDM, can be used for decision aids and can be used to provide extended patient information. For reimbursement decisions, it would be interesting to explore how the identified values can contribute. More research is needed on making explicit, for different oncological indications, if and how disease‐specific values can be systematically inventoried and incorporated to guarantee a more systematic approach to decision‐making and the deliberative processes.
Previous literature Previous studies support the broad range of values that are relevant within oncology, going beyond existing value frameworks. Our results suggest that decision‐making at the patient‐level should contain many values and patient preferences. Shared decision‐making (SDM) between physicians and patients can increase incorporation of patient preferences in treatment decisions, however, different studies reveal that SDM is not yet fully implemented and used to its full potential. , , Our results also suggest that not all values found in our study are included in reimbursement decisions. A review of US Value Frameworks by the ISPOR Value Special Task Force supports our findings as they state it is difficult for frameworks to represent values for all decision contexts. They mention that reimbursement decisions should be focused on efficiently allocating resources to maximize population health‐ and patient‐level decisions should include patients' values and preferences, within the larger constraints imposed by decisions on national levels. Different studies found differences across countries regarding reimbursement decisions. , , A study by Angelis et al. examined differences in eight European countries regarding the assessment of the value of new medicines in the context of reimbursement decisions. Most countries implement a type of economic evaluation in addition to the assessment of clinical benefit. The preferred health gain measure usually is the QALY. Additional values beyond the QALY concept are captured to a different extent between countries, explaining some of the heterogeneity in reimbursement decisions. Ease of use, nature of the treatment, public health benefit, social productivity, place in therapeutic strategy and ethical considerations are criteria considered in some countries (either implicitly or explicitly) and not considered in others.
Comparison to existing value frameworks Different frameworks exist for evaluating the value of health interventions. These frameworks can be generic or oncology specific and can inform reimbursement or patient‐level decisions. The added value of our findings to these existing frameworks is discussed below. In the United States, the Institute for Clinical and Economic Review (ICER) has been evaluating the clinical and economic values of health interventions to aid reimbursement decisions. Their value framework incorporates long‐term value and short‐term affordability. Long‐term value includes incremental cost‐effectiveness and provides the possibility of incorporating additional other benefits or disadvantages and contextual considerations through deliberative processes. Short‐term affordability includes budget impact. In Europe, the EUnetHTA published an HTA core model to aid reimbursement decisions for which different European countries collaborated. The ontology in this model covers the health problem and current use of technology, description and technical characteristics, safety, clinical effectiveness, costs and economic evaluation, ethical analysis, organizational aspect, patient and social aspects and legal aspects. The ICER value framework and the HTA core model are extensive frameworks used for the systematic assessment of new treatments for reimbursement decisions, and our study can add to these frameworks by more explicitly using the values that are specific for oncological care in the deliberative processes. Besides these models, two common oncological value frameworks are the American Society of Clinical Oncology (ASCO) and the European Society for Medical Oncology (ESMO) frameworks. These are used for facilitating shared decision‐making by patients and oncologists and could aid reimbursement decisions. The ASCO and ESMO incorporate clinical benefit, toxicity, QoL and improvement of (cancer‐related) symptoms beyond QoL. In addition, the ASCO adds treatment‐free interval, drug acquisition costs and patient co‐pay, and the ESMO adds daily well‐being, response rate and duration of response. The value frameworks are used to aid physicians in explaining treatment benefits to patients and facilitate shared decision‐making. Our study adds to these cancer‐specific value frameworks by presenting explicit values regarding the impact on daily life and future, costs for patients and loved ones, QoL and quality of treatment. In addition, the impact on loved ones and societal impact are (mostly) lacking in these frameworks.
Strengths and limitations The main strengths of this study are that we included a wide array of various stakeholders in the interviews and that multiple interviews were held per category, thus ensuring the capture of all the relevant perspectives. To our knowledge, this is the first study to explore a wider set of values in oncology. We acknowledge some limitations. First, possibilities for incorporation of the values found in current decision‐making are only explored qualitatively. Case studies on oncological treatments will provide further insights into the true extent of how these values are included in decision‐making procedures. Second, our study did not include patients from all tumour types. Different patients with different tumour types could reveal additional values mainly for the physical and psychological quality of life. Third, as this study is performed in the Netherlands the results are mainly generalizable to different countries with a comparable healthcare setting (see Box for a description). Finally, the aimal of the research was to explore value–elements for the assessment of new treatments. The themes represent those values, however, the subcategories within the main themes went beyond the scope of new treatment assessment based on the explorative and semi‐structured design of our interviews. Our approach ensures no mentioned values were missed and the focus group was used to assess the usefulness of this broad exploration of values. Despite these limitations, an overall idea of the broad range of values and their inclusion in decision‐making can be generated.
Conclusion and implications In conclusion, clinical values are not the only ones that matter to oncological patients and involved stakeholders regarding the evaluation of new treatments and decision‐making procedures in oncological care. We found a much broader range of values. The recognition and appreciation of those values might add to patient‐level decision‐making, but the usefulness for reimbursement decisions is less clear. The values add to existing value frameworks used for both patient‐level and reimbursement decisions. In addition, the values might improve patient information, empowerment and support. The results could be useful to guide clinicians and policymakers when it comes to decision‐making in oncology. We recommend exploring a more structural and explicit incorporation of values within oncology in patient‐level and reimbursement decision processes. At the patient level, the list of values can inform clinicians on which values to address in SDM, can be used for decision aids and can be used to provide extended patient information. For reimbursement decisions, it would be interesting to explore how the identified values can contribute. More research is needed on making explicit, for different oncological indications, if and how disease‐specific values can be systematically inventoried and incorporated to guarantee a more systematic approach to decision‐making and the deliberative processes.
Cilla E.J. Vrinzen: Conceptualization (equal); data curation (lead); formal analysis (lead); investigation (lead); methodology (equal); project administration (equal); resources (equal); validation (equal); writing – original draft (lead). Haiko J. Bloemendal: Conceptualization (equal); methodology (equal); supervision (equal); validation (equal); writing – review and editing (equal). Esra Stuart: Formal analysis (equal); writing – review and editing (equal). Amr Makady: Conceptualization (equal); methodology (equal); writing – review and editing (equal). Michel van Agthoven: Conceptualization (equal); methodology (equal); writing – review and editing (equal). Mariska Koster: Conceptualization (equal); methodology (equal); writing – review and editing (equal). Matthias AW Merkx: Conceptualization (equal); methodology (equal); supervision (equal); writing – review and editing (equal). Rosella P.M.G. Hermens: Conceptualization (equal); investigation (equal); methodology (equal); supervision (equal); validation (equal); writing – review and editing (equal). Patrick P.T. Jeurissen: Conceptualization (equal); funding acquisition (equal); methodology (equal); supervision (equal); validation (equal); writing – review and editing (equal).
This study is partly funded by Janssen‐Cilag BV, Breda, the Netherlands, pharmaceutical companies of Johnson & Johnson
The study was partly funded by Janssen‐Cilag B.V., Breda, the Netherlands. CV, PJ, ES, MM, HB and RH have no conflict of interest to declare. AM, MK and MA are employees of Janssen‐Cilag B.V. and MA is a stock shareholder.
Ethical approval for non‐intrusive interview studies is not mandatory by Dutch law.
Data S1 Click here for additional data file.
|
Early Termination of Oncology Clinical Trials in the United States | 6f287b46-f9cc-47a5-8d30-4d3c86e9e54b | 10028157 | Internal Medicine[mh] | INTRODUCTION Interventional clinical trials play a vital role in advancing new therapeutic approaches in medicine. Prior work has shown that a significant portion of clinical trials are terminated early. , Early termination can happen due to reasons that include, but are not limited to, poor accrual, funding issues, and emerging safety and/or efficacy signals. In the setting of safety or efficacy considerations, early termination may be appropriate, and this approach is commonly prespecified in the clinical trial statistical plan. In the case of early termination due to poor accrual, funding issues, or other logistical issues, early termination can result in utilization of resources without contributing knowledge to the scientific community. In these cases, there are ethical considerations for patients who enroll in clinical trials that do not reach a conclusion due to early termination. Despite the large and growing number of oncology clinical trials, there is limited research on early termination of cancer clinical trials. One analysis demonstrated that cancer trials have a higher likelihood of early termination when compared to trials in other disciplines. One preliminary report identified that poor accrual was the main reason for early trial termination in oncology, though the COVID‐19 pandemic was reported as a new additional reason. Another study showed that approximately one in six urologic oncology trials were terminated prematurely, with one in 10 closing early due to poor accrual. In this context, we conducted an analysis of publicly available data from ClinicalTrials.gov with the following aims. First, we sought to define the rate of early trial discontinuation of oncology trials as well as reasons for early termination. Second, we assessed potential trends in the rates of oncology trial termination. Third, we performed a comprehensive analysis of predictors of early termination, with a goal of informing efforts to improve the efficiency of the oncology clinical trials enterprise.
METHODS 2.1 Data source We conducted a cross‐sectional study of interventional clinical trials for the treatment of patients with cancer and registered in ClinicalTrials.gov with at least one site listed in the United States. Trials that included conditions other than cancer, including cancer screening or prevention studies in people without cancer, were excluded. We limited our analysis by searching for cancer interventional studies registered from September 27, 2007 to June 30, 2015. This start date aligns with start of required registration of trials in ClinicalTrials.gov . This end date was chosen to allow sufficient time for trial completion/discontinuation status to be determined and reported in ClinicalTrials.gov . All trial statuses were included except those coded as “not yet recruiting” and “withdrawn” since these trials were registered but never accrued patients, as well as “unknown” given that the status was not verified within past 2 years. The ClinicalTrials.gov query was performed on a single day (September 1, 2021). This analysis of publicly available data did not involve human subjects and therefore institutional review board review was not required. 2.2 Variables Definitions for data elements in ClinicalTrials.gov were used as per the Glossary of Common Site Terms and Clinical Trials.gov Protocol Data Element Definitions. , Trial entries in ClinicalTrials.gov provide details on the study population, condition, intervention type, start and completion dates, funding source, design characteristics, trial site(s), sex, age, phase, and current recruitment status. Investigators must periodically update these records. Certain data elements were further categorized for the purposes of analysis as described herein. The condition under study was categorized as “solid tumor,” “hematological malignancy,” “CNS tumor,” or “multiple” if cancer type was not specified or if cancer spanned more than one of the previously mentioned cancer categories. As examples, if the eligible cancer types (e.g., breast and colon), all fell under only one of the aforementioned categories (e.g., solid tumor), then it was categorized as such; however, if the eligible cancer types included more than one of the categories (e.g., lung and glioma) or was histology‐agnostic (e.g., any cancer type with specific genomic feature regardless of primary site), then it was categorized as “multiple.” Interventions were categorized as “drug/biologic,” “radiation,” “procedure,” “behavior,” “device,” “bone marrow/stem cell transplant or other cell therapy,” “other,” and “multiple” if spanning multiple categories. The focus of the intervention category was the main experimental intervention being studied, oftentimes this would be compared to the current standard of care. Interventions were further categorized as “anticancer” if treating the cancer itself, or “supportive” if treating symptoms of cancer or side effects of cancer therapy. Study duration was calculated as the time between study start date and study completion date, the last visit where data was collected for any of the study outcomes. Trials were classified as “pediatric” if accepting only patients younger than 18, “adult” if only accepting those 18 and older, and “adult and pediatric” if it spanned 18 years old. If trials accepted only patients older than 60 years of age, it was also categorized as “geriatric.” Trial phases were categorized by the indicated phase, and if the trial spanned two phases (e.g., phase 1/2 trial), the trial was categorized as the lower phase. Reasons for trial discontinuation were tabulated based on data provided in ClinicalTrials.gov and categorized as “accrual,” “funding,” “clinical efficacy,” “lack of clinical efficacy,” “side effects,” and “other” if the reason did not fit into any previous classification or no reason was cited. 2.3 Statistical considerations Descriptive statistics for rates of trial termination were calculated. Fisher exact tests were used to evaluate the association between trial termination and categorical trial characteristics, except for trial characteristics with more than four categories in which case chi‐squared tests were used. Two sample t ‐tests were used to compare continuous variables between terminated and non‐terminated trials. Two‐sided p ‐values are reported with p < 0.05 and were considered statistically significant. Logistic regression models were constructed to determine odds ratios for trial termination as a function of multiple trial characteristics, starting initially with all statistically significant variables from univariate testing and then removing variables that were no longer significant on multivariate testing to arrive at a final model containing only statistically significant variables. All statistical analyses were performed in STATA BE version 17 (StataCorp).
Data source We conducted a cross‐sectional study of interventional clinical trials for the treatment of patients with cancer and registered in ClinicalTrials.gov with at least one site listed in the United States. Trials that included conditions other than cancer, including cancer screening or prevention studies in people without cancer, were excluded. We limited our analysis by searching for cancer interventional studies registered from September 27, 2007 to June 30, 2015. This start date aligns with start of required registration of trials in ClinicalTrials.gov . This end date was chosen to allow sufficient time for trial completion/discontinuation status to be determined and reported in ClinicalTrials.gov . All trial statuses were included except those coded as “not yet recruiting” and “withdrawn” since these trials were registered but never accrued patients, as well as “unknown” given that the status was not verified within past 2 years. The ClinicalTrials.gov query was performed on a single day (September 1, 2021). This analysis of publicly available data did not involve human subjects and therefore institutional review board review was not required.
Variables Definitions for data elements in ClinicalTrials.gov were used as per the Glossary of Common Site Terms and Clinical Trials.gov Protocol Data Element Definitions. , Trial entries in ClinicalTrials.gov provide details on the study population, condition, intervention type, start and completion dates, funding source, design characteristics, trial site(s), sex, age, phase, and current recruitment status. Investigators must periodically update these records. Certain data elements were further categorized for the purposes of analysis as described herein. The condition under study was categorized as “solid tumor,” “hematological malignancy,” “CNS tumor,” or “multiple” if cancer type was not specified or if cancer spanned more than one of the previously mentioned cancer categories. As examples, if the eligible cancer types (e.g., breast and colon), all fell under only one of the aforementioned categories (e.g., solid tumor), then it was categorized as such; however, if the eligible cancer types included more than one of the categories (e.g., lung and glioma) or was histology‐agnostic (e.g., any cancer type with specific genomic feature regardless of primary site), then it was categorized as “multiple.” Interventions were categorized as “drug/biologic,” “radiation,” “procedure,” “behavior,” “device,” “bone marrow/stem cell transplant or other cell therapy,” “other,” and “multiple” if spanning multiple categories. The focus of the intervention category was the main experimental intervention being studied, oftentimes this would be compared to the current standard of care. Interventions were further categorized as “anticancer” if treating the cancer itself, or “supportive” if treating symptoms of cancer or side effects of cancer therapy. Study duration was calculated as the time between study start date and study completion date, the last visit where data was collected for any of the study outcomes. Trials were classified as “pediatric” if accepting only patients younger than 18, “adult” if only accepting those 18 and older, and “adult and pediatric” if it spanned 18 years old. If trials accepted only patients older than 60 years of age, it was also categorized as “geriatric.” Trial phases were categorized by the indicated phase, and if the trial spanned two phases (e.g., phase 1/2 trial), the trial was categorized as the lower phase. Reasons for trial discontinuation were tabulated based on data provided in ClinicalTrials.gov and categorized as “accrual,” “funding,” “clinical efficacy,” “lack of clinical efficacy,” “side effects,” and “other” if the reason did not fit into any previous classification or no reason was cited.
Statistical considerations Descriptive statistics for rates of trial termination were calculated. Fisher exact tests were used to evaluate the association between trial termination and categorical trial characteristics, except for trial characteristics with more than four categories in which case chi‐squared tests were used. Two sample t ‐tests were used to compare continuous variables between terminated and non‐terminated trials. Two‐sided p ‐values are reported with p < 0.05 and were considered statistically significant. Logistic regression models were constructed to determine odds ratios for trial termination as a function of multiple trial characteristics, starting initially with all statistically significant variables from univariate testing and then removing variables that were no longer significant on multivariate testing to arrive at a final model containing only statistically significant variables. All statistical analyses were performed in STATA BE version 17 (StataCorp).
RESULTS 3.1 Trial search outcome At the time of data collection, the ClinicalTrials.gov search yielded 9497 potential trials (Figure ). Of these, 8.53% ( n = 810) trials were excluded in which cancer was not the condition studied, trial was not exclusively cancer focused, not all patients had cancer, if there was no intervention, or if the status was “withdrawn.” The analytic cohort therefore included 8687 interventional cancer clinical trials. 3.2 Rates, reasons, and trends of early trial discontinuation Of 8687 trials, 77.3% ( n = 6712) were non‐terminated trials and 22.7% ( n = 1975) were terminated trials (Figure ). The most common reason for early termination was “Other, Multiple Reasons, or Unknown” in 36.9%, followed by accrual issues in 34.5% (Figure ). Only 1.7% of trials terminated early due to early evidence of efficacy, 7.4% terminated early due to early evidence of lack of efficacy, and 4.6% terminated early due to toxicity concern. The rate of early trial termination over time appeared stable over the course of the study period from 2007 to 2015 (Figure ). As expected, trials that were terminated early had lower mean enrollment (mean 124 vs. 48 patients; p < 0.001) and shorter duration (mean 3.3 vs. 5.7 years; p < 0.001). 3.3 Univariate predictors of trial discontinuation Table shows univariate predictors of trial discontinuation. Statistically significant predictors of early termination include cancer category ( p = 0.011), phase ( p < 0.001), funding source ( p < 0.001), location ( p < 0.001), and age ( p = 0.016). Trials studying multiple cancer categories (e.g., patients with solid tumors and hematologic malignancies were eligible) had the lowest rate of early termination (17.1%) compared to trials that included patients just from one cancer category. Phase 3 trials had the lowest rate of early termination (18.5%) compared to other phases. Industry funded trials had the lowest rate of early termination (20.1%) compared to other funding categories. Trials located only in the United States were more likely to be terminated early (24.4%) compared to trials located both in the United States and internationally (16.8%). Trials that included patients <18 years of age had lower rates of early termination compared to trials that included only adults. Since early termination due to poor accrual was the single most common discrete reason for early termination, we also evaluated predictors of early termination due to poor accrual among all terminated cancer clinical trials (Table ). On univariate analysis, we found statistically significant predictors of trial discontinuation specifically due to poor accrual to include therapeutic intent (anticancer vs. supportive care; p < 0.001), intervention type ( p < 0.001), randomization status ( p = 0.012), age ( p = 0.024), sex ( p = 0.041), phase ( p < 0.001), funding ( p < 0.001), and location ( p < 0.001). Of all trials terminated early, 33.6% of anticancer trials were terminated early due to accrual issues compared to 52.4% of supportive care trials. Radiotherapy trials had the highest rate of early termination due to accrual (53.7%) compared to other intervention types. Randomized trials had higher rates of termination due to accrual issues (35.6%) compared to non‐randomized trials (27.5%). Trials that included adults had lower rates of early termination due to accrual (33.8%) compared to trials that allowed patients <18 years. Meanwhile, trials that only included male participants had the highest rate of early termination due to accrual (43.0%) compared to terminated trials that only included females (39.8%) or those that included both males and females (33.5%). Reviewing phases, phase 4 (47.8%) and phase 2 (41.0%) trials had the highest rates of early termination due to accrual issues. Among terminated trials, industry funded trials had the lowest rate of termination due to accrual (15.7%) compared to other funding categories. Considering geography, trials located only in the US had higher rates of termination due to accrual (37.6%) compared to those located in both the US and internationally (18.6%). 3.4 Multivariable predictors of trial discontinuation In multivariable analysis, cancer category, phase, funding, and location were significant independent predictors of trial termination for any reason (Table ). The likelihood of early clinical trial termination was significantly lower for those investigating multiple cancer types than those investigating hematological cancers (odds ratio [OR] 0.70, 95% confidence interval [Cl] 0.54–0.91, p = 0.007). Phase 2 clinical trials were more likely to terminate early when compared to phase 1 (OR 1.27, 95% CI 1.14–1.41, p < 0.001). Trials funded by other sources were more likely to be terminated early than those funded by industry (OR 1.19, 95% CI 1.02–1.38, p = 0.025). Trials were less likely to be terminated if they were open internationally rather than open in the US alone (OR 0.65, 95% CI 0.55–0.76, p < 0.01). Among all terminated cancer clinical trials, we also performed multivariate analysis to identify predictors of trial discontinuation specifically due to poor accrual (Table ). Supportive care trials were more likely to be terminated due to poor accrual than anticancer trials (OR 1.71, 95% CI 1.13–2.60, p = 0.011). Phase 2 trials were more likely to be terminated due to accrual issues compared to phase 1 (OR 1.58, 95% CI 1.29–1.94, p < 0.001). Trials funded by sources other than industry were more likely to be terminated due to accrual issues than those funded by industry (ORs >2.6 and p < 0.001 for all other funding sources).
Trial search outcome At the time of data collection, the ClinicalTrials.gov search yielded 9497 potential trials (Figure ). Of these, 8.53% ( n = 810) trials were excluded in which cancer was not the condition studied, trial was not exclusively cancer focused, not all patients had cancer, if there was no intervention, or if the status was “withdrawn.” The analytic cohort therefore included 8687 interventional cancer clinical trials.
Rates, reasons, and trends of early trial discontinuation Of 8687 trials, 77.3% ( n = 6712) were non‐terminated trials and 22.7% ( n = 1975) were terminated trials (Figure ). The most common reason for early termination was “Other, Multiple Reasons, or Unknown” in 36.9%, followed by accrual issues in 34.5% (Figure ). Only 1.7% of trials terminated early due to early evidence of efficacy, 7.4% terminated early due to early evidence of lack of efficacy, and 4.6% terminated early due to toxicity concern. The rate of early trial termination over time appeared stable over the course of the study period from 2007 to 2015 (Figure ). As expected, trials that were terminated early had lower mean enrollment (mean 124 vs. 48 patients; p < 0.001) and shorter duration (mean 3.3 vs. 5.7 years; p < 0.001).
Univariate predictors of trial discontinuation Table shows univariate predictors of trial discontinuation. Statistically significant predictors of early termination include cancer category ( p = 0.011), phase ( p < 0.001), funding source ( p < 0.001), location ( p < 0.001), and age ( p = 0.016). Trials studying multiple cancer categories (e.g., patients with solid tumors and hematologic malignancies were eligible) had the lowest rate of early termination (17.1%) compared to trials that included patients just from one cancer category. Phase 3 trials had the lowest rate of early termination (18.5%) compared to other phases. Industry funded trials had the lowest rate of early termination (20.1%) compared to other funding categories. Trials located only in the United States were more likely to be terminated early (24.4%) compared to trials located both in the United States and internationally (16.8%). Trials that included patients <18 years of age had lower rates of early termination compared to trials that included only adults. Since early termination due to poor accrual was the single most common discrete reason for early termination, we also evaluated predictors of early termination due to poor accrual among all terminated cancer clinical trials (Table ). On univariate analysis, we found statistically significant predictors of trial discontinuation specifically due to poor accrual to include therapeutic intent (anticancer vs. supportive care; p < 0.001), intervention type ( p < 0.001), randomization status ( p = 0.012), age ( p = 0.024), sex ( p = 0.041), phase ( p < 0.001), funding ( p < 0.001), and location ( p < 0.001). Of all trials terminated early, 33.6% of anticancer trials were terminated early due to accrual issues compared to 52.4% of supportive care trials. Radiotherapy trials had the highest rate of early termination due to accrual (53.7%) compared to other intervention types. Randomized trials had higher rates of termination due to accrual issues (35.6%) compared to non‐randomized trials (27.5%). Trials that included adults had lower rates of early termination due to accrual (33.8%) compared to trials that allowed patients <18 years. Meanwhile, trials that only included male participants had the highest rate of early termination due to accrual (43.0%) compared to terminated trials that only included females (39.8%) or those that included both males and females (33.5%). Reviewing phases, phase 4 (47.8%) and phase 2 (41.0%) trials had the highest rates of early termination due to accrual issues. Among terminated trials, industry funded trials had the lowest rate of termination due to accrual (15.7%) compared to other funding categories. Considering geography, trials located only in the US had higher rates of termination due to accrual (37.6%) compared to those located in both the US and internationally (18.6%).
Multivariable predictors of trial discontinuation In multivariable analysis, cancer category, phase, funding, and location were significant independent predictors of trial termination for any reason (Table ). The likelihood of early clinical trial termination was significantly lower for those investigating multiple cancer types than those investigating hematological cancers (odds ratio [OR] 0.70, 95% confidence interval [Cl] 0.54–0.91, p = 0.007). Phase 2 clinical trials were more likely to terminate early when compared to phase 1 (OR 1.27, 95% CI 1.14–1.41, p < 0.001). Trials funded by other sources were more likely to be terminated early than those funded by industry (OR 1.19, 95% CI 1.02–1.38, p = 0.025). Trials were less likely to be terminated if they were open internationally rather than open in the US alone (OR 0.65, 95% CI 0.55–0.76, p < 0.01). Among all terminated cancer clinical trials, we also performed multivariate analysis to identify predictors of trial discontinuation specifically due to poor accrual (Table ). Supportive care trials were more likely to be terminated due to poor accrual than anticancer trials (OR 1.71, 95% CI 1.13–2.60, p = 0.011). Phase 2 trials were more likely to be terminated due to accrual issues compared to phase 1 (OR 1.58, 95% CI 1.29–1.94, p < 0.001). Trials funded by sources other than industry were more likely to be terminated due to accrual issues than those funded by industry (ORs >2.6 and p < 0.001 for all other funding sources).
DISCUSSION Our study demonstrates that among cancer interventional trials, early trial discontinuation is common. Over the course of study period from 2007 to 2015, the rate of early trial termination remained stable, with overall 22.7% of oncology trials terminating early. Statistically significant predictors of early termination include cancer category, phase, funding source, location, and age. In multivariable analysis, cancer category, phase, funding, and location were significant independent predictors of trial termination for any reason. Early termination due to toxicity, early efficacy, or early lack of efficacy accounted for the minority of trials that were terminated early. In contrast, poor accrual was a common reason for early termination, accounting for 34.5% of terminated trials. Statistically significant predictors of trial discontinuation specifically due to poor accrual includes therapeutic intent, intervention type, age, gender, phase, funding, and location. In multivariate analysis, therapeutic intent, phase, and funding were significant independent predictors of trial discontinuation specifically due to poor accrual. Our early termination rate of 22.7% for oncology trials is higher than those in other disciplines. Reviewing cardiovascular clinical trials, 10.9% were terminated prematurely. Meanwhile, pregnancy related trials found that 6.1% of trials were terminated. Orthopedic trials had an early termination rate of 7.7% for shoulder‐related, 12.7% elbow‐related, and 14.0% spine‐related. The only other discipline with termination rates comparable to those of oncology trials is pediatrics where 19% of trials have been reported to discontinue early. Key independent trial characteristics associated with higher rates of early termination in oncology included phase 2 trials, non‐industry funded, and US‐based only. In contrast, histology agnostic trials that spanned disease groups had lower rates of early termination. These findings are consistent with those from other disciplines. One study on cardiovascular clinical trials found that trial termination predictors included location, intervention type, phase, therapeutic intent, and year of initiation. In obstetrics, independent predictors of trial termination included number of study locations, available results, study type, randomized design, study purpose, intervention type (drug or nondrug), and study location (including locations outside USA). In orthopedics, industry sponsored trials, phase 2 trials, blinded trials, and device trials appear to be associated with higher rates of early termination. , In contrast, an evaluation of pediatric trials showed that industry funding was associated with lower rates of early termination. While differing variables serve as predictors for clinical trials termination across disciplines, the factors identified herein should be considered as risk factors for early termination during early development of oncology trials. The most common single reason for cancer clinical trial early termination was poor accrual, which is also a common finding found in other disciplines. A 2015 cross‐sectional study of terminated clinical trials on ClinicalTrials.gov found that 39% of all terminated trials were due to accrual. A 2014 study found that cardiovascular clinical trials were more likely to be terminated due to poor accrual (53.6%), which was also supported by a 2017 study (41%). The 2014 study on cardiovascular clinical trials found that mixed‐source founding and university/hospital funding were independently associated with a higher risk of study termination due to low recruitment; meanwhile, NIH/US federal funding, behavior/diet intervention, and single‐arm design were factors independently associated with lower risk for early termination due to low recruitment. Studies of obstetrics and orthopedic trials likewise show that accrual difficulty is one of the most frequently cited reason for early termination. , Taken together with the available literature, our findings suggest that study teams need to develop more realistic accrual goals and robust recruitment strategies. In oncology, this appears to be particularly critical for supportive care trials, trials funded by sources other than industry, and phase 2 and phase 4 trials. Of note, phase 4 trials are often post‐marketing studies and therefore patients may be able to access the therapy under investigation through commercial mechanisms. These trials had the highest odds ratio for early termination due to poor accrual, which further emphasizes that recruiting patients can be difficult given standards of care that can already be in place or rapidly evolving. Several limitations should be noted when interpreting our findings. Given that this study analyzed only trials in ClinicalTrials.gov , it must be considered that there might be additional cancer clinical trials not captured in our analysis, especially phase I clinical trials. The rate of nonregistration of cancer clinical trials is not known, but it is unlikely that these trials have higher rates of completion given federal and editorial policies mandating registration. Another limitation is that our cancer categories are quite broad, in the context of this pan‐cancer analysis. Our goal in categorizing cancer types into these board categories was to investigate possible patterns that might be associated with higher rates of early termination. In addition, our analysis depends on the accuracy of trial data provided to ClinicalTrials.gov by investigators and sponsors. This issue is mitigated in part by automated data validity checks and manual review of ClinicalTrials.gov to ensure data accuracy before public posting. Concurrently, there were missing data in the registry such as trial phases and reasons for discontinuation, which is mitigated by the robust size of the data set. Ultimately, we have found that cancer clinical trials are frequently discontinued early. As most cases of early termination are due to accrual and other operational issues, our study highlights the considerable inefficiency and waste associated with early termination of oncology trials. Our work highlights trial characteristics that merit a focused effort to support adequate accrual and study teams need to plan for drug supply, adequate funding, and other operational issues to avoid early termination for reasons other than for safety and/or efficacy reasons. Prior planning to reduce likelihood of early termination is of paramount importance as there are important ethical concerns associated with patients participating in trials that will not contribute to greater scientific knowledge. Monitoring this issue and routinely recording reason for trial termination will allow our findings to be revisited in future studies. Future opportunities for research include reviewing patterns of early termination specifically in trials for patients with common histologies, examining long‐term impacts of the COVID‐19 pandemic on oncology trial early termination, and exploring termination rates internationally. Ultimately, although there have been policies and interventions implemented to both increase the number of cancer clinical trials and improve trial reporting, there needs to be further consideration and action to ensure that patient participation in cancer trials advances the field.
Ellen Zhang: Conceptualization (equal); data curation (equal); formal analysis (equal); investigation (equal); methodology (equal); supervision (equal); validation (equal); writing – original draft (equal); writing – review and editing (equal). Steven G. DuBois: Conceptualization (equal); data curation (equal); formal analysis (equal); funding acquisition (lead); investigation (equal); methodology (equal); project administration (lead); resources (equal); software (equal); supervision (lead); validation (equal); visualization (equal); writing – original draft (equal); writing – review and editing (equal).
Supported by Alex's Lemonade Stand Foundation Center of Excellence award (SGD).
SGD reports consulting fees from Amgen, Bayer, Jazz, and Loxo, and travel expenses from Loxo Oncology, Roche, and Salarius.
|
Study of nonsynchronous online teaching of regional anatomy for international students integrated with medical humanities and local culture during COVID-19 pandemic | a462486b-c015-4cb5-baf8-e0b76008e993 | 10028764 | Anatomy[mh] | Regional anatomy is a core course of basic medicine and a bridge discipline between basic medicine and clinical medicine. Regional anatomy is taught to the first- and second-year medical students through didactic lectures and corpse dissection. The COVID-19 pandemic imposed unexpected disruptions to anatomical educational practice, where face-to-face (F2F) teaching in universities was not possible during the pandemic and anatomy education was shifted to the online teaching mode . E-learning platforms are online teaching tools used successfully in teaching practices in medical education that use multimedia technologies and the internet to improve the quality of learning by facilitating access to resources and remote communication . Because most international students were stuck in their own country, the teaching method for international students was online using e-learning platforms. However, online teaching of regional anatomy using e-learning platforms encountered various challenges. First, practical dissection is important for medical students to learn regional anatomy , but online teaching of anatomy cannot provide international students with experimental teaching for practical dissection. Teachers should try their best to make online education homogeneous equivalent to offline education. Second, students were unable to return to China due to the COVID-19 pandemic. Most of them have fewer chances to learn about and experience local culture. Teachers must introduce Chinese culture while teaching professional subjects. International students’ understanding of Chinese culture helps them to integrate and explore culture from different perspectives and improves their interest in learning complementary and alternative medicine (CAM), including Traditional Chinese Medicine. In recent years, because of the change in medical models and in menu of human diseases, CAM is getting popular in Western countries and is gradually accepted by mainstream medicine. Third, medical students will become a doctor in the near future. It is important to make students understand more about the associated responsibilities and medical humanities . To address the aforementioned issues, we designed and carried out the nonsynchronous online teaching of regional anatomy for international students, meanwhile, we introduced medical humanities and ethics and Chinese culture. After that, we used the questionnaire to investigate the teaching effect. The details of the survey are discussed in this article.
The research object of this study was the international students of 2019 and 2020 majoring in clinical medicine at Guangzhou Medical University (Xinzao, Panyu District, Guangzhou, China), and international students studying offline in 2017 and 2018 as controls. Due to the time difference between the location of international students and China, and the different Internet conditions, the teaching staff chose to use nonsynchronous online teaching of regional anatomy. The nonsynchronous online teaching of regional anatomy for international students includes four parts: theoretical teaching, experimental teaching, medical humanistic education, Chinese cultural context, and after-class activities. Preparation of network platform and teaching materials To begin with, we used the semi-structured questionnaire to investigate the characteristics of the participants before the regional anatomy online teaching. Using the online teaching platform (the network platform of the Super Star), we uploaded electronic textbooks, slides for teaching, videos for theoretical and experimental teaching, etc. International students could download these materials from the platform. Before each chapter, teachers released the course schedule and key points. Students checked in through the learning link of the network platform when class began. Theoretical teaching of regional anatomy for international students In regional anatomy, we introduced layers and relationships of different regions, such as the neck, upper limb, lower limb, thorax, abdomen, perineum, and back. Teachers made videos of theory chapters for online teaching. To make the videos easier to understand, each chapter was divided into 10 segments and about 9 min per segment. At the end of each video, a quiz was attached to help students revise what they had been taught. During the process, a clinical application was taught related closely to the purpose of practical use. For example, when we talked about the anatomical structures of the ventral anterolateral wall, we referred to the layers to reach the appendix from superficial to deep when doing the surgery of appendectomy. By introducing the clinical practice into the regional anatomy, students can be engaged in the study of the subject. In addition, we introduced Chinese culture while teaching the regional anatomy. For example, when explaining the anatomy of the diaphragm, we compared it to the postures of Chinese Taijiquan, which combines human physiology, anatomy, gymnastics, and Chinese traditional culture including martial arts, aiming at strengthening the body, fighting, confrontation, and other functions. Taijiquan not only vividly made students understand the relationship between diaphragm contraction and breathing, but also let them know more about Chinese culture. For another example, when teaching the anatomy of the lung, we introduced how COVID-19 infected the lung. Chinese medical staff took many measures to contain the spread of COVID-19, including the invention of a vaccine. In this way, we shortened the distance between international students and China and continue their professional development, although they were stuck in their own countries during the pandemic period. Experimental teaching of regional anatomy for international students Experimental teaching was a very important part of the regional anatomy course. Before COVID-19, students were required to dissect the body and observe anatomical structures. However, during COVID-19, most students could not return to China and could not attend in person and dissect the body by themselves. To make students understand the dissection procedure better, we selected and interpreted videos of dissections. Each video was about 8–10 min and highly consistent with the contents in theoretical teaching. At the beginning of each video, key points were focused on so that students were clear about the content. For example, when teaching the branches of the common hepatic artery, teachers explained how to accurately identify the gallbladder artery and avoid the wrong ligation. When talking about the anatomical structure of the thyroid, we emphasized how to avoid injuring the superior laryngeal nerve and recurrent laryngeal nerve when ligation of the superior and inferior thyroid arteries. At the end of each chapter, several comprehensive topics related to clinical cases were put forward for students to discuss in groups. The students were required to discuss, summarize and present their findings in a group. Before the beginning of the course, international students were divided into groups and designated or recommended as group leaders. After the beginning of the course, international students were required to ask questions in the discussion area within the specified time. The group leader collected the comments of the students in the group in the discussion area, summarizes them and share common opinions of own group. Finally, the teacher explained in the discussion area. Through the group discussion, students learned how to work in a team and as a team. Ceremony to show gratitude to donors When studying regional anatomy, students need to dissect corpses and observe the formation and relationships of anatomical structures. Corpse gave their bodies to educate medical students. The donors did not say anything in class, but they teach the student a lot. We call them “silent teachers”. At the beginning of offline local anatomy teaching, teacher and student held a silent ceremony to express their gratitude to “silent teachers”. Guangzhou Medical University had built a Thanksgiving square in memory of these corpse donors. Because international students could not attend the ceremony and visit the square during COVID-19 pandemic, we made a video to show the ceremony and square. Also, we explained how the concept of Chinese people’s donating remains changed in recent years. In Chinese traditional culture, it is believed that “The body is given by parents and children should treasure it, which is filial piety to their parents”. Few people were willing to donate their bodies after death in the preceding decades. However, “silent teachers” donated their bodies to cultivate generations of medical students. Inspired by their deeds and spirits, more and more people have donated their bodies to medical universities. We are very grateful for their contributions. In this way, international students can feel the humanities and ethics, although they could not attend the ceremony during the COVID-19 period. After-class activities We held a series of medical specimen exhibitions and science popularization lectures for citizens. Students at Guangzhou Medical University always participated as commentators. Also, students were involved in the maintenance of specimens in the exhibition hall. In addition, we also held a drawing competition of anatomical structures. Through the competition, medical students become more familiar with the anatomical structure. Considering that international students could only receive online teaching, we made a short video to introduce these activities. Furthermore, international students were invited to attend the drawing competition of anatomical structures. Through these activities, international students got a deeper understanding of the campus culture of China Medical University while enjoying the beauty of anatomy. Statistical analysis All data were summarized as mean ± SEM. To compare multiple groups on a continuous response variable, we used one-way ANOVA followed by Bonferroni post assay to compare the selected two groups. Normality and homoscedasticity were tested to ensure that parametric testing was appropriate. P ≤ 0.05 was considered statistically significant. GraphPad Prism 5 software was used to carry out all statistical analysis.
To begin with, we used the semi-structured questionnaire to investigate the characteristics of the participants before the regional anatomy online teaching. Using the online teaching platform (the network platform of the Super Star), we uploaded electronic textbooks, slides for teaching, videos for theoretical and experimental teaching, etc. International students could download these materials from the platform. Before each chapter, teachers released the course schedule and key points. Students checked in through the learning link of the network platform when class began.
In regional anatomy, we introduced layers and relationships of different regions, such as the neck, upper limb, lower limb, thorax, abdomen, perineum, and back. Teachers made videos of theory chapters for online teaching. To make the videos easier to understand, each chapter was divided into 10 segments and about 9 min per segment. At the end of each video, a quiz was attached to help students revise what they had been taught. During the process, a clinical application was taught related closely to the purpose of practical use. For example, when we talked about the anatomical structures of the ventral anterolateral wall, we referred to the layers to reach the appendix from superficial to deep when doing the surgery of appendectomy. By introducing the clinical practice into the regional anatomy, students can be engaged in the study of the subject. In addition, we introduced Chinese culture while teaching the regional anatomy. For example, when explaining the anatomy of the diaphragm, we compared it to the postures of Chinese Taijiquan, which combines human physiology, anatomy, gymnastics, and Chinese traditional culture including martial arts, aiming at strengthening the body, fighting, confrontation, and other functions. Taijiquan not only vividly made students understand the relationship between diaphragm contraction and breathing, but also let them know more about Chinese culture. For another example, when teaching the anatomy of the lung, we introduced how COVID-19 infected the lung. Chinese medical staff took many measures to contain the spread of COVID-19, including the invention of a vaccine. In this way, we shortened the distance between international students and China and continue their professional development, although they were stuck in their own countries during the pandemic period.
Experimental teaching was a very important part of the regional anatomy course. Before COVID-19, students were required to dissect the body and observe anatomical structures. However, during COVID-19, most students could not return to China and could not attend in person and dissect the body by themselves. To make students understand the dissection procedure better, we selected and interpreted videos of dissections. Each video was about 8–10 min and highly consistent with the contents in theoretical teaching. At the beginning of each video, key points were focused on so that students were clear about the content. For example, when teaching the branches of the common hepatic artery, teachers explained how to accurately identify the gallbladder artery and avoid the wrong ligation. When talking about the anatomical structure of the thyroid, we emphasized how to avoid injuring the superior laryngeal nerve and recurrent laryngeal nerve when ligation of the superior and inferior thyroid arteries. At the end of each chapter, several comprehensive topics related to clinical cases were put forward for students to discuss in groups. The students were required to discuss, summarize and present their findings in a group. Before the beginning of the course, international students were divided into groups and designated or recommended as group leaders. After the beginning of the course, international students were required to ask questions in the discussion area within the specified time. The group leader collected the comments of the students in the group in the discussion area, summarizes them and share common opinions of own group. Finally, the teacher explained in the discussion area. Through the group discussion, students learned how to work in a team and as a team.
When studying regional anatomy, students need to dissect corpses and observe the formation and relationships of anatomical structures. Corpse gave their bodies to educate medical students. The donors did not say anything in class, but they teach the student a lot. We call them “silent teachers”. At the beginning of offline local anatomy teaching, teacher and student held a silent ceremony to express their gratitude to “silent teachers”. Guangzhou Medical University had built a Thanksgiving square in memory of these corpse donors. Because international students could not attend the ceremony and visit the square during COVID-19 pandemic, we made a video to show the ceremony and square. Also, we explained how the concept of Chinese people’s donating remains changed in recent years. In Chinese traditional culture, it is believed that “The body is given by parents and children should treasure it, which is filial piety to their parents”. Few people were willing to donate their bodies after death in the preceding decades. However, “silent teachers” donated their bodies to cultivate generations of medical students. Inspired by their deeds and spirits, more and more people have donated their bodies to medical universities. We are very grateful for their contributions. In this way, international students can feel the humanities and ethics, although they could not attend the ceremony during the COVID-19 period.
We held a series of medical specimen exhibitions and science popularization lectures for citizens. Students at Guangzhou Medical University always participated as commentators. Also, students were involved in the maintenance of specimens in the exhibition hall. In addition, we also held a drawing competition of anatomical structures. Through the competition, medical students become more familiar with the anatomical structure. Considering that international students could only receive online teaching, we made a short video to introduce these activities. Furthermore, international students were invited to attend the drawing competition of anatomical structures. Through these activities, international students got a deeper understanding of the campus culture of China Medical University while enjoying the beauty of anatomy.
All data were summarized as mean ± SEM. To compare multiple groups on a continuous response variable, we used one-way ANOVA followed by Bonferroni post assay to compare the selected two groups. Normality and homoscedasticity were tested to ensure that parametric testing was appropriate. P ≤ 0.05 was considered statistically significant. GraphPad Prism 5 software was used to carry out all statistical analysis.
Investigation of international students before the online teaching of regional anatomy Guangzhou Medical University enrolled 61 international students majoring in clinical medicine in 2019. Due to the pandemic, the enrollment scale in 2020 was reduced, and 28 international students majoring in clinical medicine were enrolled. A total of 89 international students participated in the questionnaire survey, including 54 males (60.7%) and 35 females (39.3%), ranging in age from 18 to 21, and 89 questionnaires were collected. During the pandemic, 59 international students (66.29%) were from India with a Time Difference with China (TDC) of 2.5 h; 14 students (15.7%) were from Bangladesh (TDC = 2 h); six students (6.74%) were from Tanzania (TDC = 5 h); seven students (7.87%) were from Pakistan (TDC = 3 h); one student was from Zambia (TDC = 6 h); one person was from Angela (TDC = 7 h); one student was from Ghana (TDC = 8 h), as shown in Table . Forty-four students (49.4%) reported that the local network was in good condition whereas 35 students (39.3%) reported that the local network was not working well with buffering and delays. Ten students (11.2%) reported that networks were not available and they used the mobile data which incurred greater costs. However, all international students ensured that they completed the online courses on time. Effect and feedback of online teaching of regional anatomy for international students Course achievement We used formative assessment to evaluate the effects of online teaching of regional anatomy. The formative assessment was composed of four parts. The first part is the completion of online videos, which is scored according to the progress of watching the course video, accounting for 30% of the total score. The second part is the evaluation of general performance (including three items): signature, online quiz, and participation in discussion. Those who have completed three items will get 100 points, those who have completed two items will get 80 points, those who have completed one item will get 60 points, and those who have not completed any items will not get any points. This part accounts for 10% of the total score. The third part is the score of the experiment examination. Using lecture examination scores to predict experimental performance was a feasible way to identify students who may have difficulty in learning practical dissection . In this exam, students were required to find 20 structures on the pictures specified, with 5 marks for each structure. This part accounted for 30% of the total score. The fourth part is the theoretical examination at the end of the semester, accounting for 30% of the total score (Fig. ). As Fig. A shows there is no significant difference between the score of the experimental examination of nonsynchronous online teaching of regional anatomy for international students in 2019 or 2020 and that of offline teaching in 2017 or 2018. Although the theoretical examination of nonsynchronous online teaching of regional anatomy for international students in 2020 decreased significantly, the total score did not decrease but was higher than that in 2019 and 2017 or 2018. The total scores in 2020 were: 19 students (21%) got 85 points or more, 70 students (79%) got 60–80 points, and 100% of students passed the exam (Fig. ). The above results showed that students have mastered the relevant knowledge structures and key points through the implementation of nonsynchronous online teaching of a regional anatomy course. Feedback from students We also used a questionnaire to investigate the responses from students. The statements in the questionnaire were: (1) Do you agree that introducing the clinical cases in teaching regional anatomy promoted the understanding of the learning content and enhanced the learning effect? (2) Do you agree those clinical cases retain your attention in class and promote you to pay more attention to your professional spirit? (3) Do you agree that the introduction of Chinese culture in the teaching of regional anatomy enhanced your knowledge more about China? (4) Do you agree that drawing pictures related to regional anatomy attracts you to learn regional anatomy? (5) Do you expect to come back to China earlier? (6) Please list the names of the Traditional Chinese Festival. (7) Please list names of Solar terms in the traditional Chinese calendar. (8) Please list examples of Chinese scientist contribution to world medical science. (9) Please write down measures for combating the COVID-19 pandemic in China. Seventy-three students (82.0%) agreed that introducing the clinical cases in teaching regional anatomy promoted the understanding of the learning content and enhanced the learning effect. Eighty students (89.9%) strongly agreed that those cases retained their attention in class and promoted them to pay more attention to professional spirit. Eighty-two students (92.1%) reported that the introduction of Chinese culture in the teaching of regional anatomy enhanced their cognition of China to a large extent. Eighty-nine students (100%) reported that the involvement in drawing pictures attracted them to learn regional anatomy. All international students hope that the COVID-19 pandemic can be controlled as soon as possible and that they could travel to China for offline learning earlier. Fifty international students (56.2%) were able to list 4–5 Chinese traditional festivals, 29 (32.6%) could list 2–3, and 10 (11.2%) knew less than or equal to one. Fifty-three international students (59.6%) listed 4–5 Chinese seasonal solar terms, 28 students (31.5%) listed 2–3, and eight students (0.09%) could list only one. Forty-two students (47.2%) write down 4–5 examples of contributions to medical progress in China. For example, Zhong Nanshan was known for the SARS virus; Chen Wei is a Chinese epidemiologist and virologist specializing in biodefense. Tu Youyou contributed a lot to the malaria treatment. Thirty-six (40.4%) could list 2–3, and 11 (12.4%) could list only one. Among these examples, 56 (62.9%) international students mentioned traditional Chinese medicine, 48 (53.9%) mentioned acupuncture and 75 (84.3%) mentioned China’s COVID-19 vaccine. For anti-pandemic measures, 60 (67.4%) international students could list 4–5, 20 (22.5%) could list 2–3, and 9 (10.1%) could list one or less. 100% of students hope to come to China to carry out in-person education courses. Taking together, adding medical humanities and local culture into this nonsynchronous online teaching of regional anatomy is appreciated by international students, and achieved good teaching results.
Guangzhou Medical University enrolled 61 international students majoring in clinical medicine in 2019. Due to the pandemic, the enrollment scale in 2020 was reduced, and 28 international students majoring in clinical medicine were enrolled. A total of 89 international students participated in the questionnaire survey, including 54 males (60.7%) and 35 females (39.3%), ranging in age from 18 to 21, and 89 questionnaires were collected. During the pandemic, 59 international students (66.29%) were from India with a Time Difference with China (TDC) of 2.5 h; 14 students (15.7%) were from Bangladesh (TDC = 2 h); six students (6.74%) were from Tanzania (TDC = 5 h); seven students (7.87%) were from Pakistan (TDC = 3 h); one student was from Zambia (TDC = 6 h); one person was from Angela (TDC = 7 h); one student was from Ghana (TDC = 8 h), as shown in Table . Forty-four students (49.4%) reported that the local network was in good condition whereas 35 students (39.3%) reported that the local network was not working well with buffering and delays. Ten students (11.2%) reported that networks were not available and they used the mobile data which incurred greater costs. However, all international students ensured that they completed the online courses on time.
Course achievement We used formative assessment to evaluate the effects of online teaching of regional anatomy. The formative assessment was composed of four parts. The first part is the completion of online videos, which is scored according to the progress of watching the course video, accounting for 30% of the total score. The second part is the evaluation of general performance (including three items): signature, online quiz, and participation in discussion. Those who have completed three items will get 100 points, those who have completed two items will get 80 points, those who have completed one item will get 60 points, and those who have not completed any items will not get any points. This part accounts for 10% of the total score. The third part is the score of the experiment examination. Using lecture examination scores to predict experimental performance was a feasible way to identify students who may have difficulty in learning practical dissection . In this exam, students were required to find 20 structures on the pictures specified, with 5 marks for each structure. This part accounted for 30% of the total score. The fourth part is the theoretical examination at the end of the semester, accounting for 30% of the total score (Fig. ). As Fig. A shows there is no significant difference between the score of the experimental examination of nonsynchronous online teaching of regional anatomy for international students in 2019 or 2020 and that of offline teaching in 2017 or 2018. Although the theoretical examination of nonsynchronous online teaching of regional anatomy for international students in 2020 decreased significantly, the total score did not decrease but was higher than that in 2019 and 2017 or 2018. The total scores in 2020 were: 19 students (21%) got 85 points or more, 70 students (79%) got 60–80 points, and 100% of students passed the exam (Fig. ). The above results showed that students have mastered the relevant knowledge structures and key points through the implementation of nonsynchronous online teaching of a regional anatomy course. Feedback from students We also used a questionnaire to investigate the responses from students. The statements in the questionnaire were: (1) Do you agree that introducing the clinical cases in teaching regional anatomy promoted the understanding of the learning content and enhanced the learning effect? (2) Do you agree those clinical cases retain your attention in class and promote you to pay more attention to your professional spirit? (3) Do you agree that the introduction of Chinese culture in the teaching of regional anatomy enhanced your knowledge more about China? (4) Do you agree that drawing pictures related to regional anatomy attracts you to learn regional anatomy? (5) Do you expect to come back to China earlier? (6) Please list the names of the Traditional Chinese Festival. (7) Please list names of Solar terms in the traditional Chinese calendar. (8) Please list examples of Chinese scientist contribution to world medical science. (9) Please write down measures for combating the COVID-19 pandemic in China. Seventy-three students (82.0%) agreed that introducing the clinical cases in teaching regional anatomy promoted the understanding of the learning content and enhanced the learning effect. Eighty students (89.9%) strongly agreed that those cases retained their attention in class and promoted them to pay more attention to professional spirit. Eighty-two students (92.1%) reported that the introduction of Chinese culture in the teaching of regional anatomy enhanced their cognition of China to a large extent. Eighty-nine students (100%) reported that the involvement in drawing pictures attracted them to learn regional anatomy. All international students hope that the COVID-19 pandemic can be controlled as soon as possible and that they could travel to China for offline learning earlier. Fifty international students (56.2%) were able to list 4–5 Chinese traditional festivals, 29 (32.6%) could list 2–3, and 10 (11.2%) knew less than or equal to one. Fifty-three international students (59.6%) listed 4–5 Chinese seasonal solar terms, 28 students (31.5%) listed 2–3, and eight students (0.09%) could list only one. Forty-two students (47.2%) write down 4–5 examples of contributions to medical progress in China. For example, Zhong Nanshan was known for the SARS virus; Chen Wei is a Chinese epidemiologist and virologist specializing in biodefense. Tu Youyou contributed a lot to the malaria treatment. Thirty-six (40.4%) could list 2–3, and 11 (12.4%) could list only one. Among these examples, 56 (62.9%) international students mentioned traditional Chinese medicine, 48 (53.9%) mentioned acupuncture and 75 (84.3%) mentioned China’s COVID-19 vaccine. For anti-pandemic measures, 60 (67.4%) international students could list 4–5, 20 (22.5%) could list 2–3, and 9 (10.1%) could list one or less. 100% of students hope to come to China to carry out in-person education courses. Taking together, adding medical humanities and local culture into this nonsynchronous online teaching of regional anatomy is appreciated by international students, and achieved good teaching results.
We used formative assessment to evaluate the effects of online teaching of regional anatomy. The formative assessment was composed of four parts. The first part is the completion of online videos, which is scored according to the progress of watching the course video, accounting for 30% of the total score. The second part is the evaluation of general performance (including three items): signature, online quiz, and participation in discussion. Those who have completed three items will get 100 points, those who have completed two items will get 80 points, those who have completed one item will get 60 points, and those who have not completed any items will not get any points. This part accounts for 10% of the total score. The third part is the score of the experiment examination. Using lecture examination scores to predict experimental performance was a feasible way to identify students who may have difficulty in learning practical dissection . In this exam, students were required to find 20 structures on the pictures specified, with 5 marks for each structure. This part accounted for 30% of the total score. The fourth part is the theoretical examination at the end of the semester, accounting for 30% of the total score (Fig. ). As Fig. A shows there is no significant difference between the score of the experimental examination of nonsynchronous online teaching of regional anatomy for international students in 2019 or 2020 and that of offline teaching in 2017 or 2018. Although the theoretical examination of nonsynchronous online teaching of regional anatomy for international students in 2020 decreased significantly, the total score did not decrease but was higher than that in 2019 and 2017 or 2018. The total scores in 2020 were: 19 students (21%) got 85 points or more, 70 students (79%) got 60–80 points, and 100% of students passed the exam (Fig. ). The above results showed that students have mastered the relevant knowledge structures and key points through the implementation of nonsynchronous online teaching of a regional anatomy course.
We also used a questionnaire to investigate the responses from students. The statements in the questionnaire were: (1) Do you agree that introducing the clinical cases in teaching regional anatomy promoted the understanding of the learning content and enhanced the learning effect? (2) Do you agree those clinical cases retain your attention in class and promote you to pay more attention to your professional spirit? (3) Do you agree that the introduction of Chinese culture in the teaching of regional anatomy enhanced your knowledge more about China? (4) Do you agree that drawing pictures related to regional anatomy attracts you to learn regional anatomy? (5) Do you expect to come back to China earlier? (6) Please list the names of the Traditional Chinese Festival. (7) Please list names of Solar terms in the traditional Chinese calendar. (8) Please list examples of Chinese scientist contribution to world medical science. (9) Please write down measures for combating the COVID-19 pandemic in China. Seventy-three students (82.0%) agreed that introducing the clinical cases in teaching regional anatomy promoted the understanding of the learning content and enhanced the learning effect. Eighty students (89.9%) strongly agreed that those cases retained their attention in class and promoted them to pay more attention to professional spirit. Eighty-two students (92.1%) reported that the introduction of Chinese culture in the teaching of regional anatomy enhanced their cognition of China to a large extent. Eighty-nine students (100%) reported that the involvement in drawing pictures attracted them to learn regional anatomy. All international students hope that the COVID-19 pandemic can be controlled as soon as possible and that they could travel to China for offline learning earlier. Fifty international students (56.2%) were able to list 4–5 Chinese traditional festivals, 29 (32.6%) could list 2–3, and 10 (11.2%) knew less than or equal to one. Fifty-three international students (59.6%) listed 4–5 Chinese seasonal solar terms, 28 students (31.5%) listed 2–3, and eight students (0.09%) could list only one. Forty-two students (47.2%) write down 4–5 examples of contributions to medical progress in China. For example, Zhong Nanshan was known for the SARS virus; Chen Wei is a Chinese epidemiologist and virologist specializing in biodefense. Tu Youyou contributed a lot to the malaria treatment. Thirty-six (40.4%) could list 2–3, and 11 (12.4%) could list only one. Among these examples, 56 (62.9%) international students mentioned traditional Chinese medicine, 48 (53.9%) mentioned acupuncture and 75 (84.3%) mentioned China’s COVID-19 vaccine. For anti-pandemic measures, 60 (67.4%) international students could list 4–5, 20 (22.5%) could list 2–3, and 9 (10.1%) could list one or less. 100% of students hope to come to China to carry out in-person education courses. Taking together, adding medical humanities and local culture into this nonsynchronous online teaching of regional anatomy is appreciated by international students, and achieved good teaching results.
During these 2 years, the nonsynchronous online teaching combined with medical humanities and local culture was used to teach regional anatomy to international students. The students not only gained professional knowledge but also obtained enhanced exposure to local culture and professional spirit. The nonsynchronous online teaching with interpreted videos of dissections does not significantly affect students’ experimental achievement in international students. Integrating medical humanities and local culture into this teaching model is appreciated by them, and achieved good teaching results. Regional anatomy has always been taught to the first and second-year medical students through didactic lectures and corpse dissection (approximately 100 medical students for lectures and 30 students per group for corpse dissection). Because of the pandemic, the teaching mode has to be shifted to online teaching . Although Zoom video conference platform (a video conferencing tool) is a very good synchronous learning tool, due to the time difference of foreign students in different countries, we used the nonsynchronous online teaching other than Zoom for regional anatomy teaching, that is, uploading learning materials (such as electronic textbooks, videos for theoretical and experimental teaching and/or recording of narrated lecture slides using PowerPoint presentation) to the e-learning platforms, reminding students regularly, checking the progress of learning, supervising the completion of learning tasks, answering students’ questions and correcting homework from time to time. In the experimental and the total scores of regional anatomy, 100% of students passed the exam. Compared with offline teaching, the experimental results and overall evaluation results of regional anatomy of international students have not decreased significantly, which proves that the nonsynchronous online teaching can partially replace offline teaching, consistent with previous studies . Nonsynchronous online teaching has its advantages, including promoting self-directed learning and providing flexible learning opportunities that would offer continuous (24 h, 7 days a week) availability for learners, especially for international students with time differences in different countries. Some medical students worried the lack of practical teaching about corpse dissection can adversely impact training and lead to losses for students, due to the learning environment being less than ideal . Conversely, while many teachers recognize corpse dissection as a critical method for teaching and learning anatomy in medical education, the number of hours dedicated to corpse dissection has substantially declined over the years, and it has been completely removed from curricula in some cases . Srinivasan reported that effective and successful online learning was implemented using the Zoom video conferencing platform . The addition of videos of dissections can help students to understand even if lack of practical teaching with corpse dissection . Our result showed that nonsynchronous online teaching with interpreted videos of dissections does not significantly affect the experimental achievement of international students or the acquisition of anatomical knowledge by medical students, but the impact on medical students’ professional quality needs to be observed for a longer time (such as 10 or 20 years). With the application of new teaching methods, such as web-based interactive images, scanned three-dimensional (3D) models, and specimen displays on the internet, etc., it has been widely accepted by teachers and students to gradually reduce the contact time of autopsy and increase the integration of anatomy and other courses in the future anatomy teaching . We must ensure we continue to provide students with learning experiences beyond anatomical knowledge that is crucial to their development as a medical practitioner. Interestingly, we found that in asynchronous online teaching in 2020, the scores of theory examinations in the regional anatomy decreased significantly. One of the reasons may be that international students do not pay attention to preparation before the exam. Unlike offline learning, teachers will emphasize reviewing before the exam in class. The second reason may be that during the preparation period of nonsynchronous online teaching, teachers upload all learning materials to the e-learning platforms, so many international students maybe finish all learning materials in the middle of the course but forget it at the end of term, so they cannot get high scores. The improvement method is to upload learning materials step by step according to the learning progress, instead of uploading them all at once in the early stage of nonsynchronous online teaching, and students shall be advised to review them many times before the exam. In order to cultivate a humanistic, moral, and professional quality of international students, the current teaching methods are gradually changing from the traditional “teacher-centered” teaching model to a “student-centered” multi-element fusion teaching mode. Each medical university is constantly exploring its teaching mode according to its aim . International students cannot come to China because of COVID-19, most of them have fewer chances to learn about and experience the local culture. In addition to our professional knowledge, we also introduce Chinese traditional culture, traditional Chinese medicine, and the current pandemic situation, which can improve the understanding of international students about China. At the same time, we play some videos that increase humanistic quality during offline teaching, and actively invite international students to participate in after-class activities, to improve their professional quality. These measures enrich the knowledge of international students other than anatomy, which is appreciated by them, and also had a good teaching effect.
This study has only been observed for 2 years, and the effect of this regional anatomical teaching method needs to be observed for a longer time. There are a few methods and strategies to cultivate humanistic professional quality, which cannot guarantee that every student can enjoy and benefit from it. Further studies should be carried out to assess the effectiveness of various e-learning strategies for conducting tutorials, which could open a new prospect for regional anatomy education.
Integrating medical humanities and local culture into nonsynchronous online teaching of regional anatomy with interpreted videos of dissections not only effectively promotes students learning of professional knowledge, but also makes them know Chinese traditional culture and professional spirit better. This regional anatomy education model should continue to implement even post-COVID-19 pandemic as a supplement because teachers’ efforts to put forward innovative teaching methods will always be appreciated by students. At the same time, the progressive implementation of computer-based learning methods has brought out new challenges and opportunities for anatomical sciences educators.
|
Analyses of human papillomavirus, | 806e3c8f-6fa9-4e3b-9ab6-2eba56c14818 | 10029165 | Gynaecology[mh] | Sexually transmitted infections (STIs) are a major source of morbidity globally in both men and women. The use of Pap smears and human papillomavirus (HPV) testing has led to an increase in the detection of cervical dysplasia and a decrease in the incidence of cervical cancer. Mobile health programs enhance cancer screening rates in the general population and are of enormous value in poor nations. New screening technologies, such as p16/Ki67, HPV self-testing, and the application of artificial intelligence in colposcopy evaluations, should be widely used. Focusing on pre-cancerous lesions and developing tools for women at risk for persistent and recurrent pre-cancerous lesions following primary conization (also containing cryo, LEEP), and identifying groups at high risk have clinical value . HPV infections are among the most common STIs, and chronic infections with high-risk HPV (hrHPV) genotypes can lead to cervical dysplasia and invasive cervical cancer . Common STI pathogens include HPV, Chlamydia trachomatis (CT), Ureaplasma urealyticum (UU), and Neisseria gonorrhoeae (NG). HPV infections are linked to vaginal microenvironmental dysfunction. Compared to women without bacterial vaginosis (BV), the hrHPV infection rate in women with BV is increased . Furthermore, BV is associated with chronic cervical inflammation that injures the mucosal barrier and compromises immune protection to promote hrHPV infection . For example, condyloma acuminata and cervical lesions are caused by chronic HPV infection, and hrHPV has a significant role in the development of cervical cancer . HPV is known to be the leading cause of cervical cancer (CC), accounting for approximately 311,000 deaths globally in 2018 . Moreover, a small percentage of women with HPV will develop a pre-cancerous or malignant lesion . Although no statistically significant link was demonstrated between hrHPV and other STI co-infections and positive colposcopy findings in the study population, the high prevalence of STIs in hrHPV-positive women, who are at higher risk of developing the cervical disease, suggests that screening for genital infections in this population may be important . CT, UU, and NG infections can result in symptoms, including urethritis and vaginal itching, as well as reproductive health issues, including infertility or miscarriage. NG, UU, and CT infections increase the risk of HPV-mediated cervical malignancies , thus early detection of genital tract pathogens is essential for cervical cancer prevention and early cervical cancer screening. To determine the infection rates of CT, UU, and NG, and co-infections with HPV in the Haikou area of China, we analyzed the overall detection of HPV, CT, UU, and NG and the age distribution of patients in the Gynecology Outpatient Clinic of Hainan Women and Children Medical Center between January and December 2021. Study data From January to December 2021, a total of 2389 consecutive female patients (mean age, 31.21 ± 4.24 years) who attended our Gynecology Outpatient Clinic with a suspected genital tract infection were chosen. The inclusion criteria were as follows: 1. at least one symptom of vaginitis (abnormal vaginal discharge; painful or frequent urination; vaginal itching, burning, or irritation; painful or uncomfortable intercourse; and vaginal odor); 2. sexual abstinence for ≥ 24 h; 3. No vaginal douching and no medications for ≥ 72 h; 4. no current menses; and 5. signed informed consent. The exclusion criteria were as follows: atypical genital bleeding; use of medications; and unwillingness to participate in the study. Sample collection Vaginal swabs were obtained from all patients and tested for CT, UU, and NG DNA, as well as cervical exfoliated cells for HPV genotyping. The sampling procedures were carried out in accordance with the sampling tube instructions (National Instrument Note, 20,153,401,995). The sample delivery and preservation methods were standardized. The purpose of the study was explained to eligible patients, who gave informed consent before authorization by the Medical Ethics Committee of the Hainan Women and Children Medical Center (HNWCMC2021-40). Research methods Nucleic acid extraction Nucleic acids were extracted from samples using a magnetic bead technique kit from Shenzhen Ruineng Biotechnology Co., Ltd. (Shenzhen, China), and the procedure was followed precisely according to the kit's instructions (National Instruments Note, 20,143,801,491). HPV detection The sample HPV typing test was performed using a 21 HPV GenoArray Diagnostic Kit (HBGA-21PKG; Hybribio, Kowloon Bay, Hong Kong, China), which detects HPV 6, 11, 16, 18, 31, 33, 35, 39, 42, 43, 44 45, 51, 52, 53, 56, 58, 59, 66, 68, and cp8304. All steps were performed according to the manufacturer’s instructions. The detection instrument was a Hema-9600 with fully automated nucleic acid molecular hybridization technology (HBHM-9001A; Hybribio). CT, UU, and NG assays The Chlamydia trachomatis Diagnostic Kit (DA 0071), Ureaplasma urealyticum Diagnostic Kit [DA1120], and Neisseria gonorrhoeae Test Kit [DA0063] (all from The DaAn Gene Co., Ltd., Guangdong, China) were used to identify the three pathogens, and all procedures were carried out in accordance with the instructions on the kit packaging. The SLAN-96P real-time fluorescence quantitative PCR apparatus was used for DNA detection (Shanghai Toujing Life Technology Co., Shanghai, China). Statistical methods SPSS 26.0 software was used for statistical analyses. Count data are expressed as n. A χ2 or Fisher exact test was used to compare rates between groups, as indicated. The difference was considered statistically significant at a P < 0.05. From January to December 2021, a total of 2389 consecutive female patients (mean age, 31.21 ± 4.24 years) who attended our Gynecology Outpatient Clinic with a suspected genital tract infection were chosen. The inclusion criteria were as follows: 1. at least one symptom of vaginitis (abnormal vaginal discharge; painful or frequent urination; vaginal itching, burning, or irritation; painful or uncomfortable intercourse; and vaginal odor); 2. sexual abstinence for ≥ 24 h; 3. No vaginal douching and no medications for ≥ 72 h; 4. no current menses; and 5. signed informed consent. The exclusion criteria were as follows: atypical genital bleeding; use of medications; and unwillingness to participate in the study. Vaginal swabs were obtained from all patients and tested for CT, UU, and NG DNA, as well as cervical exfoliated cells for HPV genotyping. The sampling procedures were carried out in accordance with the sampling tube instructions (National Instrument Note, 20,153,401,995). The sample delivery and preservation methods were standardized. The purpose of the study was explained to eligible patients, who gave informed consent before authorization by the Medical Ethics Committee of the Hainan Women and Children Medical Center (HNWCMC2021-40). Nucleic acid extraction Nucleic acids were extracted from samples using a magnetic bead technique kit from Shenzhen Ruineng Biotechnology Co., Ltd. (Shenzhen, China), and the procedure was followed precisely according to the kit's instructions (National Instruments Note, 20,143,801,491). HPV detection The sample HPV typing test was performed using a 21 HPV GenoArray Diagnostic Kit (HBGA-21PKG; Hybribio, Kowloon Bay, Hong Kong, China), which detects HPV 6, 11, 16, 18, 31, 33, 35, 39, 42, 43, 44 45, 51, 52, 53, 56, 58, 59, 66, 68, and cp8304. All steps were performed according to the manufacturer’s instructions. The detection instrument was a Hema-9600 with fully automated nucleic acid molecular hybridization technology (HBHM-9001A; Hybribio). CT, UU, and NG assays The Chlamydia trachomatis Diagnostic Kit (DA 0071), Ureaplasma urealyticum Diagnostic Kit [DA1120], and Neisseria gonorrhoeae Test Kit [DA0063] (all from The DaAn Gene Co., Ltd., Guangdong, China) were used to identify the three pathogens, and all procedures were carried out in accordance with the instructions on the kit packaging. The SLAN-96P real-time fluorescence quantitative PCR apparatus was used for DNA detection (Shanghai Toujing Life Technology Co., Shanghai, China). Nucleic acids were extracted from samples using a magnetic bead technique kit from Shenzhen Ruineng Biotechnology Co., Ltd. (Shenzhen, China), and the procedure was followed precisely according to the kit's instructions (National Instruments Note, 20,143,801,491). The sample HPV typing test was performed using a 21 HPV GenoArray Diagnostic Kit (HBGA-21PKG; Hybribio, Kowloon Bay, Hong Kong, China), which detects HPV 6, 11, 16, 18, 31, 33, 35, 39, 42, 43, 44 45, 51, 52, 53, 56, 58, 59, 66, 68, and cp8304. All steps were performed according to the manufacturer’s instructions. The detection instrument was a Hema-9600 with fully automated nucleic acid molecular hybridization technology (HBHM-9001A; Hybribio). The Chlamydia trachomatis Diagnostic Kit (DA 0071), Ureaplasma urealyticum Diagnostic Kit [DA1120], and Neisseria gonorrhoeae Test Kit [DA0063] (all from The DaAn Gene Co., Ltd., Guangdong, China) were used to identify the three pathogens, and all procedures were carried out in accordance with the instructions on the kit packaging. The SLAN-96P real-time fluorescence quantitative PCR apparatus was used for DNA detection (Shanghai Toujing Life Technology Co., Shanghai, China). SPSS 26.0 software was used for statistical analyses. Count data are expressed as n. A χ2 or Fisher exact test was used to compare rates between groups, as indicated. The difference was considered statistically significant at a P < 0.05. Detection of HPV, CT, UU, and NG in 2389 patient samples Among the samples tested, UU had the highest detection rate (56.84% [1358/2389]), followed by HPV (17.29% [413/2389]), CT (7.99% [191/2389]), and NG (0.38% [9/2389]). Detection of different HPV subtypes A total of 413 HPV-positive samples were detected from 2398 patients. The total HPV subtype detection rate was 16.71% (69/413); HPV subtype 52 had the highest detection rate. All 21 HPV subtypes detectable in the kit were detected. The detection rates for the top 5 HPV subtypes were as follows: 52 (69/413); 16 (56/413); 39 (38/413); 51 (38/413); and 58 (36/413). Detection of pathogens in different age groups (Table ) The results showed that the HPV, CT, UU, and NG detection rates were the highest in the 15–20 year age group; the difference was statistically significant compared with the other age groups ( P < 0.05). The HPV, CT, UU, and NG detection rates in the 21–25 year age group was second only to the 15–20 year age group; the difference was statistically significant compared with the other age groups ( P < 0.05). There were no significant differences in the HPV, CT, UU, and NG detection rates between the 26–30, 31–35, 36–40, and ≥ 41 year age groups. Mixed HPV infections with CT, UU, and NG (Table ) The samples were separated into groups of HPV-positive and -negative patients based on the results of HPV tests. CT, UU, and NG detection in each group was recorded. CT [(57 (13.80%) vs. 134 (6.78%); P < 0.001], UU [323 (78.21%) vs. 1035 (52.38%); P < 0.001], and NG [6 (1.45%) vs. 3 (0.15%); P < 0.001] detection rates were significantly higher in the HPV-positive group than the HPV-negative group. Among the samples tested, UU had the highest detection rate (56.84% [1358/2389]), followed by HPV (17.29% [413/2389]), CT (7.99% [191/2389]), and NG (0.38% [9/2389]). A total of 413 HPV-positive samples were detected from 2398 patients. The total HPV subtype detection rate was 16.71% (69/413); HPV subtype 52 had the highest detection rate. All 21 HPV subtypes detectable in the kit were detected. The detection rates for the top 5 HPV subtypes were as follows: 52 (69/413); 16 (56/413); 39 (38/413); 51 (38/413); and 58 (36/413). ) The results showed that the HPV, CT, UU, and NG detection rates were the highest in the 15–20 year age group; the difference was statistically significant compared with the other age groups ( P < 0.05). The HPV, CT, UU, and NG detection rates in the 21–25 year age group was second only to the 15–20 year age group; the difference was statistically significant compared with the other age groups ( P < 0.05). There were no significant differences in the HPV, CT, UU, and NG detection rates between the 26–30, 31–35, 36–40, and ≥ 41 year age groups. ) The samples were separated into groups of HPV-positive and -negative patients based on the results of HPV tests. CT, UU, and NG detection in each group was recorded. CT [(57 (13.80%) vs. 134 (6.78%); P < 0.001], UU [323 (78.21%) vs. 1035 (52.38%); P < 0.001], and NG [6 (1.45%) vs. 3 (0.15%); P < 0.001] detection rates were significantly higher in the HPV-positive group than the HPV-negative group. The HPV detection rate in outpatients was 17.29%, which is higher than the average rate of HPV carriage among Chinese women countrywide (11.2%) . The top 5 HPV subtypes found in Haikou were 52, 16, 39, 51, and 58, which were different from the top HPV subtypes in Shanghai and Yunnan, according to the data currently available, which suggests that geographic location may have an impact on the HPV subtype detection rate . The findings of this study demonstrated that among patients with an HPV infection, there was a higher likelihood of detecting CT, UU, and NG in cervical samples. The pathogen identification frequencies among the 2389 samples were as follows: UU (58.43%); HPV (17.29%); CT (7.99%); and NG (0.38%). The prevalence of UU has been reported to be 20% in South America, 41.9% in Italy, and 51.5% in Africa . The highest prevalence of UU (53.8%) was detected in females between 34 and 39 years of age . HPV is mostly spread via sexual contact, and the likelihood of infection increases with age and sexual behavior . HPV is important, but not sufficient for cervical carcinogenesis, suggesting the involvement of other variables. The vaginal microbiota is a key factor in managing HPV infections, and depending on the composition, the vaginal microbiota can alter the vaginal mucosa microenvironment against viral infections. Numerous studies have shown an increased risk of HPV infection at younger ages . Specifically, the highest prevalence of HPV occurs among adolescents and young adults between the ages of 15 and 25 years. It has been estimated that > 75% of new HPV infections occur in individuals in this age range . This increased risk of infection in younger women has been linked to a lack of adaptive immune responses and/or a relatively large area of cervical epithelium undergoing squamous metaplasia in this age group, which may increase the opportunity for HPV DNA to infect the basal cell layer, then proliferate . In agreement with local and international studies , we showed that the 15–25 year age range in gynecologic outpatients has a higher HPV detection rate than the total HPV detection rate. The primary pathogens that cause infections of the female reproductive system are HPV, CT, UU, and NG. In the early stages of infection, these pathogens frequently cause no or moderate symptoms, which makes it easy to miss the diagnosis or misdiagnose the symptoms and cause recurring sickness, such as non-gonococcal urethritis, pelvic inflammation, gonorrhea, and chronic cervicitis . In contrast, mixed infections caused by several different microorganisms make it difficult to diagnose and treat the illness. The probability of HPV infection is increased by polymicrobial co-infections in addition to age-related variables . Cervical mucosa, which acts as a protective barrier against pathogens penetrating the upper female reproductive system, is frequently challenged by pathogens and dysbiosis . Co-infections are becoming more common in clinical practice, although the role in disease progression is unknown . Co-infections with CT are more common in patients with invasive cervical and ovarian malignancies . The impact of co-existing HPV and CT in a stem cell could be detrimental to cellular and genomic integrity, thus promoting neoplastic growth. CT and HPV generate unique transcriptional and post-translational responses, resulting in distinct reprogramming of host cell processes. Surprisingly, CT interferes with HPV-induced mechanisms that maintain cellular and genomic integrity, such as mismatch repair in stem cells . CT is recognized as a significant co-factor for HPV infections . Other STIs are increasingly being implicated as co-factors in the development of cervical cancer in HPV-positive women . Interactions between HPV and other infections with comparable mucosal locations may hasten cancer progression by increasing HPV replication and infection persistence. Recent studies, for example, have reported an elevated risk of cervical cancer in women infected with HPV and CT . Both HPV and CT cause changes in the cervical immune system and vaginal microbial ecosystem. According to Seraceni et al. , cervical cancer caused by HPV is strongly correlated with the presence of a CT infection, which makes it easier for HPV to infect cervical epithelial cells. In contrast to an HPV mono-infection, Khan et al. showed that a CT-HPV mixed infection is related to modifications in the oncogenic protein expression profile of cervical cells, which may help increase HPV infection when CT is present. The pathogen that most often infects the female reproductive system, UU, has recently been identified as one of the major risk factors for the development of HPV infection or cervical dysplasia . When UU and HPV infection are combined, Wang et al. reported an elevated incidence of high-grade squamous intraepithelial lesions plus invasive cervical cancer . Based on a meta-analysis, Ye concluded that UU enhanced the rate of high-risk HPV infections and the progression to cervical cancer. Co-infection with HPV and UU has previously been reported in males with urethritis . Although there were few NG-positive samples and a low incidence of NG infections in our study, we did show a higher rate of NG infections in the HPV-positive group. In addition, we showed a low rate of NG infections and a low number of NG-positive samples, but still found an increased rate of NG infections in the HPV-positive group. This finding suggests that NG infections also increase the risk of HPV infections and should be taken seriously and treated aggressively. Early identification of genital tract pathogens is important for the prevention and treatment of cervical cancer , and because infection with CT, UU, and NG enhances HPV-mediated cervical malignant lesions , our study showed the practical value of combined HPV, UU, CT, and NG testing. It is currently thought that genital tract infections are closely associated with age and sexual habits. An age-specific prevalence curve revealed that HPV infections first peak in women < 30 years of age, then decrease with age, but the curve surges again at approximately 50 years of age and peaks again in women ≥ 60 years of age . Due to the lack of sex education, adolescents are more likely to have high-risk sexual behavior and neglect treatment. From a biological perspective, adolescent females are more likely to be infected with reproductive tract pathogens, such as CT and HPV, due to reduced cervical mucus secretion and increased cervical ectopion . According to a study conducted from 2005–2015, the initial age of cervical cancer in Chinese women appears at approximately 25 years of age, then the incidence of cervical cancer tends to decrease . In recent years it has been shown that the time of sexual debut in the adolescent student population (15–24 years of age) in China is earlier and the proportion is increasing, and understanding the need for self-protection is inadequate . A questionnaire survey related to the sexual behavior of adolescents 14–24 years of age in Michigan (USA) pointed out that the lack of family and school education, as well as proper guidance by the media, is the cause of high-risk sexual behavior among adolescents . According to recent research, Blacks are less likely to be infected with HPV16 than Whites, based on self-reported race and genomic ancestry-relevant markers, but are more likely to be positive for other carcinogenic HPV strains . In this study we found a greater incidence of genital tract pathogens in patients between 15 and 25 years of age in the Haikou gynecologic clinics, indicating an uptick in genital tract infections in this age range. It is more important to increase reproductive health knowledge from families, schools, and networks to give early reproductive health education to young women and encourage them adopt prudent reproductive health protective measures. Indeed, this is crucial to lowering reproductive tract infections and enhancing the reproductive health of women who are ready to have children. Small quantities of RNA and DNA, as well as tumor cells, may now be detected in peripheral blood using technological advances, such as circulating tumor cells (CTCs), circulating cell-free DNA (cfDNA), circulating HPV DNA, and miRNA. Circulating molecules and biomarkers have the potential to be a useful diagnostic and prognostic tool for cervical cancer . Furthermore, the presence of hrHPV types, positive endocervical margins, HPV persistence, and the omission of HPV vaccination after conization increase the risk of developing cervical dysplasia persistence and/or recurrence independent of the risk of developing cervical dysplasia persistence and/or recurrence. A nomogram is an effective tool for counseling women about their likelihood of cervical lesion recurrence following initial conization. Triaging high-risk patients in specialist facilities and tailoring more suitable follow-up plans may be beneficial . The single hospital from which all of the patients were chosen was a limitation that may have reduced the ability to generalize the findings. Future studies should include subgroup studies on age and co-infections with HPV, CT, UU, and NG. Among Haikou gynecologic outpatients, UU is the genital tract infection with the highest detection rate. To prevent mixed infections from genital tract pathogens, integrated polymicrobial testing is warranted in HPV-positive carriers. Reproductive tract infections are on the rise in younger people, thus adolescent sexual health education must be improved. |
Population Pharmacokinetics, Pharmacogenomics, and Adverse Events of Osimertinib and its Two Active Metabolites, AZ5104 and AZ7550, in Japanese Patients with Advanced Non-small Cell Lung Cancer: a Prospective Observational Study | 829fb4d8-085c-410d-855c-2d8b59183a45 | 10030409 | Pharmacology[mh] | Osimertinib, a third-generation epidermal growth factor receptor (EGFR)-tyrosine kinase inhibitor (TKI), is a first-line treatment for patients with advanced EGFR-positive non-small cell lung cancer (NSCLC) . It has an excellent efficacy and safety profile; however, over 80% and 30% of patients administered osimertinib experience grade ≥ 2 and ≥ 3 adverse events (AEs), respectively . It was initially approved as a second or later-line treatment for patients with acquired EGFR T790M mutation, but the indication was extended to include it as a first-line treatment and adjuvant therapy after tumor resection . Therefore, long-term administration to asymptomatic early-stage patients is expected, increasing the importance of AE management for osimertinib treatment. Therapeutic drug monitoring (TDM) and the use of biomarkers can be novel strategies for AE management. The potential benefits of TDM in TKI treatments have been suggested , and osimertinib treatments may also benefit from TDM, as it has characteristics suitable for TDM: long-term administration is required ; a validated bioanalytical method is available ; high interindividual variability of exposure with coefficients of variation of 27.6–37.6% is observed ; and a correlation between the exposure to osimertinib parent compound and the occurrence of AEs exists . Osimertinib has two active metabolites, AZ5104 and AZ7550, which may be essential compounds for TDM. AZ5104 is produced by cytochrome P450 (CYP) 3A4 through demethylation of osimertinib indole N- methyl, whereas AZ7550 is produced by demethylation of the osimertinib terminal amine . Both active metabolites circulate at 10% exposure of the parent compound. However, AZ5104 has a 15-fold higher potency compared with the parent compound against wild-type EGFR. In contrast, AZ7550 has a similar potency but a longer half-life; therefore, higher accumulation is expected for it than for the parent compound . Thus, TDM of these two active metabolites may be beneficial for AE management, but the relevance of monitoring these compounds has not yet been elucidated. Some germline polymorphisms in the target EGFR , breast cancer resistance protein (BCRP/ ABCG2 ), drug transporter P-glycoprotein (MDR1/ ABCB1 ), and metabolism-related genes (cytochrome P450 oxidoreductase, POR ) are associated with the occurrence of AEs or pharmacokinetics (PK) of the first- and second-generation EGFR-TKIs (gefitinib, erlotinib, and afatinib) . These polymorphisms can act as novel biomarkers to predict osimertinib AEs, since the active metabolite of osimertinib has a high potency against wild-type EGFR , is a substrate of ABCG2 and ABCB1, and is metabolized by cytochrome P450, which requires electron transfer via POR . While there have been reports on the association between exposure to osimertinib parent compound and its efficacy/safety , no studies have explored the relevance of monitoring the two active metabolites of osimertinib for AE management. Additionally, germline polymorphisms in EGFR , ABCG2 , ABCB1 , or POR have a significant impact on the AEs of first- and second-generation EGFR-TKIs , but those of osimertinib have not been investigated. This study aimed to evaluate (1) the exposure–toxicity relationship and (2) the association of germline polymorphisms with osimertinib AEs to provide evidence for safe treatment and quality-of-life improvement for patients with NSCLC treated with long-term administration of osimertinib. Study design and patients This prospective, longitudinal observational study was designed and conducted at the Ageo Central General Hospital from February 2019 to July 2020 and Keio University Hospital from June 2020 to April 2022. The primary endpoint was the association of exposures to osimertinib, AZ5104, AZ7550, or germline polymorphisms with AE severity. Patients with EGFR -mutation-positive NSCLC aged ≥ 20 years who were orally administered osimertinib (standard dose: 80 mg tablet/day) were included in this study. The inclusion criteria did not restrict the type of EGFR mutation, disease or treatment history, or line of treatment. Patients who were mentally or physically incapable of providing informed consent were excluded from the study. The protocol of this study was reviewed and approved by the ethics committees of Ageo Central General Hospital (Approval No. 564), Keio University School of Medicine (Approval No. 20,200,098), and Keio University Faculty of Pharmacy (Approval No. 210,118–3 and 200,710–1), and written informed consent was obtained from all participants. The study was conducted with adherence to the Declaration of Helsinki. Data collection AEs were assessed during hospital stays for 3 months or at three outpatient visits, when blood was collected at 2 months or later after the initial osimertinib administration (onset of most EGFR-TKI AEs are reported to be within 2–4 months after initial administration ). The severity of AEs (Online Resource, Table D1) was scaled according to the Common Toxicity Criteria for Adverse Effects (CTCAE) version v5.0 by pharmacists and physicians. Serum samples for PK analysis and whole peripheral blood samples for genotyping were opportunistically collected (i.e., leftovers from routine laboratory blood analysis) once every 1–2 months after commencing osimertinib therapy. Patients were asked to determine the time of drug intake. The collected serum and whole peripheral blood samples were stored at − 80 °C until analysis. The serum concentrations of the osimertinib parent compound, AZ5104, and AZ7550 were analyzed as previously described . Briefly, osimertinib, AZ5104, and AZ7550 were extracted from 100 µL serum using a protein precipitation method and analyzed simultaneously using liquid chromatography–tandem mass spectrometry. Genotyping Polymorphisms in EGFR , ABCG2 , ABCB1 , and POR (Online Resource, Appendix A) were analyzed using TaqMan® probe-based assays (Applied Biosystems, Foster City, CA, USA), whereas the ABCG2 polymorphism (rs2231137) was studied using the CycleavePCR® assay (TaKaRa Bio Inc., Kusatsu, Japan). The detailed genotyping method is provided in the Online Resource (Appendix A). Exposure–toxicity and pharmacogenomics–toxicity relationship A population pharmacokinetic (PopPK) model was developed using Phoenix® NLME™ 8.3 software (Certara, Princeton, NJ, USA) to estimate the area under the serum concentration–time curve from 0 to 24 h (AUC 0–24 ) of the osimertinib parent compound, AZ5104, and AZ7550, applied as exposure measures. The detailed method for PopPK model development is provided in the Online Resource (Appendix B). The worst grade of AE that occurred in each patient scaled using CTCAE was used for the exposure–toxicity relationship analysis. The AUC 0–24 was simulated using the developed PopPK model based on individual predicted concentrations at the time closest to the occurrence of AEs. For patients who did not experience AEs, a median of three simulated AUC 0–24 from three concentration data points was applied for the analysis. Germline polymorphism and the worst grade of AE occurring in each patient were analyzed for pharmacogenomics–toxicity relationship analysis. Statistical analysis The cut-off date for data collection was April 4, 2022. Continuous variables such as AUC 0–24 and laboratory data are presented as median (interquartile range, IQR); estimates of PopPK parameters are presented as mean (standard error, SE). Exposure–toxicity and pharmacogenomics–exposure analyses (comparison of continuous variables) were performed using Mann–Whitney U test. Receiver operating characteristic (ROC) curve analysis was used to evaluate the discrimination potential of AUC 0–24 for grade ≥ 2 AEs. Significance of deviation of allele and genotype frequencies from Hardy–Weinberg equilibrium was tested, and pharmacogenomics–toxicity analysis (comparison of categorical variables) was performed using Fisher’s exact test. For the pharmacogenomics association analysis, multiple statistical analyses were performed using additive, recessive, and dominant genetic models. All p -values were two-sided, and statistical significance was set at p < 0.05. IBM® SPSS® statistics version 28.0 (SPSS, Inc., Chicago, IL, USA) was used for all the statistical analyses. This prospective, longitudinal observational study was designed and conducted at the Ageo Central General Hospital from February 2019 to July 2020 and Keio University Hospital from June 2020 to April 2022. The primary endpoint was the association of exposures to osimertinib, AZ5104, AZ7550, or germline polymorphisms with AE severity. Patients with EGFR -mutation-positive NSCLC aged ≥ 20 years who were orally administered osimertinib (standard dose: 80 mg tablet/day) were included in this study. The inclusion criteria did not restrict the type of EGFR mutation, disease or treatment history, or line of treatment. Patients who were mentally or physically incapable of providing informed consent were excluded from the study. The protocol of this study was reviewed and approved by the ethics committees of Ageo Central General Hospital (Approval No. 564), Keio University School of Medicine (Approval No. 20,200,098), and Keio University Faculty of Pharmacy (Approval No. 210,118–3 and 200,710–1), and written informed consent was obtained from all participants. The study was conducted with adherence to the Declaration of Helsinki. AEs were assessed during hospital stays for 3 months or at three outpatient visits, when blood was collected at 2 months or later after the initial osimertinib administration (onset of most EGFR-TKI AEs are reported to be within 2–4 months after initial administration ). The severity of AEs (Online Resource, Table D1) was scaled according to the Common Toxicity Criteria for Adverse Effects (CTCAE) version v5.0 by pharmacists and physicians. Serum samples for PK analysis and whole peripheral blood samples for genotyping were opportunistically collected (i.e., leftovers from routine laboratory blood analysis) once every 1–2 months after commencing osimertinib therapy. Patients were asked to determine the time of drug intake. The collected serum and whole peripheral blood samples were stored at − 80 °C until analysis. The serum concentrations of the osimertinib parent compound, AZ5104, and AZ7550 were analyzed as previously described . Briefly, osimertinib, AZ5104, and AZ7550 were extracted from 100 µL serum using a protein precipitation method and analyzed simultaneously using liquid chromatography–tandem mass spectrometry. Polymorphisms in EGFR , ABCG2 , ABCB1 , and POR (Online Resource, Appendix A) were analyzed using TaqMan® probe-based assays (Applied Biosystems, Foster City, CA, USA), whereas the ABCG2 polymorphism (rs2231137) was studied using the CycleavePCR® assay (TaKaRa Bio Inc., Kusatsu, Japan). The detailed genotyping method is provided in the Online Resource (Appendix A). A population pharmacokinetic (PopPK) model was developed using Phoenix® NLME™ 8.3 software (Certara, Princeton, NJ, USA) to estimate the area under the serum concentration–time curve from 0 to 24 h (AUC 0–24 ) of the osimertinib parent compound, AZ5104, and AZ7550, applied as exposure measures. The detailed method for PopPK model development is provided in the Online Resource (Appendix B). The worst grade of AE that occurred in each patient scaled using CTCAE was used for the exposure–toxicity relationship analysis. The AUC 0–24 was simulated using the developed PopPK model based on individual predicted concentrations at the time closest to the occurrence of AEs. For patients who did not experience AEs, a median of three simulated AUC 0–24 from three concentration data points was applied for the analysis. Germline polymorphism and the worst grade of AE occurring in each patient were analyzed for pharmacogenomics–toxicity relationship analysis. The cut-off date for data collection was April 4, 2022. Continuous variables such as AUC 0–24 and laboratory data are presented as median (interquartile range, IQR); estimates of PopPK parameters are presented as mean (standard error, SE). Exposure–toxicity and pharmacogenomics–exposure analyses (comparison of continuous variables) were performed using Mann–Whitney U test. Receiver operating characteristic (ROC) curve analysis was used to evaluate the discrimination potential of AUC 0–24 for grade ≥ 2 AEs. Significance of deviation of allele and genotype frequencies from Hardy–Weinberg equilibrium was tested, and pharmacogenomics–toxicity analysis (comparison of categorical variables) was performed using Fisher’s exact test. For the pharmacogenomics association analysis, multiple statistical analyses were performed using additive, recessive, and dominant genetic models. All p -values were two-sided, and statistical significance was set at p < 0.05. IBM® SPSS® statistics version 28.0 (SPSS, Inc., Chicago, IL, USA) was used for all the statistical analyses. Data collection Patient characteristics are summarized in Table . A total of 302 serum samples for PK analysis, median (IQR) time after administration of 6.3 (2.0–23.6) h, and 53 whole peripheral blood samples for genotyping were collected from 53 patients. The median number of serum samples per patient was five, collected within a median of 5 months. Osimertinib safety was evaluated in 51 patients; we were unable to collect enough safety data from 2 out of 53 patients because of transfer to another hospital and withdrawal of consent. The lymphocytes for one of the patients and creatine phosphokinase for 20 patients were not tested in routine clinical laboratory blood analysis. The most prevalent grade ≥ 2 AE was skin disorders (63%, Online Resource, Table D1). Approximately 25% of the patients experienced severe AEs (grade 3 or leading to dose discontinuation). The patient characteristic significantly associated with severe AEs was the EGFR-TKI treatment line, with higher frequency of severe AEs in patients receiving therapy as first-line treatment (Online Resource, Table D2, p = 0.004). Other characteristics, including sex, age, ECOG PS, and somatic EGFR mutation, were not associated with any grade ≥ 2 or ≥ 3 AEs. Development and evaluation of the PopPK model The final PopPK parameters are listed in Table . Albumin was identified as a significant covariate for the clearance of the parent compound and AZ5104, whereas body weight (BW) was identified as a significant covariate for the clearance of AZ7550 (Online Resource, Table C1–C3, p < 0.05). The PopPK parameters were estimated using models (1), (2), and (3) for osimertinib, AZ5104, and AZ7550, respectively. The estimated values were close to the mean values calculated from the bootstrap sampling of n = 974 (success rate, 97.4%), n = 1000 (success rate, 100%), and n = 1000 (success rate, 100%), respectively, and all fell within the 95% percentile confidence intervals (Table ). The detailed results for the development and evaluation of the PopPK model are provided in the Online Resource (Appendix C). Exposure–toxicity relationship The median (IQR) values of AUC 0–24 at steady state (estimated using the final PopPK model) of the osimertinib parent compound, AZ5104, and AZ7550 were 4278 (3328–5589) ng/mL*h, 414 (309–584) ng/mL*h, and 367 (275–516) ng/mL*h, respectively, regardless of dosage. There was a significant association between the AUC 0–24 of the active metabolite AZ7550 and grade ≥ 2 paronychia or grade ≥ 2 anorexia; AUC 0–24 of the osimertinib parent compound or active metabolite AZ5104 and grade ≥ 2 diarrhea; and AUC 0–24 of the parent or either of the two active metabolites and grade ≥ 2 increased creatinine. Overall, the AUC 0–24 of AZ5104 was significantly associated with any grade ≥ 2 AEs (Fig. ). The cut-off AUC 0–24 of the parent compound, AZ5104, and AZ7550 to predict grade ≥ 2 AEs were 4938 ng/mL*h, 540 ng/mL*h, and 462 ng/mL*h, respectively, based on the ROC curve. The areas under the curve (sensitivity and specificity) were 0.680 (51% and 90%), 0.705 (42% and 100%), and 0.651 (49% and 90%), respectively (Online Resource, Fig. E1). The frequency of the aforementioned AEs (grade ≥ 2) was higher in patients with ≥ 4938 ng/mL*h parent compound, ≥ 540 ng/mL*h AZ5104, or ≥ 462 ng/mL*h AZ7550 AUC 0–24 (Online Resource, Table E1–E3). Pharmacogenomics-toxicity relationship All analyzed genotypes were in Hardy–Weinberg equilibrium ( p > 0.050), except for EGFR rs4947492, EGFR rs2227983, ABCG2 rs2622604, and ABCB1 rs1045642 ( p = 0.047, p = 0.020, p = 0.049, and p = 0.028, respectively), but the allele frequency was similar to that reported by Togo Var . There was a significant relationship between polymorphisms in germline EGFR and severe AEs or anorexia. EGFR rs2293348 C > T (C/T genotype) and EGFR rs4947492 G > A (A/A genotype) were both associated with any severe AEs (Table , p = 0.019 in the additive model, and p = 0.050 in the recessive model). EGFR rs2293348 C > T (C/T genotype) was also significantly associated with grade ≥ 1 or ≥ 2 anorexia (Online Resource, Table F1, p < 0.001 or p = 0.023, respectively, in an additive model). The ABCG2 rs2231137 G > A polymorphism (G/A and A/A genotypes) was significantly associated with both AEs and PK. A higher frequency of any grade ≥ 2 AEs (Table , p = 0.008 in a dominant model) was observed in patients with ABCG2 rs2231137 G/A or A/A genotypes. Furthermore, patients with the G/A or A/A genotype had a significantly higher exposure (AUC 0–24 ) to the osimertinib parent compound (Fig. , p = 0.018). A significant association was observed between ABCB1 rs1128503 C > T polymorphism (C/T and T/T genotypes) and any grade ≥ 2 AEs (Table , p = 0.038 in a dominant model). Additionally, the ABCB1 rs2032582 G > T/A polymorphism (A/T and G/T genotypes) was associated with any grade ≥ 2 skin disorders and grade ≥ 2 lymphocyte count decrease (Online Resource, Table F2, p = 0.017; Table F3, p = 0.033; respectively, in an additive model). Polymorphisms in the POR were not significantly associated with any AEs or PK of the drug. Patient characteristics are summarized in Table . A total of 302 serum samples for PK analysis, median (IQR) time after administration of 6.3 (2.0–23.6) h, and 53 whole peripheral blood samples for genotyping were collected from 53 patients. The median number of serum samples per patient was five, collected within a median of 5 months. Osimertinib safety was evaluated in 51 patients; we were unable to collect enough safety data from 2 out of 53 patients because of transfer to another hospital and withdrawal of consent. The lymphocytes for one of the patients and creatine phosphokinase for 20 patients were not tested in routine clinical laboratory blood analysis. The most prevalent grade ≥ 2 AE was skin disorders (63%, Online Resource, Table D1). Approximately 25% of the patients experienced severe AEs (grade 3 or leading to dose discontinuation). The patient characteristic significantly associated with severe AEs was the EGFR-TKI treatment line, with higher frequency of severe AEs in patients receiving therapy as first-line treatment (Online Resource, Table D2, p = 0.004). Other characteristics, including sex, age, ECOG PS, and somatic EGFR mutation, were not associated with any grade ≥ 2 or ≥ 3 AEs. The final PopPK parameters are listed in Table . Albumin was identified as a significant covariate for the clearance of the parent compound and AZ5104, whereas body weight (BW) was identified as a significant covariate for the clearance of AZ7550 (Online Resource, Table C1–C3, p < 0.05). The PopPK parameters were estimated using models (1), (2), and (3) for osimertinib, AZ5104, and AZ7550, respectively. The estimated values were close to the mean values calculated from the bootstrap sampling of n = 974 (success rate, 97.4%), n = 1000 (success rate, 100%), and n = 1000 (success rate, 100%), respectively, and all fell within the 95% percentile confidence intervals (Table ). The detailed results for the development and evaluation of the PopPK model are provided in the Online Resource (Appendix C). The median (IQR) values of AUC 0–24 at steady state (estimated using the final PopPK model) of the osimertinib parent compound, AZ5104, and AZ7550 were 4278 (3328–5589) ng/mL*h, 414 (309–584) ng/mL*h, and 367 (275–516) ng/mL*h, respectively, regardless of dosage. There was a significant association between the AUC 0–24 of the active metabolite AZ7550 and grade ≥ 2 paronychia or grade ≥ 2 anorexia; AUC 0–24 of the osimertinib parent compound or active metabolite AZ5104 and grade ≥ 2 diarrhea; and AUC 0–24 of the parent or either of the two active metabolites and grade ≥ 2 increased creatinine. Overall, the AUC 0–24 of AZ5104 was significantly associated with any grade ≥ 2 AEs (Fig. ). The cut-off AUC 0–24 of the parent compound, AZ5104, and AZ7550 to predict grade ≥ 2 AEs were 4938 ng/mL*h, 540 ng/mL*h, and 462 ng/mL*h, respectively, based on the ROC curve. The areas under the curve (sensitivity and specificity) were 0.680 (51% and 90%), 0.705 (42% and 100%), and 0.651 (49% and 90%), respectively (Online Resource, Fig. E1). The frequency of the aforementioned AEs (grade ≥ 2) was higher in patients with ≥ 4938 ng/mL*h parent compound, ≥ 540 ng/mL*h AZ5104, or ≥ 462 ng/mL*h AZ7550 AUC 0–24 (Online Resource, Table E1–E3). All analyzed genotypes were in Hardy–Weinberg equilibrium ( p > 0.050), except for EGFR rs4947492, EGFR rs2227983, ABCG2 rs2622604, and ABCB1 rs1045642 ( p = 0.047, p = 0.020, p = 0.049, and p = 0.028, respectively), but the allele frequency was similar to that reported by Togo Var . There was a significant relationship between polymorphisms in germline EGFR and severe AEs or anorexia. EGFR rs2293348 C > T (C/T genotype) and EGFR rs4947492 G > A (A/A genotype) were both associated with any severe AEs (Table , p = 0.019 in the additive model, and p = 0.050 in the recessive model). EGFR rs2293348 C > T (C/T genotype) was also significantly associated with grade ≥ 1 or ≥ 2 anorexia (Online Resource, Table F1, p < 0.001 or p = 0.023, respectively, in an additive model). The ABCG2 rs2231137 G > A polymorphism (G/A and A/A genotypes) was significantly associated with both AEs and PK. A higher frequency of any grade ≥ 2 AEs (Table , p = 0.008 in a dominant model) was observed in patients with ABCG2 rs2231137 G/A or A/A genotypes. Furthermore, patients with the G/A or A/A genotype had a significantly higher exposure (AUC 0–24 ) to the osimertinib parent compound (Fig. , p = 0.018). A significant association was observed between ABCB1 rs1128503 C > T polymorphism (C/T and T/T genotypes) and any grade ≥ 2 AEs (Table , p = 0.038 in a dominant model). Additionally, the ABCB1 rs2032582 G > T/A polymorphism (A/T and G/T genotypes) was associated with any grade ≥ 2 skin disorders and grade ≥ 2 lymphocyte count decrease (Online Resource, Table F2, p = 0.017; Table F3, p = 0.033; respectively, in an additive model). Polymorphisms in the POR were not significantly associated with any AEs or PK of the drug. To the best of our knowledge, this is the first report to evaluate the association between exposures to the two active metabolites of osimertinib (AZ5104 and AZ7550) or germline polymorphisms ( EGFR , ABCG2 , ABCB1 , and POR ) and the severity of AEs. PopPK modeling was performed to estimate the AUC 0–24 , which was selected as an exposure measure given the suggested linear relationship between the AUC of osimertinib parent compound (at a steady state of any dosing interval) and AEs . The median estimated AUC 0–24 of osimertinib at steady state was consistent or slightly lower than that in previous reports because some patients included in our study received a reduced dose . Albumin level was identified as a significant covariate for the clearance of the parent compound and AZ5104 (positive correlation), consistent with a previous report . C-reactive protein (CRP) is a covariate of osimertinib clearance . Decrease in albumin level and an increase in CRP level occur during inflammation, which can decrease CYP3A4 activity . Therefore, decreased albumin level may be an indicator of increased exposures to the osimertinib parent compound and AZ5104—a consequence of decreased CYP3A4 activity during inflammation. However, elucidating these mechanisms is outside the scope of this study. To our knowledge, our study is the first to report a PopPK model and a significant covariate (BW) for AZ7550. The BW may have an impact on the clearance of AZ7550 because AZ7550 has a longer half-life than the parent compound and AZ5104 (72.7 h versus 59.7 and 52.6 h, respectively), and greater distribution and accumulation are expected . The exposure–toxicity relationships were different among the three compounds: osimertinib parent, AZ5104, and AZ7550. Exposure to AZ7550 was associated with grade ≥ 2 paronychia and anorexia, both of which are potentially related to direct epidermal or mucosal damage. The mechanisms underlying EGFR-TKI-induced skin disorders, such as paronychia, involve decreased keratinocyte proliferation and increased keratinocyte differentiation at the epidermis, resulting in impaired skin barrier function . However, anorexia can be caused by multiple factors, and assessing causal associations is difficult. Although the mechanism responsible for EGFR-TKI-induced anorexia has not been elucidated, one hypothesis for anorexia induction is gastric mucosal injury, since inhibition of EGFR in gastric parietal cells can interfere with gastric mucosal membrane protection and repair of mucosal injury . Moreover, cancer cachexia should be considered a confounder of the AZ7550 exposure–anorexia relationship. The main characteristics of cancer cachexia are anorexia and inflammation, and the elevated production of proinflammatory cytokines can cause a decrease in CYP3A4 activity, resulting in alterations in AZ7550 metabolism . However, because our PopPK analysis indicated that the albumin level, which may be related to inflammation, was not a significant covariate for AZ7550 clearance, we considered that inflammation was unlikely to affect the PK of AZ7550. It is more likely that the increase in exposure to AZ7550 increases direct epidermal or mucosal damage, leading to a higher severity of paronychia or anorexia. Our results also indicated that exposures to the osimertinib parent compound and AZ5104 were associated with grade ≥ 2 diarrhea and increase in creatinine level. Similarly, a previous report suggested a linear relationship between the exposure to the parent compound and the occurrence of diarrhea , whereas another report suggested that exposure to the parent compound and grade ≥ 1 diarrhea were not associated ; thus, exposure to the parent compound may be related to higher grade (grade ≥ 2) diarrhea. The potential mechanism responsible for both diarrhea and creatinine increase may be the activation of chloride secretion as a result of EGFR inhibition . The increased activation of chloride secretion by increased parent compound and AZ5104 exposures can enhance passive water movement through the gastrointestinal lumen, causing high-grade diarrhea and dehydration, which can lead to renal failure . Assessment of the causal association for the exposure–kidney failure relationship is difficult because kidney failure can influence drug excretion. On the one hand, only 1.7% of the dose is excreted in the urine as osimertinib, AZ5104, and AZ7550 (oral bioavailability of osimertinib: 69.8% ), and changes in renal clearance are unlikely to affect drug exposure . On the other hand, because the influence of renal impairment on CYP3A enzyme function has been reported , the potential impact on total clearance cannot be ignored, and further studies are warranted to confirm this influence. Overall, exposure to AZ5104 was significantly associated with any grade ≥ 2 AEs; the exposure to AZ5104 showed a higher area under the curve in ROC analysis than that to osimertinib parent compound and AZ7550. Likewise, an earlier study suggested an absence of a relationship between the trough concentration of the osimertinib parent compound and any toxicity ; thus, monitoring AZ5104 may be more beneficial than monitoring the parent compound for the management of any AEs. Additionally, the results of ROC analysis predicting any grade ≥ 2 AEs using AUC 0–24 of the parent compound, AZ5104, and AZ7550, showed low sensitivity but high specificity. Therefore, the AUC 0–24 levels may not be an absolute index for any AE prediction, but it may become a useful index to decide whether a patient experiencing AEs needs dose reduction. Since several reports suggested that there is no relationship between exposure to osimertinib and efficacy, dose reduction up to 50% of mean exposure may be considered for patients experiencing AEs and high AUC 0–24 levels; however, further validation is needed . The two intron variants in EGFR , rs2293348 C > T (C/T genotype) and rs4947492 G > A (A/A genotype), were associated with severe AEs. According to earlier studies, EGFR rs2293348 is related to erlotinib-induced rash, and rs4947492 is related to gefitinib-induced diarrhea . The role of these genetic variants has not been fully investigated; however, since the intron sequence is involved in the regulation of expression, these variants may cause alterations in EGFR expression . Thus, EGFR rs2293348 and rs4947492 may influence sensitivity to EGFR inhibition and severity of AEs. The germline polymorphisms in ABCG2 rs2231137 G > A (G/A and A/A genotypes) and ABCB1 rs1128503 C > T (C/T and T/T genotypes) were significantly associated with any grade ≥ 2 AEs, and ABCB1 rs2032582 G > T/A polymorphism (A/T and G/T genotypes) was significantly associated with any grade ≥ 2 skin disorders. Osimertinib is a substrate of ABCG2 and ABCB1; thus, polymorphisms in these genes may influence the distribution or PK of this drug . ABCG2 rs2231137 G > A (G/A and A/A genotypes) results in Val12Met substitution and is partly responsible for the functional impairment of ABCG2. The association of this variant with gefitinib-induced rash and the PK of gefitinib has been reported , which is consistent with our results for osimertinib. The two ABCB1 polymorphisms (rs1128503 and rs2032582) were associated with grade ≥ 2 AEs but not with the serum concentration of osimertinib. We hypothesized that ABCB1 polymorphisms may influence the tissue distribution and accumulation of osimertinib and its active metabolites, as an in vivo study suggested the involvement of ABCB1 and ABCG2 in tissue accumulation of other TKIs . Taken together, our results suggested that monitoring exposure to the parent compound, AZ5104, and AZ7550 and genotyping polymorphisms in EGFR, ABCG2 , and ABCB1 are potential new approaches for AE management. AEs can be effectively managed without dose interruption by adjusting the dose and increasing awareness about the need for AE management. The AEs we focused on in this study, such as paronychia, anorexia, and diarrhea, are not fatal but need to be managed to improve quality of life and encourage patients to continue osimertinib therapy for a long period. This study has several limitations to consider when interpreting the results. Because of the small sample size, we were unable to perform a multivariable analysis to compare the impact of multiple risk factors or check for potential confounding variables. A larger sample size is also required to evaluate the relationship between fatal or severe AEs (grade ≥ 3) and exposure to osimertinib or its active metabolites. Opportunistic collection of serum samples reduced the burden on patients but led to there being fewer concentration data points around the time of peak serum concentration (T max ) for PopPK analysis. Additional blood sampling is required to improve the accuracy of the volume of distribution estimate, but this was not essential for the purpose of this study. The exposure–efficacy relationship was not analyzed in this study because progression-free survival was not reached for many of the participants at the time of data cut-off. Although many reports have suggested that there is no relationship between osimertinib exposure and efficacy, further studies are needed to identify the optimal therapeutic window for osimertinib treatment. Our findings demonstrated for the first time that exposures to the two active metabolites of osimertinib (AZ5104 and AZ7550) were associated with AEs and may have a different impact than exposure to the osimertinib parent compound. Therefore, monitoring not only the parent compound but also the active metabolites is a potential approach for osimertinib AE management. Germline polymorphisms in EGFR (rs2293348 and rs4947492), ABCG2 (rs2231137), and ABCB1 (rs1128503 and rs2032582) were identified as potential biomarkers for predicting the severity of AEs and may help increase awareness of the importance of AE management for patients at higher risk. Below is the link to the electronic supplementary material. Supplementary Material 1 |
The roles of social norms and leadership in health communication in the context of COVID-19 | 4f8eb694-2c5a-48fc-b4e6-386dbd6f37aa | 10030439 | Health Communication[mh] | The roles of the leaders in health communication In crises that bring uncertainty and threats, people seek guidance from their leaders . Pandemics are highly stressful events in which individuals must confront uncertain and ambiguous situations . In the context of COVID-19, which has not only been a long-lasting threat to health but has also emerged as a broader-ranging challenge, triggering political and economic crises , individuals feel uncertain and stressed . In such a context, individuals usually direct their attention toward the suggestions and instructions of their leaders to reduce the impact of these negativities . Recent studies showed that the guidance of leaders is highly effective against COVID-19. For example, examined country-level mobile activity data, including 545 million unique devices that reflect social distancing in counties of the US, following governors' calls for compliance with the preventive measures. They found that governors' recommendations led to a significant reduction in mobility in counties. Leaders' influence in persuading the public to follow preventive measures is not confined to politicians or political figures. Religious leaders can also convince those with the same social identity (i.e., a religious community) to take preventive actions, regardless of whether or not their message contains religious arguments . Leaders sometimes negatively affect public health through misguidance that directs individuals towards risky behaviors. For example, after Jair Bolsonaro, the President of Brazil, explicitly denied the risks of COVID-19 and rejected the policy of isolation, the degree of compliance with the social distancing rule decreased more in regions with higher support for the government compared to those with lower support . Likewise, following the press conference of Andres Manuel Lopez Obrador, the Mexican President, in which he downplayed the severity of the pandemic, in pro-government municipalities, there was an increase in the geographic mobility of individuals, thus violating the social distancing measure . In the US, health communication styles also diverged sharply along partisan lines of leaders, and the individuals' levels of compliance with preventive measures differed in line with the endorsed leader. Donald Trump, as a Republican leader, used rhetoric that downplayed the risk of the pandemic, while Democratic leaders put much more emphasis on the dangers (e.g., ). Concordantly, several studies showed a remarkable difference between republican and democrat voters in adherence to the measures; Democrats, like their leaders, took the pandemic more seriously and took more precautions (e.g., ). Leadership is a mutual influence process between leaders and followers who are members of the same social group, as suggested by the New Psychology of Leadership (NPoL; ). provided a framework highlighting the psychological features that make a leader more effective. They proposed the Identity Leadership Model (ILM) to outline the four key aspects of leadership: Prototypicality, advancement, impresarioship, and entrepreneurship. Accordingly, prototypicality refers to the notion that leaders are seen as representative members of the ingroup (i.e., one of us) and as role models for ingroup members . Identity advancement is defined as a leader's capacity to advance not personal interests but ingroup interests and goals by enhancing the ingroup prestige . Identity impresarioship means the leader's ability to create events and structures for group members to engage in group-related practices that contribute to the smooth operation of the group and provide collective experiences that embed a sense of belonging to social identity . Lastly, entrepreneurship refers to a leader's capacity to create a shared sense of “we” and improve cohesion that allows ingroup members to feel that they are part of the same group and understand the meaning of the difference between “us” and “them” . Although all the dimensions of ILM might be closely related to effective leadership, identity entrepreneurship may be especially relevant during the pandemic, as it has the potential to reduce uncertainty and restore the personal sense of control. Entrepreneurship refers to the leadership's ability to shape and clarify members' understanding of what the ingroup represents rather than its role in defining and shaping ingroup stereotypes, norms, values, and ideals. Therefore, leaders who are perceived as identity entrepreneurs may be more successful in guiding how group members should behave in the pandemic context. Research implies that people need leaders who offer appropriate behavior models or define norms for how they should behave in a time of uncertainty (see ). Studies also showed that leaders who can craft an understanding of the group (i.e., who we are) become preeminent among the members (e.g., ). Therefore, a leader's ability to provide an understanding of “who we are” and “what we should do” might be crucial in the context of COVID-19. Redefining ingroup boundaries and shaping understandings of ingroup norms, values, and ideals can lead the way in this new ambiguous context. Thus, we assume that identity entrepreneurs would become effective leaders directing people to comply with preventive measures. Descriptive and injunctive norms in health communication Social norms provide frames of reference for making judgments about ambiguous stimuli and thus shape behavior . Since people in the current era have never experienced a context like COVID-19, and its uncertain multidimensional consequences, social norms may have reached an even more critical reference point in the process of making judgments and acting. Therefore, focusing on how social norms affect behaviors during the pandemic may contribute to the construction of effective health communication, increasing compliance with preventive measures. Social psychology has various norm conceptualizations; each can be quite different from the other. For example, the Theory of Planned Behavior (TPB) defines subjective norms, which represent the perceptions of how significant others (e.g., parents, spouse, friends) approve or not approve of engaging in a particular behavior . Subjective norms have an impact on health-related behaviors as well. In the case of the COVID-19 pandemic, for example, it was found that subjective norms predicted vaccination intentions positively . On the other hand, The Focus Theory of Normative Conduct distinguishes between descriptive and injunctive norms . While descriptive norms specify the most common and expected behaviors to be performed in particular situations, injunctive norms refer to which behaviors are acceptable or nonacceptable, approved or disapproved in a social group. Therefore, injunctive norms are related to social approval, and descriptive norms to belongingness or affiliation needs . Descriptive norms reflect the number of people who, as a rule, do or do not engage in the target behavior. On the other hand, Injunctive norms are concerned with their approval or disapproval of the target behavior. But at this point, we need to underline that injunctive norms do not correspond to the concept of the subjective norm in TPB, in the sense that injunctive norms are not limited to one's significant others' approval or disapproval. On the other side, the Social Identity Approach to norms, on which our study is also based, proposes that when specific group memberships become salient, individuals conform to the context-specific ingroup norms based on what they perceive as the prototypical ingroup beliefs, attitudes, feelings, and behaviors . These group norms can be either descriptive or injunctive. In other words, group norms are the prototypical features of the group that prescribe the appropriate attitudes and actions for group members, including descriptive and injunctive properties . Regarding this point, found that participants tended to act in accordance with their own attitude when this attitude is commonly held by the other ingroup members (i.e., when the descriptive norm is consistent with their attitude). Correspondingly, found that individuals tend to act less in accordance with their attitude when the injunctive norm is inconsistent with the attitude. Studies in the context of health behaviors have supported the assumptions of this approach (for example, see for flu vaccination). reported that a meta-analysis's results showed experimentally induced norms changes were associated with changes in intentions and behavior. Their findings also suggested that a change in norms was sufficient to change behavior, even without changes in personal attitudes and self-efficacy. Another meta-analysis on health behaviors indicated that descriptive norms had a greater impact on behavior than injunctive norms, especially when the behaviors were not socially approved, more socially motivated, and more pleasant . found that consistency between descriptive and injunctive norms boosts the effectiveness of social information. The results suggest that what is critical is not the superiority of one type over the other but the consistency between different normative message types. It was also shown that simultaneous activation of the two types of norms is most effective in bringing behavioral change. However, there is mixed experimental evidence on the impact of the interaction between descriptive and injunctive norms if opposed . That is, when incompatible descriptive and injunctive norms work simultaneously in a given situation, it becomes difficult to predict the individuals' behavior . Considering the particular context of COVID-19, this often came to the fore, especially in the health communication messages of various authorities. For example, leaders frequently highlighted that the people did not sufficiently comply with the preventive measures and violated hygiene or social distance rules. University students were accused of attending hall events, parties, and gatherings, ignoring social distancing rules . In a similar vein, mainstream and social media have published news or charts indicating the prevalence of vaccine reluctance, implying that resistance is a pervasive norm . In other words, the political and local leaders themselves emphasized that the descriptive norm points to the relative frequency of non-compliance with the precautions, yet the same leaders simultaneously presented a picture of consensus on the necessity and importance of compliance. Although these messages emphasized the seriousness of the situation, they also highlighted the presence of two opposing norms in society. Since common behavior becomes a moral reference, people may even display increased misbehaviors when observing others' inappropriate behaviors . In an experiment across nine countries, presented participants with vignettes about a hypothetical country affected by COVID-19 where empirical and normative expectations about preventive measures vary (i.e., what others do and what others approve of). They asked participants to estimate the level of compliance with preventive measures of the vignette characters and found that participants estimated the highest compliance when both expectations were congruently positive. However, estimates of compliance levels decreased if one norm type was opposed to the other. In other words, preventive measures are most adhered to when both types of social norms are consistently positive. However, comparisons between incongruent conditions provided mixed and nonsystematic results . Thus, previous studies provide no consistent or robust evidence in determining which type of norm (i.e., descriptive or injunctive) is more effective on behaviors in the case of a discrepancy between them in particular conditions (e.g., pandemics and societal crises). Regarding the COVID-19 context, even strong descriptive norms might have limited effects on vaccine tendency . Thus, it is important to study the role of the interaction between the different types of norms and leadership. Examining the possible interaction may contribute to identifying the factors that increase the effectiveness of health communication aimed at protecting public health. In addition, the existing literature is mainly based on survey data, which makes it difficult to draw clearly defined inferences that can serve as a guide in the field. Therefore, there is a need for strong experiments that examine both the effects of incongruency in different types of norms and the role of the leader in this process. The current study The success in controlling the pandemic within a country, and in turn across the world, appears to depend on a variety of social identity-related factors . For example, Akfirat et al. showed that national identification positively predicted acceptance of national vaccines and negatively predicted acceptance of Western vaccines; people's evaluations of their leaders also mediated both relationships. As the research findings imply, leadership and the social norms guiding group functioning and members' behaviors emerged as critical factors in the pandemic. In the current experimental study, we examined the roles of identity entrepreneurship of a national leader and descriptive and injunctive norms on compliance with COVID-19 preventive measures. In general, we expected that both descriptive and injunctive types of social norms and their interactions would positively affect adherence to preventive measures if the two types of norms support preventive measures and are not in conflict. We also examined which type of norms would be more effective in promoting preventive behaviors when in conflict (i.e., descriptive norms support anti-preventive behaviors, whereas injunctive norms disapprove of such behaviors and vice versa). In addition, we investigated the role of the leader (i.e., leader's entrepreneurship) as the source of the health messages. We consider entrepreneurship as an essential leadership quality in the struggle against the pandemic because the uncertainty and the crisis brought by the pandemic get confused about what to do and how to behave. Therefore, we assumed that entrepreneur leaders who can convey that the group has the required characteristics to overcome the situation would successfully induce people to comply with the COVID-19 preventive measures. Our research is based on a scenario experiment, which has been widely used in social psychology, particularly in group processes and leadership studies (see ; ; ). We employed a high-powered scenario experiment to examine the role of leaders as well as the main effects of different norm types (i.e., descriptive and injunctive) that encourage or hinder adherence to preventative measures and their interaction effects (i.e., congruency). We conducted an experiment with 2 (presence/lack of entrepreneurship of the leader) X 2 (supportive/obstructive descriptive norm) X 2 (supportive/obstructive injunctive norm) between-subject factorial design on the three dependent outcomes (i.e., compliance with preventive measures, vaccination intention, and prosocial behavior against COVID-19). Method The anonymized raw data is publicly available at https://osf.io/szvh2/ 4.1 Participants According to power analysis in G*Power software , assuming a small effect size ( d = 0.20, f = 0.10) and taking a two-tailed alpha as .05 with power at .95, we calculated that we needed a minimum of 1047 participants to detect an effect. We recruited undergraduate students from four universities who agreed to participate in exchange for a gift draw and course credits. We shared participation links with students through Qualtrics, an online data collection platform. Since all measures are forced, and any data cleaning method may violate the randomization of the experimental design, we included all participants who completed the study in the analysis. At the end of the data collection process, we reached 1057 participants (74% female) aged 18 to 59 with a mean of 22.1 ( SD = 5.02). As the respective population, the sample socio-economical levels seem normally distributed. More than half of the participants (62.1%) described themselves as left-oriented (Scale range between 0 = Extreme leftist , 10 = Extreme rightist ). Participants According to power analysis in G*Power software , assuming a small effect size ( d = 0.20, f = 0.10) and taking a two-tailed alpha as .05 with power at .95, we calculated that we needed a minimum of 1047 participants to detect an effect. We recruited undergraduate students from four universities who agreed to participate in exchange for a gift draw and course credits. We shared participation links with students through Qualtrics, an online data collection platform. Since all measures are forced, and any data cleaning method may violate the randomization of the experimental design, we included all participants who completed the study in the analysis. At the end of the data collection process, we reached 1057 participants (74% female) aged 18 to 59 with a mean of 22.1 ( SD = 5.02). As the respective population, the sample socio-economical levels seem normally distributed. More than half of the participants (62.1%) described themselves as left-oriented (Scale range between 0 = Extreme leftist , 10 = Extreme rightist ). Procedure Before the experiment, ethical approval was obtained by the Ethics Committee of the Dokuz Eylul University in Turkey. The study included the participants through the link generated on Qualtrics between 2020 and 11–18 and 2020-11-24 when there was uncertainty over the appropriate treatment and the vaccines were yet to be developed. Participants were randomly assigned to the experimental conditions, which were formed through manipulation of the norm type (descriptive and injunctive), norm content (supportive or obstructive), and leadership (presence/lack of entrepreneurship). We presented an imaginary scenario to the participants for manipulation purposes. We decided to locate the context of COVID-19 within a country outside Turkey, considering that participants' political and ideological attitudes may affect social norms and leadership traits expected to predict health precautions during the fight against COVID-19. We aimed to ensure that the country chosen as the context has no positive or negative history with Turkey and that little information about the country's political leader is available. Based on these criteria, the Republic of Cameroon was chosen for the experiment's framework, and Lejeune Mbella Mbella, (foreign minister) was presented as its head of state. Although at the time we collected the data, Cameroon's president was Paul Biya, and its prime minister was Joseph Ngute, we preferred neither of them as the political leader to be manipulated in our experiment because their first names had the possibility of reminding participants of Western countries. The names were Christian-originated and common in Western countries. Lejeune Mbella Mbella was not the real head of state, yet he was a real Cameroonian politician (foreign minister). Therefore, we decided to present him as the national political leader in our scenarios because the phonetic pronunciation of Lejeune Mbella Mbella sounded authentic, and the name was not familiar to Turkish respondents. The rest of the information was fabricated to manipulate the descriptive and injunctive norms and the leader's entrepreneurship quality. All six authors evaluated and discussed the contents of the vignettes to determine whether the norms and leadership manipulations correspond to the theoretical conceptualizations and relevant measures. At the end of the discussion sessions, we agreed that the scenarios fit the theoretical concepts. Before starting the experimental procedure, all participants read the following identical instruction, regardless of their assigned conditions: As known, COVID-19 (coronavirus disease) continues tospread worldwide. Cameroon, located in the Midwest of Africa, is also one of the countries where the pandemic has been spreading exponentially. Lejeuna Mblella Mbella, who is the current President of Cameroon, at every possible opportunity, appeals to the people of Cameroon to take protective measures against COVID-19. The rest of the text differed according to the conditions of independent variables, and participants were required to confirm that they read and understood the text. 5.1 Descriptive norm manipulation We manipulated descriptive norms by highlighting either their supportive or obstructive aspects. Supportive descriptive norms were defined as the behaviors commonly exhibited in society, and obstructive descriptive norms , as behaviors commonly non-exhibited in society. For instance, the common use of masks during the COVID-19 pandemic is a supportive descriptive norm, while the widespread absence of masks is an obstructive descriptive norm. We manipulated descriptive norms as being supportive or obstructive norms regarding the COVID-19 pandemic as follows: Supportive Descriptive Norm: Scientific research on the COVID-19 (coronavirus disease) pandemic in Cameroon revealed that the GREAT MAJORITY of Cameroonian society takes preventive measures against COVID-19 (coronavirus) seriously and complies with the preventive rules such as social distancing and being hygienic. Obstructive Descriptive Norm: Scientific research on COVID-19 (coronavirus disease) pandemic in Cameroon revealed that a GREAT MAJORITY of Cameroon society DO NOT take preventative measures against COVID-19 (coronavirus) seriously, and the rate of complying with preventive rules, such as social distancing and being hygienic, is VERY LOW. 5.2 Injunctive norm manipulation As in descriptive norm manipulation, we manipulated injunctive norms by presenting either supportive or obstructive aspects. Supportive injunctive norms refer to the behaviors approved and recommended by society and emphasize that punishment is required when these have not complied. On the other hand, obstructive injunctive norms are defined as behaviors approved by society, but in this case, non-compliance is not considered to be socially or legally punishable. That is, obstructive injunctive norms are those behaviors for which non-compliance is tolerated. For example, being friendly is a socially approved behavior, but its lack does not require active disapproval or legal sanctions. Within this scope, we manipulated supportive and obstructive injunctive norms as follows: Supportive Injunctive Norms: Research also shows that Cameroonians DO NOT tolerate those who do not comply with the rules fighting against COVID-19 (coronavirus), and they also warn each other to obey these rules. In addition, Cameroonians think that people who do not comply with the rules fighting against COVID-19 (coronavirus) should have sanctions imposed on them. Obstructive Injunctive Norms: Research also shows that Cameroonians are INDIFFERENT towards those who do not comply with the rules for fighting against COVID-19 (coronavirus). In addition, Cameroonians think that people who do not obey the rules for fighting against COVID-19 (coronavirus) should not have sanctions imposed on them. 5.3 Leaders' entrepreneurship We manipulated the identity entrepreneurship quality of the national political leader, who was in a position to direct people to comply with the preventive measures against COVID-19. We created two conditions regarding identity entrepreneurship manipulation, portraying the leader either as having entrepreneurship qualities or presenting no information about the leadership characteristics, as below: In his speeches, President Mbella, emphasizes Cameroonians as responsible, self-sacrificing, and resilient against difficulties, creating a sense of unity and integrity in the country. He also states that he believes the country will overcome the pandemic in a much shorter time than other countries because Cameroon people have these particular values and distinctive characteristics. In the control condition, we provided no information on leadership characteristics in terms of identity entrepreneurship. Descriptive norm manipulation We manipulated descriptive norms by highlighting either their supportive or obstructive aspects. Supportive descriptive norms were defined as the behaviors commonly exhibited in society, and obstructive descriptive norms , as behaviors commonly non-exhibited in society. For instance, the common use of masks during the COVID-19 pandemic is a supportive descriptive norm, while the widespread absence of masks is an obstructive descriptive norm. We manipulated descriptive norms as being supportive or obstructive norms regarding the COVID-19 pandemic as follows: Supportive Descriptive Norm: Scientific research on the COVID-19 (coronavirus disease) pandemic in Cameroon revealed that the GREAT MAJORITY of Cameroonian society takes preventive measures against COVID-19 (coronavirus) seriously and complies with the preventive rules such as social distancing and being hygienic. Obstructive Descriptive Norm: Scientific research on COVID-19 (coronavirus disease) pandemic in Cameroon revealed that a GREAT MAJORITY of Cameroon society DO NOT take preventative measures against COVID-19 (coronavirus) seriously, and the rate of complying with preventive rules, such as social distancing and being hygienic, is VERY LOW. Injunctive norm manipulation As in descriptive norm manipulation, we manipulated injunctive norms by presenting either supportive or obstructive aspects. Supportive injunctive norms refer to the behaviors approved and recommended by society and emphasize that punishment is required when these have not complied. On the other hand, obstructive injunctive norms are defined as behaviors approved by society, but in this case, non-compliance is not considered to be socially or legally punishable. That is, obstructive injunctive norms are those behaviors for which non-compliance is tolerated. For example, being friendly is a socially approved behavior, but its lack does not require active disapproval or legal sanctions. Within this scope, we manipulated supportive and obstructive injunctive norms as follows: Supportive Injunctive Norms: Research also shows that Cameroonians DO NOT tolerate those who do not comply with the rules fighting against COVID-19 (coronavirus), and they also warn each other to obey these rules. In addition, Cameroonians think that people who do not comply with the rules fighting against COVID-19 (coronavirus) should have sanctions imposed on them. Obstructive Injunctive Norms: Research also shows that Cameroonians are INDIFFERENT towards those who do not comply with the rules for fighting against COVID-19 (coronavirus). In addition, Cameroonians think that people who do not obey the rules for fighting against COVID-19 (coronavirus) should not have sanctions imposed on them. Leaders' entrepreneurship We manipulated the identity entrepreneurship quality of the national political leader, who was in a position to direct people to comply with the preventive measures against COVID-19. We created two conditions regarding identity entrepreneurship manipulation, portraying the leader either as having entrepreneurship qualities or presenting no information about the leadership characteristics, as below: In his speeches, President Mbella, emphasizes Cameroonians as responsible, self-sacrificing, and resilient against difficulties, creating a sense of unity and integrity in the country. He also states that he believes the country will overcome the pandemic in a much shorter time than other countries because Cameroon people have these particular values and distinctive characteristics. In the control condition, we provided no information on leadership characteristics in terms of identity entrepreneurship. Measures 6.1 Manipulation check We asked three questions to check the manipulations (for descriptive norm: Most Cameroonians take the measures in the fight against COVID-19 seriously and comply with the rules of social distance and hygiene ; for injunctive norm: Most Cameroonians think that sanctions should be imposed on those who do not comply with the rules against COVID-19 ; and for identity entrepreneurship: The President of Cameroon, Lejeune Mbella Mbella creates a sense of unity and integrity in the country by emphasizing the values and characteristics of the Cameroonian people in the period of combat against COVID-19 ). All three items were Likert type with 5 points (1 = Strongly disagree, 5 = Strongly agree ). 6.2 Preventive measures We required the participants to answer the dependent measures from the point of view of Cameroonians. We explicitly instructed them to put themselves in the position of Cameroonians when answering the questions. We generated nine items to measure compliance with the preventive measures related to hygiene and physical distancing (e.g., If I were a Cameroonian, I would maintain physical distance from others ). Participants answered the items on a slider from 0 ( definitely I would not ) to 100 ( definitely I would ). The results of exploratory factor analysis showed that the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy was 0.934, and Bartlett's Test of Sphericity was statistically significant ( p < .001) level, meaning that the scale was deemed suitable for factor analysis. Thus, principal component analysis results with varimax rotation yielded a single factor with an Eigenvalue of 6.99, accounting for 77.73% of the variance. Cronbach's alpha for the scale was 0.96. 6.3 Intention to be vaccinated We used a single-item measure of vaccine intention ( If a vaccine was developed, I, as a Cameroonian, … ) using a 5-Point Likert scale (1 = definitely would not get vaccinated , 5 = definitely would get vaccinated ). 6.4 Prosociality We measured prosocial behavioral tendencies using a single item. Participants were asked to imagine themselves as Cameroonian and indicated how much they would be prepared to donate to a foundation fighting against COVID-19 if they had 100 units of money. A higher value of the donation is considered to indicate a greater tendency for prosocial behavior. We used the same measures for all conditions. Manipulation check We asked three questions to check the manipulations (for descriptive norm: Most Cameroonians take the measures in the fight against COVID-19 seriously and comply with the rules of social distance and hygiene ; for injunctive norm: Most Cameroonians think that sanctions should be imposed on those who do not comply with the rules against COVID-19 ; and for identity entrepreneurship: The President of Cameroon, Lejeune Mbella Mbella creates a sense of unity and integrity in the country by emphasizing the values and characteristics of the Cameroonian people in the period of combat against COVID-19 ). All three items were Likert type with 5 points (1 = Strongly disagree, 5 = Strongly agree ). Preventive measures We required the participants to answer the dependent measures from the point of view of Cameroonians. We explicitly instructed them to put themselves in the position of Cameroonians when answering the questions. We generated nine items to measure compliance with the preventive measures related to hygiene and physical distancing (e.g., If I were a Cameroonian, I would maintain physical distance from others ). Participants answered the items on a slider from 0 ( definitely I would not ) to 100 ( definitely I would ). The results of exploratory factor analysis showed that the Kaiser–Meyer–Olkin (KMO) measure of sampling adequacy was 0.934, and Bartlett's Test of Sphericity was statistically significant ( p < .001) level, meaning that the scale was deemed suitable for factor analysis. Thus, principal component analysis results with varimax rotation yielded a single factor with an Eigenvalue of 6.99, accounting for 77.73% of the variance. Cronbach's alpha for the scale was 0.96. Intention to be vaccinated We used a single-item measure of vaccine intention ( If a vaccine was developed, I, as a Cameroonian, … ) using a 5-Point Likert scale (1 = definitely would not get vaccinated , 5 = definitely would get vaccinated ). Prosociality We measured prosocial behavioral tendencies using a single item. Participants were asked to imagine themselves as Cameroonian and indicated how much they would be prepared to donate to a foundation fighting against COVID-19 if they had 100 units of money. A higher value of the donation is considered to indicate a greater tendency for prosocial behavior. We used the same measures for all conditions. Results 7.1 Manipulation check To test the effectiveness of the manipulations, we conducted three different independent t-tests. According to the results, regarding the leader's entrepreneurship, experimental group participants ( M = 5.15, SD = 1.20) perceived the leader as significantly more entrepreneurial than control group participants ( M = 3.75, SD = 1.61), t (1055) = 15.98, p < .001, Cohen's d = 0.98, which indicates identity entrepreneurship manipulation was effective. Similarly, participants in the supportive descriptive norm condition ( M = 4.84, SD = 1.52) perceived the rate of the Cameroonians' complying with the preventive rules is significantly higher than those in the obstructive descriptive norm condition ( M = 2.17, SD = 1.48), t (1055) = 28.90, p < .001, Cohen's d = 1.78. This result also shows the success of supportive descriptive norm manipulation. Lastly, participants in the positive injunctive norm condition ( M = 4.83, SD = 1.38) perceived the rate of Cameroonians demanding sanctions against those who violated the COVID-19 preventive measures as significantly higher than those in the negative injunctive norm condition ( M = 2.17, SD = 1.55), t (1055) = 29.43, p < .001, Cohen's d = 1.81. Therefore, we concluded that our manipulations were effective. 7.2 Main analyses We conducted a three-way multivariate analysis of variance (MANOVA) to examine the effects of a leader's entrepreneurship, supportive and obstructive forms of the descriptive norm and injunctive norm, and their interactions on the combination of outcomes. Results indicated a significant main effect of the descriptive norm, Wilks' Lambda = .883, F (3, 1047) = 46.249, p < .001, η p 2 = 0.12, a significant main effect of the injunctive norm, Wilks' Lambda = 0.966, F (3,1047) = 12.430, p < .001, η p 2 = 0.03, but no significant main effect of identity entrepreneurship, Wilks' Lambda = 0.996, F (3,1047) = 1.319, p = .267. A significant interaction emerged between descriptive norm and injunctive norm, Wilks' Lambda = 0.988, F (3,1047) = 4.226, p = .006, partial η2 = 0.01. A series of follow-up univariate analyses of variance (ANOVAs) for each dependent variable were conducted to examine multivariate analyses results. 7.3 Compliance with preventive measures There was a significant main effect of the descriptive norm on compliance with preventive measures, F (1, 1049) = 127.11, p < .001, η p 2 = 0.108. Accordingly, compliance with preventive measures was significantly higher for the supportive descriptive norm condition than for the obstructive descriptive norm condition ( M difference = 17.3, SE = 1.53, p bonferroni < .001, d = 0.693). In a similar vein, there was a significant main effect of the injunctive norm, F (1, 1049) = 33.54, p < .001, η p 2 = 0.031. In the supportive injunctive norm condition, intention for compliance with preventive measures was significantly higher than in the obstructive injunctive norm condition ( M difference = 8.87, SE = 1.53, p bonferroni < .001, d = 0.357). There was also a significant interaction between descriptive and injunctive norm, F (1, 1049) = 4.58, p = .040, η p 2 = 0.004. Specifically, when the two types of norms were congruently supportive, there was the highest level of compliance with preventive measures ( M = 88.4, SE = 1.53), and this condition indicated greater compliance with preventive measures than the condition in which both norms are obstructive ( M difference = −26.26, SE = 2.17, p bonferroni < .001, d = −1.053). The next strongest compliance with measures was found in the condition where the descriptive norm is supportive ( M = 82.7, SE = 1.53), and higher compliance was indicated than for the condition of both norms being obstructive ( M difference = −20.58, SE = 2.17, p bonferroni < .001, d = −0.825). In the case where all norms were incongruent, the condition in which the descriptive norm was supportive and the injunctive norm obstructive was found more effective than the reverse condition ( M difference = 8.33 SE = 2.17, p bonferroni < .001, d = −0.334). All other results are displayed in , and interaction is depicted in . There was also a significant interaction between the leader's entrepreneurship and descriptive norm, F (1, 1049) = 6.67, p = .010, η p 2 = 0.006. When there was an entrepreneur leader, and the descriptive norm was supportive, participants indicated the highest compliance with preventive measures ( M = 86.7, SE = 1.55), and this condition resulted in higher compliance than the condition in which the descriptive norm was obstructive, and leader's entrepreneurship was lacking (i.e., control condition), M difference = −15.51 SE = 2.19, p bonferroni < .001, d = −0.623. All other interactions are presented in and depicted in . 7.4 Vaccine intention For the vaccine intention, a significant main effect emerged for both the descriptive norm, F (1, 1049) = 44.31, p < .001, η p 2 = 0.041, and the injunctive norm, F (1, 1049) = 10.56, p < .001, η p 2 = 0.010. Participants in the supportive descriptive norm condition had significantly higher vaccine intention than those in the obstructive descriptive norm condition ( M difference = −0.438, SE = 0.066, p bonferroni < .001, d = −0.410). Similarly, the supportive injunctive norm condition had higher vaccine intention than the obstructive injunctive norm group ( M difference = −0.214, SE = 0.066, p bonferroni < .001, d = −0.200). There was no significant interaction between conditions. All univariate tests can be seen in . 7.5 Prosociality Regarding prosociality, there was a significant main effect of the descriptive norm, F (1, 1049) = 31.06, p < .001, η p 2 = 0.029, and a significant main effect of the injunctive norm, F (1, 1049) = 9.79, p = .002, η p 2 = 0.009 on intention to donate to the foundation fighting against COVID-19. Participants in the supportive descriptive norm condition showed significantly higher intention to donate than participants in the obstructive descriptive norm condition ( M difference = −8.80, SE = 1.58, p bonferroni < .001, d = −0.343). Similarly, participants in the positive injunctive norm condition showed significantly higher intention to donate than participants in the negative, injunctive norm condition ( M difference = −4.94, SE = 1.58, p bonferroni = .002, d = −0.193). There was no significant interaction, and all univariate tests are presented in . Manipulation check To test the effectiveness of the manipulations, we conducted three different independent t-tests. According to the results, regarding the leader's entrepreneurship, experimental group participants ( M = 5.15, SD = 1.20) perceived the leader as significantly more entrepreneurial than control group participants ( M = 3.75, SD = 1.61), t (1055) = 15.98, p < .001, Cohen's d = 0.98, which indicates identity entrepreneurship manipulation was effective. Similarly, participants in the supportive descriptive norm condition ( M = 4.84, SD = 1.52) perceived the rate of the Cameroonians' complying with the preventive rules is significantly higher than those in the obstructive descriptive norm condition ( M = 2.17, SD = 1.48), t (1055) = 28.90, p < .001, Cohen's d = 1.78. This result also shows the success of supportive descriptive norm manipulation. Lastly, participants in the positive injunctive norm condition ( M = 4.83, SD = 1.38) perceived the rate of Cameroonians demanding sanctions against those who violated the COVID-19 preventive measures as significantly higher than those in the negative injunctive norm condition ( M = 2.17, SD = 1.55), t (1055) = 29.43, p < .001, Cohen's d = 1.81. Therefore, we concluded that our manipulations were effective. Main analyses We conducted a three-way multivariate analysis of variance (MANOVA) to examine the effects of a leader's entrepreneurship, supportive and obstructive forms of the descriptive norm and injunctive norm, and their interactions on the combination of outcomes. Results indicated a significant main effect of the descriptive norm, Wilks' Lambda = .883, F (3, 1047) = 46.249, p < .001, η p 2 = 0.12, a significant main effect of the injunctive norm, Wilks' Lambda = 0.966, F (3,1047) = 12.430, p < .001, η p 2 = 0.03, but no significant main effect of identity entrepreneurship, Wilks' Lambda = 0.996, F (3,1047) = 1.319, p = .267. A significant interaction emerged between descriptive norm and injunctive norm, Wilks' Lambda = 0.988, F (3,1047) = 4.226, p = .006, partial η2 = 0.01. A series of follow-up univariate analyses of variance (ANOVAs) for each dependent variable were conducted to examine multivariate analyses results. Compliance with preventive measures There was a significant main effect of the descriptive norm on compliance with preventive measures, F (1, 1049) = 127.11, p < .001, η p 2 = 0.108. Accordingly, compliance with preventive measures was significantly higher for the supportive descriptive norm condition than for the obstructive descriptive norm condition ( M difference = 17.3, SE = 1.53, p bonferroni < .001, d = 0.693). In a similar vein, there was a significant main effect of the injunctive norm, F (1, 1049) = 33.54, p < .001, η p 2 = 0.031. In the supportive injunctive norm condition, intention for compliance with preventive measures was significantly higher than in the obstructive injunctive norm condition ( M difference = 8.87, SE = 1.53, p bonferroni < .001, d = 0.357). There was also a significant interaction between descriptive and injunctive norm, F (1, 1049) = 4.58, p = .040, η p 2 = 0.004. Specifically, when the two types of norms were congruently supportive, there was the highest level of compliance with preventive measures ( M = 88.4, SE = 1.53), and this condition indicated greater compliance with preventive measures than the condition in which both norms are obstructive ( M difference = −26.26, SE = 2.17, p bonferroni < .001, d = −1.053). The next strongest compliance with measures was found in the condition where the descriptive norm is supportive ( M = 82.7, SE = 1.53), and higher compliance was indicated than for the condition of both norms being obstructive ( M difference = −20.58, SE = 2.17, p bonferroni < .001, d = −0.825). In the case where all norms were incongruent, the condition in which the descriptive norm was supportive and the injunctive norm obstructive was found more effective than the reverse condition ( M difference = 8.33 SE = 2.17, p bonferroni < .001, d = −0.334). All other results are displayed in , and interaction is depicted in . There was also a significant interaction between the leader's entrepreneurship and descriptive norm, F (1, 1049) = 6.67, p = .010, η p 2 = 0.006. When there was an entrepreneur leader, and the descriptive norm was supportive, participants indicated the highest compliance with preventive measures ( M = 86.7, SE = 1.55), and this condition resulted in higher compliance than the condition in which the descriptive norm was obstructive, and leader's entrepreneurship was lacking (i.e., control condition), M difference = −15.51 SE = 2.19, p bonferroni < .001, d = −0.623. All other interactions are presented in and depicted in . Vaccine intention For the vaccine intention, a significant main effect emerged for both the descriptive norm, F (1, 1049) = 44.31, p < .001, η p 2 = 0.041, and the injunctive norm, F (1, 1049) = 10.56, p < .001, η p 2 = 0.010. Participants in the supportive descriptive norm condition had significantly higher vaccine intention than those in the obstructive descriptive norm condition ( M difference = −0.438, SE = 0.066, p bonferroni < .001, d = −0.410). Similarly, the supportive injunctive norm condition had higher vaccine intention than the obstructive injunctive norm group ( M difference = −0.214, SE = 0.066, p bonferroni < .001, d = −0.200). There was no significant interaction between conditions. All univariate tests can be seen in . Prosociality Regarding prosociality, there was a significant main effect of the descriptive norm, F (1, 1049) = 31.06, p < .001, η p 2 = 0.029, and a significant main effect of the injunctive norm, F (1, 1049) = 9.79, p = .002, η p 2 = 0.009 on intention to donate to the foundation fighting against COVID-19. Participants in the supportive descriptive norm condition showed significantly higher intention to donate than participants in the obstructive descriptive norm condition ( M difference = −8.80, SE = 1.58, p bonferroni < .001, d = −0.343). Similarly, participants in the positive injunctive norm condition showed significantly higher intention to donate than participants in the negative, injunctive norm condition ( M difference = −4.94, SE = 1.58, p bonferroni = .002, d = −0.193). There was no significant interaction, and all univariate tests are presented in . Discussion In the present study, we focused on the roles of social norms and leadership in complying with the calls for preventive measures regarding health communication. Specifically, we aimed to examine the effects of descriptive/injunctive types of social norms and the leader's entrepreneurship characteristics on adherence to COVID-19 preventive measures. We also investigated the effects of conditions in which there was a conflict between the two types of norms (one supportive and the other obstructive) and their interaction with a leader's entrepreneurship. In a high-powered experiment, we found that supportive forms of descriptive and injunctive norms positively predicted compliance with preventive measures, vaccine intention, and prosocial tendencies against COVID-19. The results also showed an interaction between descriptive and injunctive norms, indicating the highest intention to comply with preventive measures occurred when there was consistency, i.e., both types of norms supported compliance with preventive measures. In addition, when the norms were incongruent, the obstructive injunctive and supportive descriptive norm conditions produced results indicating greater compliance than the supportive and descriptive norm conditions. In other words, if there is a conflict, the descriptive norm in opposition to the adherence to measures creates greater obstacles in terms of health communication aimed at individuals' compliance. As for leadership, results did not indicate any significant main effect for the leader's entrepreneurship, but there was a significant interaction between a leader's entrepreneurship and the descriptive norm, indicating that compliance results in the highest scores when the descriptive norms are supportive, and the leader emphasizing the ingroup has the required characteristics. The current findings can potentially contribute to the existing literature in various respects. To begin with, the current study provides a comprehensive perspective on the power of injunctive and descriptive norms. Descriptive norms were found as more effective than injunctive norms regarding compliance with COVID-19 preventive measures. Descriptive norms accentuate the behaviors of the majority and reinforce the idea that “this behavior is right” . Since a pandemic involves intense uncertainty and creates the need for guidance, it is appropriate to determine one's course of action heuristically by observing the behavior of those around . Previous studies also supported this viewpoint and showed that descriptive norms reduce uncertainty and are especially functional in unfamiliar and ambiguous contexts (e.g., ). Concerning the effects of norm valence, we found that when individuals were exposed to supportive descriptive and injunctive norms rather than obstructive norms, they were more likely to comply with COVID-19 preventive measures. A study based on evolutionary predisposition by showed that obstructive descriptive norms exerted a stronger influence over people than supportive descriptive norms. That is, participants were influenced more by behaviors avoided by others than by behaviors performed by others. Our findings support that individuals are more likely to adopt preventive behaviors when exposed to what the majority generally does rather than does not. To our knowledge, the present study is the first high-powered experiment that examines how the effects of descriptive and injunctive norms differ when they are incongruent in health communication. The extant literature indicated mixed results when there is a conflict between descriptive and injunctive norms. However, there seems to be inadequate research on the complexity of social norms and the effects of this complexity. The interaction between the descriptive and injunctive norms, especially when conflicting, appears as neglected in previous literature . We found that the intention to comply with COVID-19 preventive measures in the condition the descriptive norm supports, but the injunctive norm obstructs preventive measures is higher than in the reverse condition. This finding is consistent with previous accounts that when a situation is ambiguous, people seek proper behaviors in their environment and tend to take action preferred by a large majority . According to , the concept of “repeat assembly” operates at many levels in human evolution, not just in terms of genes. That is, characteristics or behaviors that are “repeated” by a number of individuals and groups promote the adaptiveness of group life and increase survival chances. On the other hand, the interaction between descriptive and injunctive norms was not significant in either vaccine intention or prosocial tendency. This difference may stem from the dependent variables related to the contemporary context and personal/collective outcomes. That is, at the time of this study, widespread vaccination had not yet begun, and vaccine development and approval were in progress. In other words, the social norm did not have time to produce severe effects on an issue that had yet to materialize fully. In addition, vaccination intention may be strengthened more by emphasizing individual risk rather than social benefits. Social norms include group-level aspects in terms of their content, and in the norm manipulations, this study did not emphasize the individual health risk in the necessity of complying with the precautions. Some preliminary findings also implied that emphasizing individual health risks is more effective in increasing vaccination rates (e.g., ; ). As for a prosocial tendency, this variable prioritizes social interests rather than individual gains compared to other positive behaviors against COVID-19. indicated that people are prone to credit descriptive norms for their own risk-taking but recommend others act in parallel with injunctive norms. Decision-making studies emphasize that real behavioral measures of prosociality (i.e., incentivized games including real decisions with real money) can provide more accurate results and that different cognitive biases affect responses to the two measures . Therefore, we encourage future studies to use real behavioral measures when examining prosocial behaviors. Another crucial finding of the current study is a significant interaction between the leader's entrepreneurship and the supportive descriptive norm, indicating that compliance with preventive measures resulted in the highest scores. Such a finding in the COVID-19 context was unsurprising regarding the notion that a leader's identity entrepreneurship characteristic is likely to meet uncertainty reduction needs and might function to restore people's sense of control by shaping and clarifying their understanding of ingroup stereotypes, norms, values, and ideals (e.g., ). Besides, people might consider a leader successful and effective because, in this experimental condition, the majority were presented as complying with the preventive measures as demanded. From this point, the reason for the non-significant interaction between obstructive conditions becomes clearer. That is, the participants might have perceived the leader's identity entrepreneurship effort to be ineffective had they not observed that most followed their leader's call. We did not find a significant main effect of a leader's entrepreneurship on outcomes. There may be two factors underlying this result: leadership process and health context. Firstly, we considered only the entrepreneurship dimension of the ILM, which deals with the leadership processes as a whole, in terms of four different but interrelated leadership qualities. Employing dimensions of ILM other than entrepreneurship might be more effective in guiding people to healthy behavior. Group members evaluate leaders on multiple aspects; therefore, examining leadership dimensions by considering their interaction may provide a more comprehensive perspective rather than relying on a single dimension. On the other hand, the level of ingroup identification might have considerable effects on evaluating the leader and compliance with the leader's guidance . These notions might also explain why no significant main effect of the leader's entrepreneurship was found on any outcomes. Secondly, the lack of effectiveness of the identity entrepreneurship quality of the national political leader even compared to the control condition, may be related to the health theme. Previous studies show that experts, scientists, and health authorities are more effective in guiding people to healthy behaviors than national, political, and government leaders regarding healthy behaviors (e.g., ; ), including compliance with preventive measures against COVID-19 (e.g., ). This may relate to people's greater trust in health authorities in matters that directly concern their health and require expert knowledge. Concordantly, previous studies showed that individuals are skeptical of health messages from government authorities , finding them less persuasive than scientists regarding vaccination decisions . Thus, we encourage future studies that compare public trust in health experts and national political leaders. We consider that the present study has valuable outcomes for practical implications in health communication. It seems crucial to take more care over-delivering health messages to the masses in ambiguous and crisis times, as negative portrayals of the majority, such as selfish, panic-buying consumers, might create the opposite effect that was intended. To encourage people to comply with the preventive measures in daily life, authorities or media often emphasize that it is not sufficient to advocate behavioral adherence to measures and that the role of goodwill should be considered when emphasizing the seriousness of the risks. It should also be noted that not only in crisis but also at other times when a new health behavior is required to be adopted within a society, it seems insufficient to explain the importance of this behavior or the risks that may arise in its absence. Rather, the behavior in question should be presented as widely accepted, performed, and supported by others. Also, although we did not find a significant main influence regarding political leadership, the interaction effect of leadership and descriptive norm implies that positive framing would be more effective if voiced by leaders. As a result, it seems critical to be aware of the importance of positively framed descriptive and injunctive norms in public messages. Rather than obstructive content, positive and supportive social norms should be emphasized, especially if mass behavior change is intended. Our study builds on a tradition of imaginary scenario experiments used frequently in social psychology. . This methodology has both advantages and disadvantages. Implementing this method allows for describing a more realistic and naturalistic context. Therefore, the responses given while imagining oneself in a hypothetical situation were usually reported as very similar to the actual behaviors in the real world . Again, the findings of the scenario experiments were consistent with the results of the survey studies and the laboratory experiments (e.g., ). As a result, there seem to be enough reasons to consider our method valid. And yet, to control confounding factors such as political orientations or ideologies, presenting Cameroon rather than Turkey as the experimental framework does not automatically guarantee that participants will distance themselves from the norms operating in Turkey and from its political leader. In this regard, we must emphasize that replication of this study, whether through surveys or laboratory experiments, would supplement our findings. The current study is not free of other shortcomings. First, we were not able to include group identification variables in the study. Previous studies indicated that group identification is a crucial factor that affects various group-level phenomena. For example, group members' prototypicality perceptions of the leader vary according to their identification levels, which eventually affects the endorsement of group leaders . However, the experimental design we implemented in this study did not allow us to measure participants' national identification levels. On the other hand, our scenarios were fiction, and participants were required to imagine themselves as Cameroonians. The phrase “If I were Cameroonian … " was used so that participants could hypothetically categorize themselves in a national group, i.e., Cameroonians, which constituted the base for our manipulations. Yet, such a formulation might have been influenced by the participants' identifications with their actual national identities. When interpreting the results, this point should be taken into account. Second, we couldn't examine the roles of background variables such as personal attitudes, traits, values, emotions (fear), perceived risks, and demographics. Each of these variables influences whether particular health behaviors are performed . In particular, the fact that the study sample consists of university students is a significant limitation. Since university students may have psychological motivations and personality traits that differ from the general population , it is necessary to be careful about the representativeness of the inferences based on the findings presented by the data consisting of the student sample . Although implementing an experimental procedure (random assignment to the experimental conditions) may eliminate the confounding effects of these individual-level variables, we strongly encourage studies based on representative samples. The corresponding author is responsible for ensuring that the descriptions are accurate and agreed by all authors.; Serap Akfırat participated to the process of designing and conducting the study, data gathering, deciding appropriate analyses, writing the whole parts of the manuscript; Fatih Bayrak participated to the process of designing and conducting the study, data gathering, deciding appropriate analyses, analyzing the data, writing the whole parts of the manuscript; Emir Üzümçeker participated to the process of designing and conducting the study, data gathering, deciding appropriate analyses, analyzing the data, writing the whole parts of the manuscript.; Tolga Ergiyen participated to the process of designing and conducting the study, deciding appropriate analyses, analyzing the data, the writing the “Method” and “Results” sections of the manuscript; Taylan Yurtbakan Tolga Ergiyen participated to the process of designing and conducting the study, deciding appropriate analyses, writing the “Method” part section of the manuscript; Mete Sefa Uysal participated to the process of designing and conducting the study, deciding appropriate analyses, writing the “Introduction” and “Discussion” parts of the manuscript. |
Claude Bernard and life in the laboratory | 0b647371-64cc-4e63-aa02-e6d9ca16b27f | 10030445 | Physiology[mh] | Physiological laboratories properly speaking are an invention of the second half of the nineteenth century. They became only gradually established along with the process of the institutional separation of physiology from anatomy that European universities underwent in the second third of that century (Schiller, ). In his book Wissenschaft in der Maschinenstadt. Emil Du Bois-Remyond und seine Laboratorien in Berlin , Sven Dierig has described the coming into being of a physiological laboratory in detail and situated it in the context of nineteenth century industrialization and urbanization. In the first decades of the nineteenth century, medical students specializing in physiological questions usually still had to conduct their experiments – if they experimented at all — in a corner of their private homes and chambers. Dierig has aptly called this phase “chamber physiology” – “ Stubenphysiologie ” (Dierig, ). In the second third of the century, anatomists pursuing physiological questions, such as, e.g., Johannes Müller in Berlin, who had followed anatomist Karl Asmund Rudolphi on his chair at Berlin’s University in 1833, and who was also in charge – as it then used to be — of the anatomical collection, had to find a space in the collection rooms, if he aimed at doing experiments besides his teaching duties (Lohff, ; Otis, ). It is only in the last third of the century that physiologists started to be endowed with institutes suited for their special purposes, of which the “Physiologische Anstalt” of Carl Ludwig in Leipzig was one of the first and most richly equipped (Lenoir, ). It opened its doors in 1869 and became a paradigm of a physiological working space for Europe as well as for overseas, especially the United States. In his later addresses, Claude Bernard did not miss to point to these developments in neighboring Germany, claiming that his own country would be the legitimate place to house such institutions, being the country in which modern physiology had originated. In his course on general physiology that he pronounced from 1869 to 1876 at the Muséum d’Histoire Naturelle, where he had succeeded Marie-Jean-Pierre Flourens after the latter’s death in 1868, and that later appeared as the famous Leçons sur les phénomènes de la vie communs aux animaux et aux végétaux , he gave the following brief description of Ludwig’s lab: “I put before your eyes the plan of one of these laboratories, that of Leipzig directed by Ludwig, which is sketched here in the beautiful report of M. Wurtz: I wished that on this example you can see the richness of these scientific installations, of which we do not even have an idea in France. In the basement there are the cellars, the rooms kept at constant temperature, the distillation apparatuses, a steam engine that sustains movement everywhere, the atelier of a mechanic attached to the laboratory, a magazine for the chemical products, a hospital for the dogs. – In the first floor there are the laboratories of vivisection, those of physics and biological chemistry, the chambers for handling mercury, the rooms for the microscopes, for the histological studies, for the spectroscope, and so on. – The library, the rooms for the courses, the lodging of the professor, are part of the same building; add a horse stable, an aviary, numerous aquaria, and we have enumerated the essential parts of this magnificent establishment erected for science” (Bernard, , pp. 15–16). And in the report on general physiology that he wrote on the instigation, in 1867, of Victor Duruy, then Minister of Public Instruction, he stated: “One can see that, for instance, if Germany has the biggest share in contemporary publications in the science of physiology, this is a result of the fact that the cultural means of experimental physiology are considerable and well instituted there.” And he did not forget to add: “French physiology reclaims nothing else than what is only fair to give it; it never lacked the physiological genius” (Bernard , pp. 210–211). However, Paris had to wait until 1880 for its first full-fledged state-supported physiological laboratory to be built in town for Etienne-Jules Marey who had taken over the chair of Pierre Flourens at the Collège de France in 1869. Claude Bernard’s first laboratory experience dates back to 1840, when he entered the Collège de France as an assistant and subsequently became “préparateur” to François Magendie in 1841. Magendie had been appointed professor for physiology and general pathology at the Collège de France in 1836. He had managed to install a modest laboratory at the Collège, the first and only laboratory space of physiology in France at the time. Paul Bert, a student of Bernard and his successor on the chair for general physiology at the Sorbonne, on the occasion of Bernard’s death in 1878, once succinctly voiced this turning point in Bernard’s life as follows: “As soon as he had put his foot in the laboratory at the Collège de France, his path was traced. The famous physiologist’s [Magendie’s] intrepid, albeit a bit disordered, experimentation, his implacable critique, his skepticism that included his own discoveries, made a deep, creative impression on the spirit of the young Claude Bernard” (Bert, , p. 17). It is here that over the course of the following decade Bernard developed his own style as a physiological experimenter. In his magisterial work on Claude Bernard and Animal Chemistry , Frederic Holmes has meticulously followed, on the basis of his early laboratory notebooks, the “investigative pathway” of Bernard that led him, among other things, to the discovery of the glycogenic function of the animal liver. Bernard did not have a circumscribed research question to begin with. Rather, his early steps in the laboratory appear as “a bit disordered” like those of his teacher, to use the wording of Paul Bert. Among the questions he approached were animal combustion, a theme virulent in physiology since the heroic days of Antoine Lavoisier. With respect to the latter, in his late Leçons still, Bernard confessed: “The discovery of respiratory combustion by Lavoisier has been, one can say, more fruitful than the majority of anatomical discoveries” (Bernard, , p. 7). Furthermore, efforts to determine the nutritive value of gelatin were imposed on Magendie’s laboratory by a commission of the government that sought to determine cheap nutrients that were hoped to be able to replace the expensive meat nutrition in Parisian hospitals and asylums. He worked on dynamic aspects of blood circulation as well as on the uptake and the course of drugs through the body. And not least, his interest was attracted by the decomposition of nutrients, such as sugars, in the animal’s organs (Holmes, ; Rheinberger, , Ch. 6). It was an amalgam of questions on which Bernard tried his hands in these early years, without conspicuous results to begin with. The ensemble of trials was, however, held together by the conviction that if one wanted to know what was specifically physiological, or biological, about these processes, one had to experiment on and with the living animal body. This conviction was to be condensed, in the course of the 1850s, into his notion of “milieu intérieur” (d’Hombres, /2016). This went against the grain of physiological chemistry of the time, as particularly represented by Jean-Baptiste Dumas in France and Justus von Liebig in Germany that was essentially based on the chemical analysis of metabolic products, a comparison of inputs and outputs of sorts. Bernard was not at all opposed to chemical analysis; on the contrary, he frequently sought the help of chemists for secondary analysis of materials he had recovered from his experimental animals. However, he held on to the maxim that it was foremost necessary to identify the intermediate products of metabolism that were specifically formed in the living body’s internal environment. Accordingly, he sought to develop an arsenal of procedures that would allow him to introduce substances into the body in a localized manner and to retrieve substances such as gastric juice in a similarly localized way and according to a particular time regime. He thus established what became known as in vivo experimental physiology. This first decade of his experimental activities led him to the unsettlement of a deep-seated dogma of contemporary physiology: the conviction that the synthesis of organic materials was restricted to plants, whereas their decomposition was the privilege of animals. He had started out, among other strands of work he pursued, to analyze the decomposition of sugar in the animal body; what he ended up with was an investigation of the opposite: the synthetic glycogenic function of the liver. I think that we can draw two lessons from Claude Bernard’s laboratory regime as it developed at the Collège de France in the course of the 1840s, and that would also guide Bernard’s experimental efforts in later years (Grmek, , ). The first is that Bernard developed a style of experimentation that we can address as specifically exploratory. As the laboratory notebooks of the time between 1840 and 1850 document (Holmes, , ), to stay with this example, Bernard conducted several lines of experimentation in parallel. He developed what Gerald Geison, in a review of Holmes’ book on Claude Bernard and Animal Chemistry , called “Bernard’s real methods”, contrasting them with the “idealized prescriptions for scientific research” laid down in his Introduction to the Study of Experimental Medicine (1865) (Geison, , p. 639). He thus could explore the range of action of the experimental gadgets and devices that he had developed, and he could adapt and modify them accordingly. He could also use the findings of one series of experiments and implement them, if possible and feasible, in another one. In this way, he was able to develop an experimental network that again and again led him to surprising revelations. He was also ready, if with the actual state of knowledge and equipment he ended up in an impasse, to temporarily abandon a particular stream of experimentation and switch to another one, until a new finding or a novel experimental device allowed him to switch over again to the one he had put aside for the time being. It is an experimental strategy that works, so it appears, particularly well if a new field of research is in the process of being opened and delineated. The history of the sciences knows of a number of variants of that strategy (Rheinberger, ; Holmes, ). Bernard himself, at the end of his career, summarized it in the following words: “In order to tackle experimental criticism and to get to know all the conditions of a physiological phenomenon, one must have groped for a long time, one must have been deceived a thousand times, one must have, in a word, grown old in experimental practice” (Bernard, , p. 19). The second lesson is that Bernard came to be convinced that minutiae matter and that it is decisive to develop a sensorium for them. In the introduction to his classic published in 1865, the Introduction , he put it as follows: “In scientific investigation, the most minute procedures are of the highest importance. The lucky choice of an animal, an instrument construed in a specific manner, the use of one reagent instead of another, will often suffice to resolve general questions of a highest order. […] In one word, the greatest scientific truths have their roots in the details of experimental investigation that constitute in some sort the soil in which these truths develop.” And he concluded: “One must have been brought up and have lived in laboratories in order to grasp the full importance of all these details of procedure in investigation that are often ignored and despised by those false scholars who call themselves generalizers” (Bernard, , p. 44). This feeling for the details, the attention to the contingencies going along with the intricacies of experimentation was a lesson that Bernard learned early on and that he held in highest esteem throughout his later career. In his Preface to the Introduction , François Dagognet has summarized Bernard’s attitude toward experimentation and his receptivity with regard to the empirical detail as follows: “Theory, in turn, may play the role of a springboard as well as that of an obstacle. One discovers less with ideas than against them, because the learned scholar must become a “doubter” who tries to understand the language of nature beyond the interpretation that, aiming to reveal it, also disguises and confuses it. Yes, one must question life, but one has to be attentive above all to the answers it gives in the margins or even outside the expected discourse” (Dagognet, , p. 18). Bernard walked these margins, and they revealed to him their riches over and over again. Throughout his life, Bernard had to work under precarious laboratory conditions. Since the beginning of the 1840s, he had François Magendie’s modestly equipped laboratory at the Collège de France at his disposal. Before that time, Magendie himself had to give “private courses in experimental physiology,” as Bernard euphemistically called these sessions. And he recollects: “It is only after 1830 that, having become professor of medicine at the Collège de France, he was able to establish the largely insufficient laboratory that there still exists […]” (Bernard, , p. 10). The list of people with whom Bernard cooperated throughout the 1840s, among them Jean Poiseuille, Jean-Baptiste-Rozier Coze, and Charles-Louis Barreswil, amply demonstrates that he depended on the equipment of others, above all chemists, in particular for the secondary analysis of the bodily products that he retrieved from his in vivo experiments (Holmes, ). The situation did not much improve when he became, in 1855, the successor of Magendie. “After having become successor of Magendie at the Collège de France, I have fought like him against the lack of resources. […] Back then, the laboratory at the Collège de France was the only one that existed” (Bernard, , p. 11). No laboratory facilities were attached to the chair of general physiology at the Sorbonne either, to which Bernard had been appointed in 1854. Consequently, he later moved to the chair left vacated by the death of Pierre Flourens at the Muséum d’histoire naturelle in 1868, where he could finally install another laboratory of physiology in Paris. There, he proudly introduced his opening lecture in the summer semester of 1870 with the following words: “The introduction of general physiology in the renowned establishment that houses the natural sciences, and the creation of a laboratory annexed to the chair mark a notable progress in teaching experimental physiology.” And he looked back, not without reproach to the responsible authorities: “This completely modern science that originated in France under the fruitful impulse of Lavoisier, Bichat, Magendie, etc., was until now, it must be said, left practically without encouragement, whereas in the neighboring countries, in contrast, it received considerable sustainment” (Bernard, , p. 1). And then, he undergirded the importance of laboratory work, not only for teaching, but in particular for research: “Finally, the majority of scientific questions are resolved by the invention of an adequate tool: he who discovers a new procedure, a new instrument, often achieves more for experimental physiology than the deepest-minded philosopher or the most powerful generalizer” (Bernard, , p. 12). With a few interspersed exceptions, Bernard’s personal laboratory experience remains in the background of the Introduction to Experimental Medicine . Overall and in general, the Introduction depicts a rationalized image of the laboratory process at large and of experimentation in particular, an image that is in accordance with contemporary philosophy of science. The papers that Bernard left document that he studied, excerpted, and commented upon, among others, the philosophy of Auguste Comte, one of the leading positivist philosophers of the time (Bernard, ). Bernard operates with the traditional categories of induction and deduction, proof and counter-proof, of truth and error, hypothesis and fact, determinism and indeterminism. I will not add here to the abundant literature about the Introduction . It is one of the rare cases where historians, philosophers and scientists crossed paths and shared their reflections on the modern sciences again and again. Instead, I will restrict myself to Bernard’s remarks on the specificities of a physiological laboratory. The second part of the Introduction contains a long chapter devoted to experimental considerations particularly concerning the manipulation of living animals. As already said, Bernard founded his way of practicing physiology on experimentation in vivo. It was clear for him that physiology had to use the means and procedures of physics and of chemistry, if only “with a great number of inherent difficulties” (Bernard, , p. 145) for he assumed that the physico-chemical processes, i.e. their “determinism,” were the same in the non-living and in the living world. On the other hand, there was something irreducibly special about living beings: “In one word, biology has its own problem and its specific point of view; it only makes use of the help and borrows the methods from the other sciences, but not their theories” (Bernard, , p. 144). Bernard stayed with this non-reducible difference and its epistemological consequences, which he came to attribute to the organism’s peculiar “internal organic environment,” to which life was inextricably linked (Bernard, , Part II, Chap. 1, § 3). His position gave rise to an endless quarrel about whether and if, to what extent, he can be regarded as a “vitalist” (Virtanen, ; Canguilhem, , , ; Bange, ). I do not pursue this debate here; rather, what appear to me to be important are two things that connect to his vivisectionist approach. The first point is that Bernard was convinced that this non-reducible difference, that is, the specific way organisms made use of the physico-chemical determinism, as he called it, could only be grasped in the living body itself. He had a hunch that the specificity of the living body’s internal environment hung together with its use of ferments, of which, however, knowledge was still scarce in the 1860 and 1870s. “We have thus to consider, besides the physico-chemical conditions indispensable for the manifestation of life, the special evolutionary physiological conditions which are the quid proprium of the biological science” (Bernard, , p. 149). Experimentation in vivo appeared thus as a logical consequence of this conviction. The second point, neither to be neglected, is that Bernard was working in the tradition of medical physiology, that is, with the human body in mind. Experimentation on humans, and on the living human body in particular, was, however, utterly restricted. Experimentation on higher animals, as practiced by Bernard, thus appeared as a viable option to get as near as possible to the human body in terms of experimental analysis. Keeping this reasoning in mind, it is not by chance if Bernard claimed that an adequate physiological laboratory had to be of a complex nature. “The laboratory of the medical physiologist must be the most complicated of all laboratories, since he has to experiment on the phenomena of life, which are the most complex of all natural phenomena” (Bernard, , p. 199). In his later Lectures on the Phenomena of Life Common to Animals and Plants , and being now offered the opportunity to establish a laboratory of his own, he differentiated this claim further and specified it at the same time: “The laboratory of the physiologist is necessarily complex in view of the complexity of the phenomena that are studied here. It has naturally to be arranged for three different orders of work: 1. the work of vivisection ; 2. the physico-chemical works; 3. the anatomical-histological works” (Bernard, , p. 16). The internal structure of the physiological laboratory was not the only point of concern for Bernard, however. Another point of attention and reflection was the relation between the laboratory of the experimental physiologist and the clinic. He took care not to confound the two: “The hospital, or better, the ward, is not the laboratory of the physician, as is often believed; the latter is, as we have said already, only his field of observation […]” (Bernard, , p. 205). And he explains this relation further: “The physician who is eager to merit this designation in a scientific sense must, having left the hospital, go into his laboratory, and it is here that he must try, by way of experiments on animals, to account for what he has observed on his patients, be it with respect to the mechanism of the illness, be it with regard to the action of drugs, or be it concerning the origin of the morbid lesions of the organs or tissues” (Bernard, , p. 206). And finally, Bernard does not forget to count the library among the knowledge spaces between which the practitioner of scientific medicine has to move. “The libraries could also be considered to form part of the laboratory of the scientist and the medical experimenter.” But he immediately adds his caveat : “This is, however, only the case if he reads in order to know and control by nature the experiences or theories of his predecessors, and not in order to find in the books the ready-made opinions that would dispense him from working […]” (Bernard, , p. 199). The working environment of the experimental physiologist is thus tripartite in a double sense. Internally, there is a partition and circulation of knowledge between vivisection, physico-chemistry, and anatomy/histology. And the experimental laboratory itself is part of another triple: the laboratory, the clinic, and the library. Cahier rouge If we understand, with Mirko Grmek , Bernard’s laboratory notes between 1850 and 1860, also called the Cahier rouge , as a part of his intellectual preparation for the composition of his later book on Experimental Medicine , we can nevertheless observe an opposite tendency. Whereas in the Introduction , Bernard positions himself as a theoretician, or philosopher, or epistemologist of the experiment; in the Cahier rouge , he writes as a practitioner immersed in his daily work. Whereas in the Introduction , he generally follows what could be called an ‘idea first’ protocol, in the Notebook , an ‘idea follows observation’ protocol prevails. It can of course be argued, and rightly so, that all the elements of Bernard’s discourse on experimentation to be identified in the notebook can be found in the introduction as well. It all depends, however, on the relative weight attributed to them in the unpublished notes and in the published chef d’oeuvre . As a rule, in the Introduction , Bernard avoids, with a few exceptions, to point to what he calls “groping experiments” ( expériences de tâtonnement ) or “experiments for the sake of seeing if” ( expériences pour voir ) and qualifies them as being inferior and belonging to “a science in its infancy” (Bernard, , p. 50). The general tenor of the book, however, is characteristically captured by the following succinct and unequivocal, reassuring statement: “It is the idea that constitutes, as we shall see, the starting point or the primum movens of all scientific reasoning, and it is the idea that is likewise the goal in the aspiration of the scientific spirit toward the unknown” (Bernard, , p. 56). In the Cahier rouge , whose character conveys something like the intimacy of a laboratory confession not destined for a public readership, the tendency is the other way around. Here, the poetological aspect of the laboratory prevails (Rheinberger, ; Sattar, ). What is presented in a tamed form and between the lines in the Introduction , here it finds its spontaneous expression. But this also implies that from the perspective of the Notebook , we can read the Introduction with new eyes. In his introduction to the Cahier rouge , Mirko Grmek has rightly stressed the double face of these notes. On the one hand, they contain reflections that Bernard himself subsumed under the heading “scientific philosophy.” On the other hand, the majority of entries consist of spontaneous considerations concerning experiments to be made. Mostly, the two kinds of entries do not appear to be directly related to each other. They are likely to be the result of what Bernard himself described with the following words in the Cahier : “The ideas develop themselves spontaneously in the mind, and if one lets oneself go with them, one is like a man at the window who regards people passing […]. It does not require any effort; it is even charming” (Bernard , p. 89). Grmek has concluded that “we are confronted with a thought that is in the process of being formed, and not a finished thought” (Grmek, , p. 13). And he summarized: “The experimental protocols let us see how the laboratory of Claude Bernard functioned; the book bound in a red envelope, on its part, allows us to enter into the interior laboratory of his thoughts” (Grmek, , p. 17). Let us first give a brief impression of Bernard’s remarks on what he calls “scientific philosophy.” Under the heading “Ideas to develop,” we can read: “[…] State that in fact one never makes a discovery by looking for it directly […] Science proceeds by way of revolutions, not by way of pure and simple additions […]” (Bernard , p. 149). Accordingly, he describes his own scientific itinerary as follows: “I came to the field of science on a devious route, and I have rid myself of rules by launching myself across fields, something that others would probably not have dared to do. But I think that in physiology this has not been bad, because it has led me to new insights” (Bernard , pp. 128–129). Consequently, in these notes, Bernard sings the praises of ignorance. The following statement sounds like a confession: “I am not a materialist. – I am not a vitalist either. – The vitalists claim; the materialists claim the opposite. – I say: I claim nothing, I know nothing; it’s the truth, and it is this state of ignorance in which I am that allows me to make hypotheses, to poetize, to indulge in my feelings and to follow my nature” (Bernard , p. 118). Later on in the Notebook , he stresses once again: “There is a certain pleasure in ignoring, because then the imagination can work” (Bernard , p. 157). One is reminded here of Bernard’s later notebook Philosophie , which ends on the following note: “If [man] needs to know, he no less needs to ignore in order to aim at knowing. If man knows everything, he will be annihilated. As Pascal says, man is made for searching after the truth and not for its possession” (Bernard, , p. 43). The first task of the experimenter appears thus to be keeping oneself open for the unexpected. After glossing over some of these more general reflections, insofar as they concern life in the laboratory, let us now have a brief look into the parts of the Notebook that concern themselves with Bernard’s sketches of experiments to be pursued or carried out in the future. Most conspicuously, they reveal to us an aspect of Bernard’s laboratory regime, or style of experimentation that he obviously practiced over and over again. We could, for want of a better expression, call it experimenting as difference making, differential experimentation, or searching for robust differences. We could also call it comparative experimentation, as Bernard himself does later in the Introduction under the title “Comparative Experimentation” (Bernard, , pp. 181–185). A quick overall count reveals at least a dozen places in the Cahier rouge , where Bernard conceived of and vividly described experiments directed at exposing differences that he hoped would serve as the starting points for further experimentation. Let us have a brief look at one of them, which Bernard put under the heading of “comparative digestion.” Here we read: “In order to know the role of the intestinal fluids, one must not let them always act in isolation from each other, as is the usual practice, but successively one after another and in their proper order, on the three classes of nutritive substances. Thus: 1 0 saliva , on salivated or non-salivated nutrients; 2 0 gastric juice , on salivated nutrients with and without gastric juice; 3 0 bile , on nutrients treated with saliva and gastric juice with and without bile […]” (Bernard , p. 40–41). And he continued the list with the sap of the pancreas, of the intestines, of the appendix, and so on. We could also call this regime of experimentation the ‘principle of with and without,’ or the “principle of exhaustion” as I have called it elsewhere (Rheinberger, ). Whereas the experimental principle just described can be seen as characterizing a certain way of doing experiments in general, we find, in the Cahier rouge , also remarks on the specificity of physiological experimentation, in contrast to experiments in chemistry or in physics. Mostly, experiments that follow these particular advices are simply described and not commented on. There are two entries, however, where Bernard gets explicit. In the first, under the heading of “A Physiological Principle,” he gives a reason for preferring experimentation on lower organisms over that on higher organisms: “The lower the organization, the more variety there is in less unity […] The higher the organization, the more variety there is in more unity […] It follows that experiments should be, as far as possible, conducted on lower organisms” (Bernard , p. 53). Although Bernard is best known for his experiments on mammals, he also experimented extensively with frogs and even with yeast. The second entry is of a more technical nature and concerns the principle of experimentation in physiology itself. “Experimentation in physiology is effected by ablation or by isolation . 1 0 by ablation , one seeks to see the trouble one produces in the ensemble (cutting the nerve). 2 0 by isolation , one seeks to see the organ function independently of the ensemble. Examples: isolated muscles; sublingual gland.” And then he adds, without having announced it, a third point: “3 0 by exaggeration of the function of the organ; section of nerves, liver, spleen […]” (Bernard , p. 172). These are three strategies of biological experimentation that actually have continued to be of relevance to this day. Ablation addresses the fact that by abolishing a biological entity – knocking it out — one can learn in which function it is implicated. Isolation is a strategy that pertains not only to biology, but it acquires a particular meaning here: Since organs, or cellular processes, are interconnected in bodies and in cells, it is often not easy to discern which function correlates with which structure. Isolation can help to solve this problem. Exaggeration , finally, points to a phenomenon that occupied Bernard in particular in his numerous investigations of the nervous system with its various subsystems. He found that he could not only ablate nervous actions, but in depressing one, he found that he could enhance another and make it easier to be grasped. We see thus that the Cahier rouge is a rich source of insight into Bernard’s experimental thinking and vivid description of the way he practiced it in his physiological laboratory setting. Perhaps a most surprising aspect of these notes is Claude Bernard’s repeated allusion to the arts and the way he tries to connect these two realms of human creativity, the sciences and the arts. Early on in the Notebook , he states: “The human spirit proceeds in the same way in all its productions. Everywhere, in music, painting, discourses of all kinds, in the sciences as well as in the arts, there is one and the same principle for the presentation of their objects. And it is this [common] part which constitutes the artist .” An additional remark is to be read as a reminder to himself for working out “the art of the sciences, considered in their exposition” (Bernard , p. 36). Later in the Cahier , we find the following statement: “One says: This is a beautiful creation, an inspiration. An artist never knows how he arrives at things [in advance]. Even so, a scientist does not know how to find things. Once found, however, one reasons and one applies; but one needs the starting point, one needs to find where one does no longer know, for one always needs premises, and they are unknown” (Bernard , p. 135). The “art of investigation” or “art of experimental research,” as he called it in the Introduction (Bernard, , pp. 35, 39, 42), that is, the characterization of research as akin to an artistic process – and the other way around, looking at the arts as a research process, we might add — appears to have been constantly on Bernard’s mind. In the Introduction , it would take the following form. In the first chapter of its first part, he deals with the various definitions of observation and experimentation, respectively. Starting with a slogan attributed to Georges Cuvier – “the observer listens to nature, the experimenter interrogates it and forces it to reveal itself” —, he remarks that as soon as one descends into experimental practice, this clear-cut distinction starts to become blurry. And he explains: “This results, it appears to me, from having confounded the art of investigation that researches and states facts, with the art of reasoning that treats them logically in pursuit of the truth. But in investigation, one can have an activity of the mind and of the senses at the same time, be it for making observations or for doing experiments” (Bernard, , pp. 34–35). “Here,” in the investigative process itself, he continues a few pages later, “we can no longer distinguish the observer from the experimenter by the nature of the research procedures applied” (Bernard, , p. 42). The space of investigation, that is, defies our efforts to achieve clean distinctions, and the image of science as derived from the space of representation, where logical reasoning prevails, cannot stand in for an appropriate characterization of the research process. This paper aimed at tracing, in a selected number of – published as well as unpublished — writings, what Claude Bernard had to say about and how he experienced the physiological laboratory, including the experimental work carried out in these spaces that were only emerging during his lifetime. The focus on the laboratory was meant to make more tangible and to develop a nuanced picture of Bernard’s epistemological position. He was staunchly opposed to metaphysics. But he also intuitively felt that in his equally staunch defense of scientific knowledge, he had to avoid a metaphysical position himself. His epistemological conviction is beautifully expressed in the following sentences with which I would like to end my parcours. They sound like an echo from Immanuel Kant’s Critique of Pure Reason : “We are looking for the laws of the phenomena, that is, for what is stable, invariable, permanent, eternal in them. […] We represent things in an abstractive manner in order to circumvent difficulties. Things are not rigorously such in nature, but we are obliged to conceive them in this way, and to say that one needs procedures that are nearer to the nature of things means nothing, because we do not know the nature of things, and these procedures need only to be in accord with the nature of our mind” (Bernard , p. 58). |
Can we design the next generation of digital health communication programs by leveraging the power of artificial intelligence to segment target audiences, bolster impact and deliver differentiated services? A machine learning analysis of survey data from rural India | ce338d47-3d5f-4fb2-a35b-69ad60012d1c | 10030469 | Health Communication[mh] | Digital health solutions have the potential to address critical gaps in information access and service delivery, which underpin high mortality. Mobile health communication programmes, which provide information directly to beneficiaries, are among the few examples of digital health solutions to have scaled widely in a range of settings. Historically, these solutions have been designed as ‘blunt instruments’—providing the same content, with the same frequency, using the same digital channel to large target populations. While this approach has enabled solutions to scale, it has contributed to variability in their reach and impact, due in part to differences in women’s access to and use of mobile phones, particularly in low-income and middle-income countries. Despite near ubiquitous ownership of mobile phones at a household level, a growing body of evidence suggests that there is a substantial gap between men and women’s ownership, access to and use of mobile phones. In India, there is a 45% gap between women’s reported access to a phone and ownership at a household level. Variations in the size of the gap have been observed across states and urban/rural areas, and by sociodemographic characteristics, including education, caste and socioeconomic status. Among women with reported access to a mobile phone, the gender gap further persists in the use of mobiles, in part because of patriarchal gender norms and limited digital skills. Collectively, these gender gaps underscore the need to consider inequities in phone access and use patterns when designing and implementing direct to beneficiary (D2B) mobile health communication programmes. Kilkari, designed and scaled by BBC Media Action in collaboration with the Ministry of Health and Family Welfare, is India’s largest D2B mobile health information programme. When BBC Media Action transitioned Kilkari to the national government in April 2019, it had been implemented in 13 states and reached over 10 million women and their families. Evidence on the programme’s impact from a randomised control trial conducted in Madhya Pradesh, India, between 2018 and 2021, suggests that across study arms, Kilkari was associated with a 3.7% increase in modern reversible contraceptive use (RR: 1.12, 95% CI: 1.03 to 1.21, p=0.007), and a 2.0% decrease in the proportion of males or females sterilised since the birth of the child (RR: 0.85, 95% CI: 0.74 to 0.97, p=0.016). The programme’s impact on contraceptive use, however, varied across key population subgroups. Among women exposed to 50% or more of the Kilkari content as compared with those not exposed, differences in reversible method use were greatest for those in the poorest socioeconomic strata (15.8% higher), for those in disadvantaged castes (12.0% higher), and for those with any male child (9.9% higher). Kilkari’s overall and varied impact across beneficiary groups raises important questions about whether the differential targeting of women and their families might lead to efficiency gains and deepen impact. In this manuscript, we argue that to maximise reach, exposure and deepen impact, the future design of mobile health communication solutions will need to consider the heterogeneity of beneficiaries, including within husband–wife couples, and move away from a one-size-fits all model towards differentiated programme design and delivery. Drawing from husbands’ and wives’ survey data captured as part of a randomised controlled trial (RCT) of Kilkari in Madhya Pradesh India, we used a three-step process involving K-Means clustering and Least Absolute Shrinkage and Selection Operator (Lasso) regression to segment couples into distinct clusters. We then assess differences in health behaviours across respondents in both study arms of the RCT. Findings are anticipated to inform future efforts to capture data and refine methods for segmenting beneficiary populations and in turn optimising the design and delivery of mobile health communication programmes in India and elsewhere globally. Kilkari program overview Kilkari is an outbound service that makes weekly, stage-based, prerecorded calls about reproductive, maternal, neonatal and child health (RMNCH) directly to families’ mobile phones, starting from the second trimester of pregnancy until the child is 1 year old. Kilkari is comprised of 90 min of RMNCH content sent via 72 once weekly voice calls (average call duration: 1 min, 15 s). Approximately 18% of cumulative call content is on family planning; 13% on child immunisation; 13% on nutrition; 12% on infant feeding; 10% on pregnancy care; 7% on entitlements; 7% on diarrhoea; 7% on postnatal care; and the remainder on a range of topics including intrapartum care, water and sanitation, and early childhood development. BBC Media Action designed and piloted Kilkari in the Indian state of Bihar in 2012–2013, and then redesigned and scaled it in collaboration with the Ministry of Health and Family Welfare between 2015 and 2019. Evidence on the evaluation design and programme impact are reported elsewhere. Setting Data used in this analysis were collected from four districts of the central Indian state of Madhya Pradesh as part of the impact evaluation of Kilkari described elsewhere. Madhya Pradesh (population 75 million) is home to an estimated 20% of India’s population and falls below national averages for most sociodemographic and health indicators. Wide differences by gender and between urban and rural areas persist for wide range of indicators including literacy, phone access and health seeking behaviours. Among men and women 15–49 years of age, 59% of women (78% urban and 51% rural) were literate as compared with 82% of men in 2015–2016. Among literate women, 23% had 10 or more years of schooling (44% urban and 14% rural). Despite near universal access to phones at a household level, only 19% of women in rural areas and 50% in urban had access to a phone that they themselves could use in 2015. Among pregnant women, over half (52%) of pregnant women received the recommended four antenatal care (ANC) visits in urban areas as compared with only 30% in rural areas. Despite high rates of institutional delivery (94%) in urban areas, only 76% of women in rural areas reported delivering in a health facility in 2015. These disparities underscore the population heterogeneity within and across Madhya Pradesh. Sample population The samples for this study were obtained through cross-sectional surveys administered between 2018 and 2020 to women (n=5095) with access to a mobile phone and their husbands (n=3842) in four districts of Madhya Pradesh. At the time of the first survey (2018–2019), the women were 4–7 months pregnant; the latter survey (2019–2020) reinterviewed the same women at 12 months post partum. Their husbands were only interviewed once, during the latter survey round. The surveys spanned 1.5 hours in length. In this analysis, modules on household assets and member characteristics; phone access and use, including observed digital skills (navigate interactive voice response (IVR) prompts, give a missed call, store contacts on a phone, open SMS, read SMS) were used to develop models. Data on practice for maternal and child health behaviours, including infant and young child feeding, family planning, pregnancy and postpartum care were used to explore the differential impact of Kilkari across clusters but not used in the development of clusters. Approach to segmentation presents a framework used for developing homogenous clusters of men and women in four districts of rural Madhya Pradesh India. describes the steps undertaken at each point in the framework in detail. We started with data elements collected on phone access and use as well as population sociodemographic characteristics collected as part of a cross-sectional survey described elsewhere. Unsupervised learning was undertaken using K-Means cluster and strong signals were identified. Strong signals were defined as variables that had at least a prevalence of 70% in one or more clusters and differed from another cluster by 50% or more. For example, 6% of men own a smart phone in Cluster 1, 88% in Cluster 2 and 75% in Cluster 3. Therefore, having a smart phone can be considered as a strong signal. Additional details are summarised in . Once defined, we then explored differences in healthcare practices across study clusters among those exposed and not exposed to Kilkari within each cluster. Box 1 Stepwise process for developing and refining a machine learning approach for population segmentation Data collected from special surveys like the couple’s dataset used here are relatively smaller in terms of sample size but large with regard to the number of data elements available. In such high-dimensional data, there are many irrelevant dimensions which can mask existing clusters in noisy data, making more difficult the development of effective clustering methods. Several approaches have been proposed to address this problem. They can be grouped into two categories: static or adaptive dimensionality reduction , including principal components analysis and subspace clustering consisting on selecting a small number of original dimensions (features) in some unsupervised way or using expert knowledge so that clusters become more obvious in the subspace. In this study, we combined subspace clustering using expert knowledge and adaptive dimensionality reduction to find subspace where clusters are most well separated and well defined. Therefore, as part of subspace clustering, we chose to start with couples’ survey data, including variables related to sociodemographic characteristic, phone ownership, use and literacy . Emergent clusters were overlapping. We decided to use men’s survey data on phone access and use as a starting point. Step 1. Defining variables which characterise homogenous groups Analyses started with a predefined set of data elements captured as part of a men’s cross-sectional survey including sociodemographic characteristics and phone access and use. K-Means clustering was used to identify clusters and the elbow method was used to define the optimal number of clusters. Strong signals were then identified. Variables which had at least a prevalence of 70% in one or more clusters and differed from another cluster by 50% or more were considered to have a strong signal. Step 2. Model strengthen through the identification and addition of new variables Once an initial model was developed drawing from the predefined set of data from the men’s survey and strong signals were identified, we reviewed available data from the combined dataset (data from the men’s survey and women’s survey). Signal strength was used as an outcome variable or target in a linear regression with L1 regularisation or Lasso regression (Least Absolute Shrinkage and Selection Operator). Regularisation is a technique used in supervised learning to avoid overfitting. Lasso regression adds absolute value of magnitude of coefficient as penalty term to the loss function. The loss function becomes: L o s s = E r r o r ( y , y ) + α ∑ i = 1 N | ω i | where ω i are coefficients of linear regression y = ω 1 x 1 + ω 2 x 2 + … + ω N x N + b . Lasso regression works well for selecting features in very large datasets as it shrinks the less important features of coefficients to 0. Merged women’s survey and men’s survey data were used as predictors for the regression, excluding variables related to heath knowledge and practices. We ended up with a sample of 3484 rows and 1725 variables after data preprocessing. Step 3. Refining clusters using supervised learning We then reran K-Means clustering with three clusters (K=3) using important features selected by Lasso regression. This methodology was used to refine the clusters and subsequently identify new strong signals. After step 3 was conducted, we repeated step 2, and kept on iteratively repeating step 2 and 3 until there was no gain in strong signals. Data preparation and results formatting have been conducted in R V.4.1.1, K-Means clustering has been performed in Python V.3.8.5. 10.1136/bmjopen-2022-063354.supp1 Supplementary data Patient and public involvement Patients were first engaged on identification in their households as part of a household listing carried out in mid/late 2018. Those meeting eligibility criteria were interviewed as part of the baseline survey, and ultimately randomised to the intervention and control arms. Prior to the administration of the baseline, a small number of patients were involved in the refinement of survey tools through qualitative interviews, including cognitive interviews, which were carried out to optimise survey questions, including the language and translation used. Finalised tools were administered to patients at baseline and endline, and for a subsample of the study population, additional interviews carried out over the phone and via qualitative interviews between the baseline and endline surveys. Unfortunately, because travel restrictions associated with COVID-19, findings were not disseminated back to community members. K-Means algorithm As part of steps 1 and 3, K-Means algorithms were used . We chose to use K-Means algorithm because of its simplicity and speed to handle large dataset compared with hierarchical clustering. A K-Means algorithm is one method of cluster analysis designed to uncover natural groupings within a heterogeneous population by minimising Euclidean distance between them. When using a K-Means algorithm, the first step is to choose the number of clusters K that will be generated. The algorithm starts by selecting K points randomly as the initial centres (also known as cluster means or centroids) and then iteratively assigns each observation to the nearest centre. Next, the algorithm computes the new mean value (centroid) of each cluster’s new set of observation. K-Means reiterates this process, assigning observations to the nearest centre. This process repeats until a new iteration no longer reassigns any observations to a new cluster (convergence). Four metrics have been used for the validation of clustering: within cluster sum of squares, silhouette index, Ray-Turi criterion and Calinski-Harabatz criterion. Elbow method was used to find the right K (number of clusters). is a chart showing the within-cluster sum of squares (or inertia) by the number of groups (k value) chosen for several executions of the algorithm. Inertia is a metric that shows how dissimilar the members of a group are. The less inertia there is, the more similarity there is within a cluster (compactness). The main purpose of clustering is not to find 100% compactness, it is rather to find a fair number of groups that could explain with satisfaction a considerable part of the data (k=3 in this case). Silhouette analysis helped to evaluate the goodness of clustering or clustering validation . It can be used to study the separation distance between the resulting clusters. The silhouette plot displays a measure of how close each point in one cluster is to points in the neighbouring clusters. This measure has a range of [−1, 1]. Silhouette coefficients near+1 indicate that the sample is far from the neighbouring clusters. A value of 0 indicates that the sample is very close to the decision boundary between two neighbouring clusters and negative values indicate that those samples might have been assigned to the wrong cluster. shows that choosing three clusters was more efficient than four for the data from the available surveys for two reasons: (1) there were less points with negative silhouettes and (2) the cluster size (thickness) was more uniform for three groupings. Other criteria used to evaluate quality of clustering are obtained by combining the ‘within-cluster compactness index’ and ‘between-cluster spacing index’. Calinski-Harabatz criterion is given by: C ( k ) = T r a c e ( B ) ( n − k ) T r a c e ( W ) ( k − 1 ) and Ray-Turi criterion is given by r ( k ) = d i s t a n c e ( W ) d i s t a n c e ( B ) , where B is the between-cluster covariance matrix (so high values of B denote well-separated clusters) and W is the within-cluster covariance matrix (so low values of W correspond to compact clusters). They both ended up with same conclusions that three clusters were the best choice for the data we had. gives different metrics used and values obtained for various clusters. Kilkari is an outbound service that makes weekly, stage-based, prerecorded calls about reproductive, maternal, neonatal and child health (RMNCH) directly to families’ mobile phones, starting from the second trimester of pregnancy until the child is 1 year old. Kilkari is comprised of 90 min of RMNCH content sent via 72 once weekly voice calls (average call duration: 1 min, 15 s). Approximately 18% of cumulative call content is on family planning; 13% on child immunisation; 13% on nutrition; 12% on infant feeding; 10% on pregnancy care; 7% on entitlements; 7% on diarrhoea; 7% on postnatal care; and the remainder on a range of topics including intrapartum care, water and sanitation, and early childhood development. BBC Media Action designed and piloted Kilkari in the Indian state of Bihar in 2012–2013, and then redesigned and scaled it in collaboration with the Ministry of Health and Family Welfare between 2015 and 2019. Evidence on the evaluation design and programme impact are reported elsewhere. Data used in this analysis were collected from four districts of the central Indian state of Madhya Pradesh as part of the impact evaluation of Kilkari described elsewhere. Madhya Pradesh (population 75 million) is home to an estimated 20% of India’s population and falls below national averages for most sociodemographic and health indicators. Wide differences by gender and between urban and rural areas persist for wide range of indicators including literacy, phone access and health seeking behaviours. Among men and women 15–49 years of age, 59% of women (78% urban and 51% rural) were literate as compared with 82% of men in 2015–2016. Among literate women, 23% had 10 or more years of schooling (44% urban and 14% rural). Despite near universal access to phones at a household level, only 19% of women in rural areas and 50% in urban had access to a phone that they themselves could use in 2015. Among pregnant women, over half (52%) of pregnant women received the recommended four antenatal care (ANC) visits in urban areas as compared with only 30% in rural areas. Despite high rates of institutional delivery (94%) in urban areas, only 76% of women in rural areas reported delivering in a health facility in 2015. These disparities underscore the population heterogeneity within and across Madhya Pradesh. The samples for this study were obtained through cross-sectional surveys administered between 2018 and 2020 to women (n=5095) with access to a mobile phone and their husbands (n=3842) in four districts of Madhya Pradesh. At the time of the first survey (2018–2019), the women were 4–7 months pregnant; the latter survey (2019–2020) reinterviewed the same women at 12 months post partum. Their husbands were only interviewed once, during the latter survey round. The surveys spanned 1.5 hours in length. In this analysis, modules on household assets and member characteristics; phone access and use, including observed digital skills (navigate interactive voice response (IVR) prompts, give a missed call, store contacts on a phone, open SMS, read SMS) were used to develop models. Data on practice for maternal and child health behaviours, including infant and young child feeding, family planning, pregnancy and postpartum care were used to explore the differential impact of Kilkari across clusters but not used in the development of clusters. presents a framework used for developing homogenous clusters of men and women in four districts of rural Madhya Pradesh India. describes the steps undertaken at each point in the framework in detail. We started with data elements collected on phone access and use as well as population sociodemographic characteristics collected as part of a cross-sectional survey described elsewhere. Unsupervised learning was undertaken using K-Means cluster and strong signals were identified. Strong signals were defined as variables that had at least a prevalence of 70% in one or more clusters and differed from another cluster by 50% or more. For example, 6% of men own a smart phone in Cluster 1, 88% in Cluster 2 and 75% in Cluster 3. Therefore, having a smart phone can be considered as a strong signal. Additional details are summarised in . Once defined, we then explored differences in healthcare practices across study clusters among those exposed and not exposed to Kilkari within each cluster. Box 1 Stepwise process for developing and refining a machine learning approach for population segmentation Data collected from special surveys like the couple’s dataset used here are relatively smaller in terms of sample size but large with regard to the number of data elements available. In such high-dimensional data, there are many irrelevant dimensions which can mask existing clusters in noisy data, making more difficult the development of effective clustering methods. Several approaches have been proposed to address this problem. They can be grouped into two categories: static or adaptive dimensionality reduction , including principal components analysis and subspace clustering consisting on selecting a small number of original dimensions (features) in some unsupervised way or using expert knowledge so that clusters become more obvious in the subspace. In this study, we combined subspace clustering using expert knowledge and adaptive dimensionality reduction to find subspace where clusters are most well separated and well defined. Therefore, as part of subspace clustering, we chose to start with couples’ survey data, including variables related to sociodemographic characteristic, phone ownership, use and literacy . Emergent clusters were overlapping. We decided to use men’s survey data on phone access and use as a starting point. Step 1. Defining variables which characterise homogenous groups Analyses started with a predefined set of data elements captured as part of a men’s cross-sectional survey including sociodemographic characteristics and phone access and use. K-Means clustering was used to identify clusters and the elbow method was used to define the optimal number of clusters. Strong signals were then identified. Variables which had at least a prevalence of 70% in one or more clusters and differed from another cluster by 50% or more were considered to have a strong signal. Step 2. Model strengthen through the identification and addition of new variables Once an initial model was developed drawing from the predefined set of data from the men’s survey and strong signals were identified, we reviewed available data from the combined dataset (data from the men’s survey and women’s survey). Signal strength was used as an outcome variable or target in a linear regression with L1 regularisation or Lasso regression (Least Absolute Shrinkage and Selection Operator). Regularisation is a technique used in supervised learning to avoid overfitting. Lasso regression adds absolute value of magnitude of coefficient as penalty term to the loss function. The loss function becomes: L o s s = E r r o r ( y , y ) + α ∑ i = 1 N | ω i | where ω i are coefficients of linear regression y = ω 1 x 1 + ω 2 x 2 + … + ω N x N + b . Lasso regression works well for selecting features in very large datasets as it shrinks the less important features of coefficients to 0. Merged women’s survey and men’s survey data were used as predictors for the regression, excluding variables related to heath knowledge and practices. We ended up with a sample of 3484 rows and 1725 variables after data preprocessing. Step 3. Refining clusters using supervised learning We then reran K-Means clustering with three clusters (K=3) using important features selected by Lasso regression. This methodology was used to refine the clusters and subsequently identify new strong signals. After step 3 was conducted, we repeated step 2, and kept on iteratively repeating step 2 and 3 until there was no gain in strong signals. Data preparation and results formatting have been conducted in R V.4.1.1, K-Means clustering has been performed in Python V.3.8.5. 10.1136/bmjopen-2022-063354.supp1 Supplementary data Analyses started with a predefined set of data elements captured as part of a men’s cross-sectional survey including sociodemographic characteristics and phone access and use. K-Means clustering was used to identify clusters and the elbow method was used to define the optimal number of clusters. Strong signals were then identified. Variables which had at least a prevalence of 70% in one or more clusters and differed from another cluster by 50% or more were considered to have a strong signal. Once an initial model was developed drawing from the predefined set of data from the men’s survey and strong signals were identified, we reviewed available data from the combined dataset (data from the men’s survey and women’s survey). Signal strength was used as an outcome variable or target in a linear regression with L1 regularisation or Lasso regression (Least Absolute Shrinkage and Selection Operator). Regularisation is a technique used in supervised learning to avoid overfitting. Lasso regression adds absolute value of magnitude of coefficient as penalty term to the loss function. The loss function becomes: L o s s = E r r o r ( y , y ) + α ∑ i = 1 N | ω i | where ω i are coefficients of linear regression y = ω 1 x 1 + ω 2 x 2 + … + ω N x N + b . Lasso regression works well for selecting features in very large datasets as it shrinks the less important features of coefficients to 0. Merged women’s survey and men’s survey data were used as predictors for the regression, excluding variables related to heath knowledge and practices. We ended up with a sample of 3484 rows and 1725 variables after data preprocessing. We then reran K-Means clustering with three clusters (K=3) using important features selected by Lasso regression. This methodology was used to refine the clusters and subsequently identify new strong signals. After step 3 was conducted, we repeated step 2, and kept on iteratively repeating step 2 and 3 until there was no gain in strong signals. Data preparation and results formatting have been conducted in R V.4.1.1, K-Means clustering has been performed in Python V.3.8.5. Patients were first engaged on identification in their households as part of a household listing carried out in mid/late 2018. Those meeting eligibility criteria were interviewed as part of the baseline survey, and ultimately randomised to the intervention and control arms. Prior to the administration of the baseline, a small number of patients were involved in the refinement of survey tools through qualitative interviews, including cognitive interviews, which were carried out to optimise survey questions, including the language and translation used. Finalised tools were administered to patients at baseline and endline, and for a subsample of the study population, additional interviews carried out over the phone and via qualitative interviews between the baseline and endline surveys. Unfortunately, because travel restrictions associated with COVID-19, findings were not disseminated back to community members. As part of steps 1 and 3, K-Means algorithms were used . We chose to use K-Means algorithm because of its simplicity and speed to handle large dataset compared with hierarchical clustering. A K-Means algorithm is one method of cluster analysis designed to uncover natural groupings within a heterogeneous population by minimising Euclidean distance between them. When using a K-Means algorithm, the first step is to choose the number of clusters K that will be generated. The algorithm starts by selecting K points randomly as the initial centres (also known as cluster means or centroids) and then iteratively assigns each observation to the nearest centre. Next, the algorithm computes the new mean value (centroid) of each cluster’s new set of observation. K-Means reiterates this process, assigning observations to the nearest centre. This process repeats until a new iteration no longer reassigns any observations to a new cluster (convergence). Four metrics have been used for the validation of clustering: within cluster sum of squares, silhouette index, Ray-Turi criterion and Calinski-Harabatz criterion. Elbow method was used to find the right K (number of clusters). is a chart showing the within-cluster sum of squares (or inertia) by the number of groups (k value) chosen for several executions of the algorithm. Inertia is a metric that shows how dissimilar the members of a group are. The less inertia there is, the more similarity there is within a cluster (compactness). The main purpose of clustering is not to find 100% compactness, it is rather to find a fair number of groups that could explain with satisfaction a considerable part of the data (k=3 in this case). Silhouette analysis helped to evaluate the goodness of clustering or clustering validation . It can be used to study the separation distance between the resulting clusters. The silhouette plot displays a measure of how close each point in one cluster is to points in the neighbouring clusters. This measure has a range of [−1, 1]. Silhouette coefficients near+1 indicate that the sample is far from the neighbouring clusters. A value of 0 indicates that the sample is very close to the decision boundary between two neighbouring clusters and negative values indicate that those samples might have been assigned to the wrong cluster. shows that choosing three clusters was more efficient than four for the data from the available surveys for two reasons: (1) there were less points with negative silhouettes and (2) the cluster size (thickness) was more uniform for three groupings. Other criteria used to evaluate quality of clustering are obtained by combining the ‘within-cluster compactness index’ and ‘between-cluster spacing index’. Calinski-Harabatz criterion is given by: C ( k ) = T r a c e ( B ) ( n − k ) T r a c e ( W ) ( k − 1 ) and Ray-Turi criterion is given by r ( k ) = d i s t a n c e ( W ) d i s t a n c e ( B ) , where B is the between-cluster covariance matrix (so high values of B denote well-separated clusters) and W is the within-cluster covariance matrix (so low values of W correspond to compact clusters). They both ended up with same conclusions that three clusters were the best choice for the data we had. gives different metrics used and values obtained for various clusters. Sample characteristics summarise the sample characteristics by cluster for men and women interviewed. and presents select characteristics with ‘strong signals’ for each cluster. Cluster 1 (n=1408) constitutes 40% of the sample population and was comprised of men and women with low levels of digital access and skills . This cluster included the poorest segment of the sample population: 36% had a primary school or lower education and 40% were from a scheduled tribe/caste. Most men owned a feature (68%) or brick phone (22%); used the phone daily (89%); and while able to navigate IVR prompts (91%), only 29% were able to perform all of the five basic digital skills assessed. Women in this cluster similarly had lower levels of education as compared with other clusters (39% have primary school or less education); used feature (74%) or brick phones (8%); and had low digital skills (15% were able to perform the five basic digital skills assessed). Cluster 2 (n=666; 19% of sample population) is comprised of men with mid-level and women with low digital access and skills. In this cluster, 75% of men owned smartphones, 65% were observed to successfully perform the five basic digital skills assessed and 36% could perform a basic internet search. Men in Cluster 2 also self-reported accessing videos from YouTube (84%) and using WhatsApp (95%). Women in Cluster 2 had low phone ownership; nearly half of women reported owning a phone (38% owned a phone and did not share it, 22% owned and shared a phone)—findings which contradict their husbands’ reports of 0% women’s phone ownership. Only 21% of women in this cluster were observed to be able to successfully perform the five basic digital skills assessed. However, based on husband’s reporting of their wives’ digital skills, 36% of women could search the internet, 37% used WhatsApp, and 66% watched shows on someone else’s phone. Cluster 3 (n=1410; 40% of sample population) is comprised of couples with high-level digital access among both husbands and wives, and lower-level digital skill among wives . An estimated 67% of couples in this cluster were in the richer or richest socioeconomic strata, while 71% of men and 58% of women had high school or higher levels of education. Men in this cluster reported using the internet frequently (85%), were observed to own smart phones (88%) and had high levels of digital skills: 77% could perform the five basic digital skills assessed, 77% could perform a basic internet search and 85% could send a WhatsApp message. When reporting on their wife’s digital access and skills, all men in this cluster reported that their wives’ owned phones (100%), but often shared these phones with their husbands (77%), using them to watch shows (75%), search the internet (55%) or use WhatsApp (57%). However, a much lower level of women interviewed in this cluster were observed to own Feature (57%) or Smart phones (34%) and had moderate digital skills with 41% being able to successfully perform the five basic digital skills assessed. Differences in health outcomes by cluster presents differences in health outcomes by Cluster among those exposed and not exposed to Kilkari as part of the RCT in Madhya Pradesh. Findings suggest that the greatest impact was observed among those exposed to Kilkari in Cluster 2, which is the smallest cluster identified (19% of the sample population). Among this population, differences between exposed and not exposed were 8% for reversible modern contraceptive methods, 7% for immunisation at 10 weeks, 3% for immunisation at 9 months, and 4% for timely immunisation at 10 weeks and 9 months. Additionally, an 8% difference between exposed and not exposed was observed for the proportion of women who report being involved in the decision about what complementary foods to give child. Among Clusters 1 and 3, improvements were observed among those exposed to Kilkari for a small number of outcomes. In Cluster 1, those exposed to Kilkari had a 3%–4% higher rate of immunisation at 6, 10, 14 weeks than those not exposed. In both Clusters 1 and 3, the timeliness of immunisation improved at 10 weeks among those exposed. No improvements were observed for use of modern reversible contraception in either cluster. summarise the sample characteristics by cluster for men and women interviewed. and presents select characteristics with ‘strong signals’ for each cluster. Cluster 1 (n=1408) constitutes 40% of the sample population and was comprised of men and women with low levels of digital access and skills . This cluster included the poorest segment of the sample population: 36% had a primary school or lower education and 40% were from a scheduled tribe/caste. Most men owned a feature (68%) or brick phone (22%); used the phone daily (89%); and while able to navigate IVR prompts (91%), only 29% were able to perform all of the five basic digital skills assessed. Women in this cluster similarly had lower levels of education as compared with other clusters (39% have primary school or less education); used feature (74%) or brick phones (8%); and had low digital skills (15% were able to perform the five basic digital skills assessed). Cluster 2 (n=666; 19% of sample population) is comprised of men with mid-level and women with low digital access and skills. In this cluster, 75% of men owned smartphones, 65% were observed to successfully perform the five basic digital skills assessed and 36% could perform a basic internet search. Men in Cluster 2 also self-reported accessing videos from YouTube (84%) and using WhatsApp (95%). Women in Cluster 2 had low phone ownership; nearly half of women reported owning a phone (38% owned a phone and did not share it, 22% owned and shared a phone)—findings which contradict their husbands’ reports of 0% women’s phone ownership. Only 21% of women in this cluster were observed to be able to successfully perform the five basic digital skills assessed. However, based on husband’s reporting of their wives’ digital skills, 36% of women could search the internet, 37% used WhatsApp, and 66% watched shows on someone else’s phone. Cluster 3 (n=1410; 40% of sample population) is comprised of couples with high-level digital access among both husbands and wives, and lower-level digital skill among wives . An estimated 67% of couples in this cluster were in the richer or richest socioeconomic strata, while 71% of men and 58% of women had high school or higher levels of education. Men in this cluster reported using the internet frequently (85%), were observed to own smart phones (88%) and had high levels of digital skills: 77% could perform the five basic digital skills assessed, 77% could perform a basic internet search and 85% could send a WhatsApp message. When reporting on their wife’s digital access and skills, all men in this cluster reported that their wives’ owned phones (100%), but often shared these phones with their husbands (77%), using them to watch shows (75%), search the internet (55%) or use WhatsApp (57%). However, a much lower level of women interviewed in this cluster were observed to own Feature (57%) or Smart phones (34%) and had moderate digital skills with 41% being able to successfully perform the five basic digital skills assessed. presents differences in health outcomes by Cluster among those exposed and not exposed to Kilkari as part of the RCT in Madhya Pradesh. Findings suggest that the greatest impact was observed among those exposed to Kilkari in Cluster 2, which is the smallest cluster identified (19% of the sample population). Among this population, differences between exposed and not exposed were 8% for reversible modern contraceptive methods, 7% for immunisation at 10 weeks, 3% for immunisation at 9 months, and 4% for timely immunisation at 10 weeks and 9 months. Additionally, an 8% difference between exposed and not exposed was observed for the proportion of women who report being involved in the decision about what complementary foods to give child. Among Clusters 1 and 3, improvements were observed among those exposed to Kilkari for a small number of outcomes. In Cluster 1, those exposed to Kilkari had a 3%–4% higher rate of immunisation at 6, 10, 14 weeks than those not exposed. In both Clusters 1 and 3, the timeliness of immunisation improved at 10 weeks among those exposed. No improvements were observed for use of modern reversible contraception in either cluster. Evidence on the impact of D2B mobile health communication programmes is limited but broadly suggests that they can cost-effectively improve some reproductive, maternal and child health practices. This analysis aims to serve as a proof of concept for segmenting beneficiary populations to support the design of more targeted mobile health communication programmes. We used a three-step iterative process involving a combination of supervised and unsupervised learning (K-Means clustering and Lasso regression) to segment couples into distinct clusters. Three identifiable groups emerge each with differing health behaviours. Findings suggest that exposure the D2B programme Kilkari may have a differential impact among the clusters. Implications for designing future digital solutions Findings demonstrate that the impact of the D2B solution Kilkari varied across homogenous clusters of women with access to mobile phones and their husbands in Madhya Pradesh. Across delivery channels, our analysis indicates that mobile health communication could not be effectively delivered to husbands and wives in Cluster 1 using WhatsApp, because smartphone ownership and WhatsApp use in this cluster are negligible. IVR, on the other hand, could be used to reach couples in Cluster 1, but reach is likely to be sporadic because of high levels of phone sharing with others (78% among men and 57% among women). On the other hand, WhatsApp and YouTube are likely to be effective digital channels for communicating with both husbands and wives in Cluster 3, where most men and women own or use smartphones and WhatsApp. Beyond delivery channels, study findings raise a number of important learnings for content development as well as optimising beneficiary reach and exposure. The creative approach to content created for Cluster 3, where 40% of women are from the richest socioeconomic status and only 17% have never been to school or have a primary school education or less, would need to be very different from the creative approach to content created for Cluster 1, where 53% have a poorest or poorer socioeconomic status, and 39% have never been to school or have a primary school education or less. Similarly, this analysis adds to qualitative findings and provides important insights into how gender norms related to women’s use of mobile phones may effect reach and impact. While few (13–15%) husbands indicated that ‘adults’ need oversight to use mobile phones, men’s perceptions varied when asked about specific use cases. Across all Clusters, nearly half of husbands indicated that their wives needed permission to pick up phone calls from unknown numbers—an important insight for IVR programmes which may make outbound calls without prewarning to beneficiaries. In Clusters 1 and 2, 25% and 29% of husband’s, respectively, report that their wives need permission to answer calls from health workers—as compared with 15% in Cluster 3. While restrictions on SMS and WhatsApp were lower than making or receiving calls, these channels are less viable given women’s limited access to smartphones, low literacy and digital skills. Overall, men’s perceptions on the restrictions needed on the receipt and placement of calls by women was lower for Cluster 3. However, despite the relative wealth of beneficiaries in Cluster 3 (67% were in the richer or richest socioeconomic strata), 48% of women had zero balance on their mobile phones at the time of interview. Collectively, these findings highlight the immense challenges which underpin efforts to facilitate women’s phone access and use. They too underline the criticality of designing mobile health communication content for couples, rather than just wives to ensure the buy-in of male gatekeepers, and for continuing to prioritise face to face communication with women on critical health issues. Approach to segmentation Data in our sample were captured as part of special surveys carried out through the impact evaluation of Kilkari. Future programmes may be tempted to apply the approach undertaken here to existing datasets, including routine health information systems or other forms of government tracking data. In the India context, while these data are likely to be less costly than special surveys, they are comparatively limited in terms of data elements captured—particularly in terms of data ownership of different types of mobile devices, digital skill levels and usage of specific applications or social media platforms. Data quality may also be a significant issue in existing datasets. For example, we estimate that SIM change in our study population was 44% over a 12-month period—a factor which when coupled with the absence of systems to update government tracking registries raises important questions about who is retained in these databases, and therefore able to receive mobile health communications—and who is missing. Among the variables used, men’s phone access and use were most integral to developing distinct clusters. We recommend that future surveys seeking to generate data for designing digital services for women ensure that data elements are captured on men’s phone access and use practices as well as their perception of their wife’s phone access and use. In addition to underlying data, our analytic approach differed from other segmentation analyses. Our work is relatively new in global health literature related to digital health programmes that are positioned as D2B programmes. While similar ML models are being tested in various domains related to public health, they consist exclusively of unsupervised learning or supervised learning, this analysis is the first of its kind focusing on the use of a combination of supervised and unsupervised learning to identify homogenous clusters for targeting of digital health programmes. Data collected from special surveys like the couple’s dataset used here are comparatively smaller in terms of sample size but large with regard to the number of data elements available. An alternative approach to that described in this manuscript might be to develop strata based on population characteristics. Indeed, findings from the impact evaluation published elsewhere suggest that women with access to phones in the most disadvantaged sociodemographic strata (poorest (15.8% higher) and disadvantaged castes (12% higher)) had greater impact when exposed to 50% or more of the Kilkari content as compared with those not exposed. With an approach to segmentation based on these strata of highest impact, we know and understand what divides or groups respondents (eg, socioeconomic status, education) but this may not be enough when they do not explain the underlying reasons for change. In the approach used here, the study population is segmented using multiple characteristics (sociodemographic, digital access and use) simultaneously. The results are clusters comprised of individuals with mixed sociodemographic characteristics which may help to explain the reduced impact observed on health outcomes. Designing a strategy based on previously known/identifiable strata alone has been the basis of targeting in public health but has not maximised reach, exposure and effect to its fullest potential. The approach used here may better group beneficiaries based on their digital access and use characteristics which may serve to increase reach and exposure. However, further research is needed to determine how to deepen impact within these digital clusters. Findings demonstrate that the impact of the D2B solution Kilkari varied across homogenous clusters of women with access to mobile phones and their husbands in Madhya Pradesh. Across delivery channels, our analysis indicates that mobile health communication could not be effectively delivered to husbands and wives in Cluster 1 using WhatsApp, because smartphone ownership and WhatsApp use in this cluster are negligible. IVR, on the other hand, could be used to reach couples in Cluster 1, but reach is likely to be sporadic because of high levels of phone sharing with others (78% among men and 57% among women). On the other hand, WhatsApp and YouTube are likely to be effective digital channels for communicating with both husbands and wives in Cluster 3, where most men and women own or use smartphones and WhatsApp. Beyond delivery channels, study findings raise a number of important learnings for content development as well as optimising beneficiary reach and exposure. The creative approach to content created for Cluster 3, where 40% of women are from the richest socioeconomic status and only 17% have never been to school or have a primary school education or less, would need to be very different from the creative approach to content created for Cluster 1, where 53% have a poorest or poorer socioeconomic status, and 39% have never been to school or have a primary school education or less. Similarly, this analysis adds to qualitative findings and provides important insights into how gender norms related to women’s use of mobile phones may effect reach and impact. While few (13–15%) husbands indicated that ‘adults’ need oversight to use mobile phones, men’s perceptions varied when asked about specific use cases. Across all Clusters, nearly half of husbands indicated that their wives needed permission to pick up phone calls from unknown numbers—an important insight for IVR programmes which may make outbound calls without prewarning to beneficiaries. In Clusters 1 and 2, 25% and 29% of husband’s, respectively, report that their wives need permission to answer calls from health workers—as compared with 15% in Cluster 3. While restrictions on SMS and WhatsApp were lower than making or receiving calls, these channels are less viable given women’s limited access to smartphones, low literacy and digital skills. Overall, men’s perceptions on the restrictions needed on the receipt and placement of calls by women was lower for Cluster 3. However, despite the relative wealth of beneficiaries in Cluster 3 (67% were in the richer or richest socioeconomic strata), 48% of women had zero balance on their mobile phones at the time of interview. Collectively, these findings highlight the immense challenges which underpin efforts to facilitate women’s phone access and use. They too underline the criticality of designing mobile health communication content for couples, rather than just wives to ensure the buy-in of male gatekeepers, and for continuing to prioritise face to face communication with women on critical health issues. Data in our sample were captured as part of special surveys carried out through the impact evaluation of Kilkari. Future programmes may be tempted to apply the approach undertaken here to existing datasets, including routine health information systems or other forms of government tracking data. In the India context, while these data are likely to be less costly than special surveys, they are comparatively limited in terms of data elements captured—particularly in terms of data ownership of different types of mobile devices, digital skill levels and usage of specific applications or social media platforms. Data quality may also be a significant issue in existing datasets. For example, we estimate that SIM change in our study population was 44% over a 12-month period—a factor which when coupled with the absence of systems to update government tracking registries raises important questions about who is retained in these databases, and therefore able to receive mobile health communications—and who is missing. Among the variables used, men’s phone access and use were most integral to developing distinct clusters. We recommend that future surveys seeking to generate data for designing digital services for women ensure that data elements are captured on men’s phone access and use practices as well as their perception of their wife’s phone access and use. In addition to underlying data, our analytic approach differed from other segmentation analyses. Our work is relatively new in global health literature related to digital health programmes that are positioned as D2B programmes. While similar ML models are being tested in various domains related to public health, they consist exclusively of unsupervised learning or supervised learning, this analysis is the first of its kind focusing on the use of a combination of supervised and unsupervised learning to identify homogenous clusters for targeting of digital health programmes. Data collected from special surveys like the couple’s dataset used here are comparatively smaller in terms of sample size but large with regard to the number of data elements available. An alternative approach to that described in this manuscript might be to develop strata based on population characteristics. Indeed, findings from the impact evaluation published elsewhere suggest that women with access to phones in the most disadvantaged sociodemographic strata (poorest (15.8% higher) and disadvantaged castes (12% higher)) had greater impact when exposed to 50% or more of the Kilkari content as compared with those not exposed. With an approach to segmentation based on these strata of highest impact, we know and understand what divides or groups respondents (eg, socioeconomic status, education) but this may not be enough when they do not explain the underlying reasons for change. In the approach used here, the study population is segmented using multiple characteristics (sociodemographic, digital access and use) simultaneously. The results are clusters comprised of individuals with mixed sociodemographic characteristics which may help to explain the reduced impact observed on health outcomes. Designing a strategy based on previously known/identifiable strata alone has been the basis of targeting in public health but has not maximised reach, exposure and effect to its fullest potential. The approach used here may better group beneficiaries based on their digital access and use characteristics which may serve to increase reach and exposure. However, further research is needed to determine how to deepen impact within these digital clusters. Study findings sought to identify distinct clusters of husbands and wives based on their sociodemographic, phone access and use characteristics, and to explore the differential impact of a maternal mobile messaging programme across these clusters. Three identifiable groups emerge each with differing levels of digital access and use. Descriptive analyses suggest that improvements in some health behaviours were observed for a greater number of outcomes in Cluster 2, than in Clusters 1 and 3. These findings suggest that one size fits all mobile health communications solutions may only engage one segment of a target beneficiary population, and offer much promise for future D2B and other digital health programmes which could see greater reach, exposure and impact through differentiated design and implementation. More quantitative and qualitative work is needed to better understand factors driving the differences in impact and what is likely to motivate adoption of target behaviours in different clusters. Our work opens up a new avenue of research into better targeting of beneficiaries using data on variety of domains including sociodemographics, mobile phone access and use. Future work will entail evaluation of the actual platform used for targeting and delivery of the programme in pilot projects. Successful pilots can be scaled up to larger swathes of the population in India and similar setting around the world. Reviewer comments Author's manuscript |
Thermal acclimation of methanotrophs from the genus | f08574f0-a6f2-41ea-9132-47d4fdba1df4 | 10030640 | Microbiology[mh] | The only known biological CH 4 sink are methanotrophs that utilize CH 4 as an energy and carbon source . Aerobic methanotrophs belong to Gammaproteobacteria, Alphaproteobacteria, and Verrucomicrobia. These bacteria are found in soils, wetlands, water, rice paddies, landfills, sewage, and sediments, consuming subsurface biogenic or thermogenic CH 4 or harvesting the gas directly from air . Methanotrophs oxidize CH 4 using a cytoplasmic soluble methane monooxygenase or a membrane-bound particulate methane monooxygenase (sMMO or pMMO), and fixate carbon via the ribulose monophosphate pathway, serine pathway or the Calvin-Benson-Bassham Cycle . Members of the gammaproteobacterial genus Methylobacter have been identified as the most abundant and active methanotrophs in many ecosystems including peat soils, tundra, ponds, lakes, sub-glacial sediments, lake sediments, rice paddies, and landfills [ – ]. These environments are among the biggest contributors to global methane (CH 4 ) emissions, responsible for 431 Tg CH 4 yr -1 (bottom up estimated global mean for years 1978–2019) . High-emitting ecosystems often have high dissolved CH 4 concentrations (>100 µM) [ – ], meaning that Methylobacter species are often exposed to, and presumably adapted to, CH 4 uptake saturation. Soils experience a mix of stable temperatures and large temperature fluctuations driven by diurnal cycles, weather changes and season, and depending on soil type, depth, latitude, and altitude [ – ]. Highly variable temperature effects on soil CH 4 oxidation rates have been observed, ranging from, for example, strong temperature responses in landfills and marine sediments to variable responses in permafrost soils and minor effects under atmospheric CH 4 concentrations in forest soils or high CH 4 concentrations in peat . Temperature responses in methanotrophs have been suggested to depend on the soil type and CH 4 concentration . Also, CH 4 oxidation has been observed to be more temperature sensitive than CH 4 production in active layer tundra soils, but the underlying cellular mechanisms of this temperature sensitivity were not investigated . Despite the apparent importance of microbial thermal acclimation (also known as thermal acclimatization and thermal adaptation ), we do not yet have a clear understanding of how methanotrophs adjust physiologically to temperature changes and how this relates to substrate turnover and growth. Temperature responses in bacteria occur at many levels, for example, genetic changes that influence the amino acid composition of enzymes , horizontal gene transfer that influences cell structure, metabolic potential, or enzyme kinetics , modifications of membrane fatty acid composition , and changes in gene expression patterns or regulatory networks . For example, in response to short-term cold shock, some bacteria adapt DNA curvature and favor translation of cold-response transcripts . Bacterial cells have also been shown to grow to larger sizes and sustain larger intracellular pools of ATP to compensate for kinetic limitations at low temperatures . Some studies suggest that accumulation of carbon and energy storage polymers can have a role in thermal acclimation of microorganisms, as seen by, for example, the transcriptional upregulation of polyhydroxyalkanoate (PHA) storage in anoxic peat at low temperature (<10 °C) , PHA accumulation at high temperature (30–37 °C) in methanotrophic enrichments dominated by Methylocystis , and glycogen storage as a survival mechanism at low temperature in E. coli . Central to the cellular response of bacteria to altered conditions, including temperatures, is the genetic information processing machinery , including the transcription of DNA into RNA and the translation of RNA into proteins. Ribosomal tuning was shown through studies on E. coli to be a major cellular tool to optimize growth rates and physiological states in response to changes in their external conditions, including temperature (e.g., ). Recently, studies of soil microbial responses to long-term warming have revealed that the higher growth rates at higher temperatures were accompanied by a reduction in the number of ribosomes , confirming that adjustment of the protein biosynthesis machinery is an important thermal acclimation mechanism in nature. However, so far, our understanding of thermal acclimation via ribosomal regulation is mostly limited to studies on the model organism E. coli . Due to a considerable CH 4 uptake, including a consumption of up to 90% of produced CH 4 in wetlands and atmospheric CH 4 uptake in soils , methanotrophs constitute one of the most important CH 4 -sinks on Earth . As these microorganisms are frequently exposed to temperature change in soil surface layers, their means of physiological adjustment may significantly influence global CH 4 cycling. Based on the frequent reports of their detection , members of the genus Methylobacter may be a particularly important contributor to the biological CH 4 sink. However, our knowledge of temperature effects on methanotroph physiology is very limited. Here, we have investigated how Methylobacter acclimate to different temperatures and CH 4 concentrations in comparative growth and CH 4 oxidation kinetics experiments with strains from three Methylobacter species. Furthermore, to learn which cellular machineries are adjusted for thermal acclimation, we performed a series of temperature experiments including transcriptomics and measurement of cell sizes and cell contents with one of these strains, Methylobacter tundripaludum SV96 T .
Environmental distribution of Methylobacter and temperature effects on growth and CH 4 oxidation kinetics in three Methylobacter strains By screening the data available in the Earth Microbiome Project (EMP) , we observed a widespread distribution of Methylobacter -like 16S rRNA gene sequences (Fig. S ), covering 17% of all 23,813 screened EMP samples (Supplementary dataset (SD) A: Table A ) and including 256 out of 1355 unique geographic locations (SD A: Table A ). Methylobacter-like sequences were most frequently observed in freshwater (64% of all freshwater samples in the database) (SD A: Table A ) and were also most abundant in freshwater (SD A: Table A ). In line with this, members of Methylobacter are often observed in environmental studies and the type strain species M. tundripaludum (SV96 T ) is common in wetlands, lakes, and other high-CH 4 environments [ – , ]. The high abundance and global distribution Methylobacter in freshwater environments, suggests a globally important role in CH 4 cycling as these are hotspots for CH 4 production, contributing ~42% of total anthropogenic and natural CH 4 emissions . To study how thermal acclimation affects growth and CH 4 oxidation in these environmentally important microorganisms we compared growth kinetics and CH 4 oxidation kinetics of three different Methylobacter strains, Methylobacter tundripaludum SV96 T , Methylobacter sp. G7, and M. luteus ACM 3304 T (see Fig. S for experimental setup). Methylobacter tundripaludum SV96 T originates from Arctic peat and its temperature range for growth suggests psychrotolerance with a growth optimum between 15 and 25 °C . Phylogeny of its pmoA gene (common marker gene encoding the beta subunit of pMMO) places this strain in a different clade than most of the characterized Methylobacter species, including Methylobacter luteus and Methylobacter whittebury (Fig. S ) as previously also shown using phylogenomics . In experiments with M. tundripaludum SV96 T , we compared growth and CH 4 oxidation Michaels-Menten kinetics between 8, 15, 21, and 27 °C, across seven dissolved CH 4 concentrations ranging from 0.003 to 0.25 mM ( n = 84 incubations with ~4 timepoint measurements per incubation). At CH 4 -uptake saturation (V max(app) : “app” indicating cellular V max ), specific growth was fastest at 15 °C, followed by 21, 8, and 27 °C (Fig. , Fig. S , SD B: Table B ). At CH 4 concentrations below ~0.005 µM, the specific growth rate at 8 °C was higher than at the other temperatures (Fig. , Fig. S ). CH 4 oxidation rates also changed with both temperature and CH 4 concentration (Fig. , SD B: Table B ). At CH 4 -uptake saturation, highest cellular CH 4 consumption rates (V max(app) ) were detected at 15 °C, followed by 21, 8, and 27 °C (Fig. and Fig. S , SD B: Table B ). At lower CH 4 concentrations (below 0.005 µM), we observed only minor differences in CH 4 consumption between temperatures. Consequently, the growth efficiency (number of cell divisions per CH 4 molecules oxidized, modelled with four-parameter logistics curves) at CH 4 saturation was much lower at 8 and 15 °C than at 21 and 27 °C (Fig. , SD B: Table B ). However, growth efficiency increased with decreasing CH 4 concentrations at all temperatures, with the largest increase at 8 °C, followed by 15, 27, and 21 °C, respectively. This observation, that at high CH 4 concentrations the highest growth efficiency was found at high temperatures (21 and 27 °C), while at low CH 4 concentrations the highest growth efficiency was found at low temperatures (8 and 15 °C), implies large shifts in cellular resource allocation and therefore CH 4 consumption due to changes in temperature and CH 4 concentration. This also disagrees with a proposed model in which fast and slow growth are suggested to be inefficient while the highest efficiency is obtained at intermediate growth rates . Methylobacter sp. G7 is closely related to M. tundripaludum SV96 T , pmoA gene phylogeny placing these strains within the same clade (Fig. S ). It originates from a biofilm inside the coalmine G7, close to Longyearbyen, Svalbard (78°13′00″N 15°38′00″E). Methylobacter sp. G7 has not been published as a novel species to date, but its gene identities to M. tundripaludum SV96 T (94.4% for pmoA and 98.9% for the 16 S rRNA gene) and different phospholipid fatty acid (PLFA) profile compared to M. tundripaludum SV96 T (SD B: Table B ) suggest that it may represent a novel Methylobacter species. The experiments with Methylobacter sp. G7 were set up the same way as for M. tundripaludum SV96 T ( n = 84). However, this strain did not grow at 27 °C, and therefore these experiments were carried out at 4, 8, 15, and 21 °C. The highest growth rate for Methylobacter sp. G7 under CH 4 saturation was at 8 °C, followed by 4, 15, and 21 °C (Fig. , Fig. S , SD B: Table B ). The highest CH 4 oxidation rates, under CH 4 saturation, were at 21 and 8 °C, followed by 15 and 4 °C (Fig. , Fig. S , SD B: Table B ). This indicates the highest growth efficiency at 4 °C, followed by 8, 15, and 21 °C, when CH 4 concentrations are saturated (Fig. , Fig. S , SD B: Table B ), opposite to that of M. tundripaludum SV96 T (Fig. ). However, like M. tundripaludum SV96 T , the growth efficiency of Methylobacter sp. G7 increased substantially with decreasing CH 4 concentrations. Methylobacter luteus is a mesophilic species that clusters separately from the two other strains (Fig. S ) and is found in, for example, freshwater, marine sediments, sewage, and landfills . M. luteus ACM 3304 T has a different PLFA profile than the two other strains (Supplementary dataset B, Table B ). The experiments ( n = 84) with M. luteus ACM 3304 T (originally isolated from sewage ) demonstrated the highest cell division and CH 4 oxidation rates under CH 4 saturation at 27 °C, followed by 21, 15, and 8 °C (Fig. G and , Fig. S , SD B: Tables B and B ). The highest growth efficiency under CH 4 saturation was observed at 15 °C (Fig. , Fig. S , SD B: Table B ), followed by 21, 8, and 27 °C. We observed only small changes in growth efficiency with changing CH 4 concentration in M. luteus ACM 3304 T , and apart from the large increase in growth efficiency at 15 °C, we saw little change with temperature, confirming that the thermal acclimation strategy of M. luteus ACM 3304 T is very different from the two other strains. However, a large influence of temperature on CH 4 consumption per cell was common to all three strains (Fig. ). By comparing the growth efficiencies of the three Methylobacter strains we can demonstrate how thermal acclimation can affect CH 4 uptake rates. For example, we see that substantial additional quantities of CH 4 are consumed by M. tundripaludum SV96 T at 15 and 8 °C to support growth at higher CH 4 concentrations, relative to 21 and 27 °C. The same is seen at 21 °C relative to the other temperatures for Methylobacter sp. G7 (Fig. ). Maximizing growth rates can be physiologically inefficient, for example in cases where more proteins are required to speed up the rates , but this strategy can be beneficial when sufficient resources are available. On the other hand, M. tundripaludum SV96 T and Methylobacter sp. G7 have both evolved acclimation strategies where the most efficient growth occurs at low temperatures and low CH 4 concentrations (Fig. ). Efficient growth means lower CH 4 oxidation rates, and thus our results suggest that transitions from inefficient to efficient growth due to temperature changes in nature can lead to lower CH 4 oxidation rates, or vice versa. For M. luteus ACM 3304 T , the effect of temperature on CH 4 consumption is especially evident in the much more efficient growth at 15 °C relative to the other temperatures. This implies that a drop in temperature from 25 to 15 °C would reduce the size of a M. luteus ACM 3304 T -dominated CH 4 sink considerably, on a per cell basis, not because the strain cannot function at the temperature, but because it is growing more efficiently. This further indicates that the competitiveness of a strain does not necessarily correlate with its CH 4 consumption. Thus, while CH 4 production rates are often expected to increase with increasing temperatures , our data demonstrate that CH 4 oxidation rates might not always correlate, for physiological reasons, possibly explaining some of the divergent observations of temperature effects on CH 4 oxidation in environmental studies (e.g., ). Cell sizes and RNA and DNA content of M. tundripaludum SV96 T We directed our attention toward M. tundripaludum SV96 T for investigating in more depth the cellular mechanisms that are involved in thermal acclimation. For many decades, observations have been made of microorganisms, including algae and bacteria, that change size in response to temperature (e.g., ). For the temperatures 8 and 20 °C, we tested whether this was also the case for M. tundripaludum SV96 T . We observed slight, but significant, decreases of width (5% on average, n = 222–263) and length (10% on average, n = 222–263) with increasing temperature (Fig. ). However, we did not see consistent differences in optical density of the cultures despite these cell size differences, possibly due to the combination of low percentage change and technical variability in optical density measurements. We then prepared for running a transcriptomics experiment at 8, 15, 21, and 27 °C and saturating CH 4 concentrations (>0.1 mM dissolved CH 4 ) during exponential growth. A growth experiment (Fig. S , SD B, Table B ) was first carried out to test whether the same temperature response could be observed under these conditions as in the growth kinetics experiment (Fig. and Fig. S ). The results of the growth experiment showed no significant differences between the growth rates at 8 and 27 °C (0.035 and 0.038 cell divisions cell −1 h −1 , respectively), and no significant differences between the rates at 15 and 21 °C (0.085 and 0.095 cell divisions cell −1 h −1 , respectively). While these rates were lower than indicated by the model predictions from the growth kinetics experiment (Growth V max(app) : 0.0699 and 0.0612 cell divisions cell −1 h −1 at 8 °C and 27 °C, and 0.1128 and 0.1044 cell divisions cell −1 h −1 at 15 and 21 °C, respectively), the ratios between the rates at different temperatures were similar. This confirmed that temperature responses were consistent and thus comparable between experiments, so we proceeded to extract total nucleic acids from the cells incubated for transcriptomics. Prior to RNA purification we measured the quantities of RNA and DNA. The DNA quantities per 10 8 cells did not significantly differ between temperatures ( p > 0.08) while the RNA quantities were significantly different between all temperatures ( p < 0.05), except 21 and 15 °C ( p = 0.06), with highest RNA content at 15 and 21 °C, followed by 27 and 8 °C (Fig. , SD B: Table B ). Correspondingly, the RNA to DNA ratio in M. tundripaludum SV96 T varied between temperatures, with mean of 2.3, 3.7, 3.6 and 2.8, at 8, 15, 21, and 27 °C, respectively (Fig. , SD B: Table B ) and RNA:DNA ratios were significantly lower at 8 and 27 °C compared to 15 and 21 °C ( p < 0.05). Changes in cellular total RNA reflect changes in ribosomal RNA (rRNA) as most bacterial RNA is rRNA (82–90%) . Ribosomal RNA can constitute up to 20% of cell dry weight and ribosomal proteins make up 20–40% of total proteins . Thus, tuning of the cellular ribosome concentration is a key factor in optimizing growth rates in E. coli . Our results suggest that M. tundripaludum SV96 T reduces its cellular rRNA concentration at the higher (27 °C) and lower (8 °C) end of the temperature range that allows growth, while maintaining similar concentrations in the optimum range (15 and 21 °C). Similarly, in E. coli , cellular ribosome concentrations were suggested to remain constant at temperatures around the growth optimum but decline at higher and lower temperatures as some of the resources were diverted toward other physiological processes, including stress responses, that are needed to maintain high growth rates under sub-optimal conditions . Central carbon and energy metabolisms To identify which physiological components, other than rRNA, are adjusted for thermal acclimation in M. tundripaludum SV96 T , we sequenced three mRNA enriched transcriptomes from each of the temperatures (Fig. S ), resulting in a total of 12 mRNA libraries. The mRNA fraction ranged from 30 to 99% mRNA after rRNA depletion, leaving between 2.2 million and 15.3 million mRNA reads aligned to the M. tundripaludum SV96 T genome (SD B: Table B ). We observed upregulation (upregulation or increased gene expression referring to significantly increased transcript numbers) of genes encoding pMMO and methanol dehydrogenase at 15, 21, and 27 °C compared to 8 °C (Fig. , SD C: Table C ). The high oxidation rates (V max(app) ) at 8 °C (Fig. S ) despite low expression of pMMO (Fig. ) and fewer ribosomes per cell at this temperature (Fig. ) relative to 21 and 27 °C (Fig. S ), suggest that M. tundripaludum SV96 T may carry a low temperature-adapted pMMO. The combination of high pMMO expression (Fig. ) and high cellular ribosome content (Fig. ) at 15 °C corresponded to the highest cellular CH 4 oxidation rate (V max(app) = 0.36 µmol CH 4 10 8 cells −1 h −1 ) (Fig. S ). However, at 21 °C, similarly high ribosome content and pMMO expression as at 15 °C resulted in a much lower cellular V max(app) (Fig. S ), also supporting the proposition of a low temperature-adapted pMMO. Perhaps an inefficient pMMO at high temperature is one of the reasons for the overall higher pMMO gene expression at temperatures above 8 °C. We observed upregulation at 15 °C of genes in the NADH-generating tetrahydromethanopterin (H 4 MPT) pathway, used for C 1 transfer during formaldehyde oxidation to CO 2 (Fig. ). This may suggest higher rates of energy conservation. However, there was no consistent upregulation of oxidative phosphorylation at 15 °C (Fig. ). Rather, we observed increased expression, at different temperatures, of genes for different enzyme systems with the same functional role (e.g., the NUO and NQR versions of the NADH dehydrogenase complex I). Isozymes, enzymes with different amino acid sequences that catalyze the same reaction, have previously been suggested as an important part of microbial thermal acclimation . However, the discrete functional roles of different functionally analogous but non-homologous enzyme systems with roles in oxidative phosphorylation are generally not well understood . At 15 °C, we also observed upregulation of a large proportion of the genes leading from CH 4 oxidation through the Ribulose monophosphate pathway (RuMP) fructose bisphosphate branch 1 (FBP 1) toward nucleotide biosynthesis (Fig. ). Considering that nucleotides are the main building blocks for RNA, this matched the high cellular RNA content at 15 °C (Fig. ). Nucleotide biosynthesis, transcription, and translation Nucleotide biosynthesis pathways receive the precursors 2-Oxo-glutarate and D-ribulose-5P from the tricarboxylic acid cycle and RuMP, respectively. We observed upregulation at 15 °C for 30 out of 46 pathway steps needed for nucleotide biosynthesis from these two precursors, matching the upregulation of RuMP toward nucleotide biosynthesis (Fig. S , SD C: Table C ). Among the genes encoding the DNA-directed RNA polymerase, only the β subunit, rpoB (α; rpoA , β ’ ; rpoC , β; rpoB ) was upregulated at 15 °C (Fig. S , SD C: Table C ). The genes for three out of four enzymes that compose the RNA degradosome; RNase E, Enolase, and Pnpase (polynucleotide phosphorylase); were also upregulated at 15 °C (Fig. S ). Only one of the genes encoding an ATP dependent RNA helicase ( rhlB ), was upregulated at 15 °C (although not relative to 8 °C), rhlE was not. Furthermore, genes encoding enzymes for protein folding ( groE , groL , and dnaK ) were upregulated at 15 °C (Fig. S ). These increased investments into transcription, RNA degradation, and protein folding at 15 °C matched an overall upregulation of genes for the small subunit (SSU) and large subunit (LSU) ribosomal proteins, with 15 out of 29 LSU and 14 out of 19 SSU proteins being significantly higher expressed at 15 °C (Fig. , SD C: Table C ). This was further supported by the upregulation of genes for many individual amino acid biosynthesis pathway steps (43 out of 79) at 15 °C relative to the other temperatures (Fig. S , SD C: Table C ). These findings are in accordance with previous observations of transcriptional adjustment of genes for translation, amino acid biosynthesis, nucleotide metabolism, and protein metabolism in E. coli exposed to changing temperatures . The upregulation of these genes in M. tundripaludum SV96 T was observed exclusively at 15 °C (Figs. , S , S , S ), while similar cellular rRNA concentrations were found at 15 and 21 °C (Fig. ). Thus, it seems that to maintain a similar cellular rRNA concentration and similar growth rate at 15 and 21 °C, the cell must tune its gene expression differently. In this context, it is important to note that the overall upregulation of protein biosynthesis machinery genes at 15 °C also does not imply that we find the highest protein synthesis rates at 15 °C. The reason is that all synthesis rates, including the synthesis of ribosomes and proteins, are directly influenced by temperature . However, it does imply higher protein biosynthesis rates at 15 °C, relative to the rates that would have been obtained at 15 °C without upregulation. Such catalytic compensation comes at a cost. Perhaps this is one of the reasons for the lower growth efficiency under CH 4 saturation at 15 °C, relative to 21 °C (Figs. and S ). Maybe the similar cellular rRNA concentrations and growth rates at 15 and 21 °C was only possible due to the increased cellular CH 4 consumption and larger investment into protein biosynthesis at 15 °C. Catalytic compensation at low temperature was also previously observed, for example, in the upregulation of ATP synthesis and CO 2 fixation via RuBisCO in bacteria and plants, respectively , and suggested as part of the bacterial response to soil warming . Glycogen storage Several bacteria are known to store superfluous carbon and energy as, for example, glycogen or PHA , and temperature effects on storage have been observed previously . In searching the transcriptome of M. tundripaludum SV96 T for carbon storage pathway genes we identified all steps leading from CH 4 oxidation, via the initial steps of RuMP-FBP (Fig. ), to glycogen synthesis (Fig. , SD C: Table C ). Starting from formaldehyde, the genes for six out of the eight steps needed to produce glycogen were upregulated at 8 °C and 15 °C (Fig. ). However, for the final step from amylose to glycogen, catalyzed by a 1,4-alpha-glucan branching enzyme (EC: 2.4.1.18), we counted two gene copies. These copies were expressed in opposite patterns in relation to temperature (Fig. ), suggesting this as an important regulatory step for glycogen synthesis. This also made us question whether the expression patterns really meant that glycogen synthesis is upregulated at low temperature. Thus, in a follow-up experiment, cells were harvested in exponential phase (the same physiological state of cells harvested for transcriptomics) for measurement of cellular glycogen. We could show that cells had accumulated highest concentrations of glycogen at 21 °C, followed by 27, 15, and 8 °C (Fig. , SD B: Table B ). Thus, our measurements matched the transcriptional pattern of only one of the gene copies encoding a 1,4-alpha-glucan branching enzyme, not the overall expression pattern of the genes in the pathway (Fig. ). Regulatory adjustments of growth, membranes, cell walls, and exopolysaccharides After having observed the adjustment of several central cellular mechanisms to different temperatures we surveyed those functions most directly related to growth. FtsZ is a key cytoskeleton protein involved in bacterial cell division, forming the Z-ring, a constricting structure at the division site . The relative gene expression pattern for FtsZ and related proteins reflected the growth rates of the cultures at different temperatures with similar expression levels at 15 and 21 °C, and lower expression at 8 and 27 °C (Fig. S , SD C: Table C ). This further corresponded to the upregulation of genes for cell wall synthesis at both 15 and 21 °C (Fig. S , SD C: Table C ). At 27 °C we observed a strong upregulation of genes for exopolysaccharides (Fig. S , SD C: Table C ). As M. tundripaludum SV96 T does not grow above 30 °C , these responses might reflect adjustments needed for survival at high temperatures. Increased expression of exopolysaccharides is often associated with sub-optimal conditions, including temperatures outside the growth optimum range and can have a role as cryoprotectant in microorganisms . Exopolysaccharides are also thought to protect cells from high temperatures by strengthening their structural integrity . On the contrary, desaturation of fatty acids to increase membrane fluidity is a common response in bacteria exposed to low temperature, for example, . Correspondingly, we observed significant upregulation of fatty acid desaturase ( desA ) at 8 °C and 15 °C, relative to 21 and 27 °C (Fig. S , SD C: Table C ). Implications of thermal acclimation In this study, we have demonstrated different thermal acclimation strategies in three members of the widespread methanotroph genus Methylobacter . We have also shown that the effect of temperature on growth and CH 4 consumption depends on the CH 4 concentration, and that CH 4 oxidation rates do not correlate with temperature and growth. This means that higher temperatures can, in some instances, lead to more efficiently growing methanotrophs that oxidize less CH 4 per cell over time. If the growth of such methanotrophs should be further inhibited by for example nutrient or oxygen limitations, increasing temperatures could lead to lower CH 4 oxidation rates in some ecosystems, despite the microorganisms themselves being better adapted to the higher temperature. These counter-intuitive insights may explain why temperature effects on CH 4 oxidation rates in nature vary so much. Furthermore, we have provided insights into the physiological adjustments that underlie thermal acclimation of CH 4 oxidation and growth, focusing on M. tundripaludum SV96 T . These physiological adjustments include changes in cell sizes, the protein biosynthesis machinery, glycogen storage, cell division, and exopolysaccharide expression. Considering the scale of temperature change globally, including daily, seasonal, and long-term climate change, thermal acclimation of methanotrophs might exert considerable influence on global CH 4 cycling. We show that the temperature effect on CH 4 consumption is directly related to how different strains adjust their physiology during thermal acclimation. Furthermore, large differences in thermal acclimation strategies and temperature effects on CH 4 consumption between strains suggest that also community shifts could have a substantial impact on the size of the biological CH 4 sink.
Methylobacter and temperature effects on growth and CH 4 oxidation kinetics in three Methylobacter strains By screening the data available in the Earth Microbiome Project (EMP) , we observed a widespread distribution of Methylobacter -like 16S rRNA gene sequences (Fig. S ), covering 17% of all 23,813 screened EMP samples (Supplementary dataset (SD) A: Table A ) and including 256 out of 1355 unique geographic locations (SD A: Table A ). Methylobacter-like sequences were most frequently observed in freshwater (64% of all freshwater samples in the database) (SD A: Table A ) and were also most abundant in freshwater (SD A: Table A ). In line with this, members of Methylobacter are often observed in environmental studies and the type strain species M. tundripaludum (SV96 T ) is common in wetlands, lakes, and other high-CH 4 environments [ – , ]. The high abundance and global distribution Methylobacter in freshwater environments, suggests a globally important role in CH 4 cycling as these are hotspots for CH 4 production, contributing ~42% of total anthropogenic and natural CH 4 emissions . To study how thermal acclimation affects growth and CH 4 oxidation in these environmentally important microorganisms we compared growth kinetics and CH 4 oxidation kinetics of three different Methylobacter strains, Methylobacter tundripaludum SV96 T , Methylobacter sp. G7, and M. luteus ACM 3304 T (see Fig. S for experimental setup). Methylobacter tundripaludum SV96 T originates from Arctic peat and its temperature range for growth suggests psychrotolerance with a growth optimum between 15 and 25 °C . Phylogeny of its pmoA gene (common marker gene encoding the beta subunit of pMMO) places this strain in a different clade than most of the characterized Methylobacter species, including Methylobacter luteus and Methylobacter whittebury (Fig. S ) as previously also shown using phylogenomics . In experiments with M. tundripaludum SV96 T , we compared growth and CH 4 oxidation Michaels-Menten kinetics between 8, 15, 21, and 27 °C, across seven dissolved CH 4 concentrations ranging from 0.003 to 0.25 mM ( n = 84 incubations with ~4 timepoint measurements per incubation). At CH 4 -uptake saturation (V max(app) : “app” indicating cellular V max ), specific growth was fastest at 15 °C, followed by 21, 8, and 27 °C (Fig. , Fig. S , SD B: Table B ). At CH 4 concentrations below ~0.005 µM, the specific growth rate at 8 °C was higher than at the other temperatures (Fig. , Fig. S ). CH 4 oxidation rates also changed with both temperature and CH 4 concentration (Fig. , SD B: Table B ). At CH 4 -uptake saturation, highest cellular CH 4 consumption rates (V max(app) ) were detected at 15 °C, followed by 21, 8, and 27 °C (Fig. and Fig. S , SD B: Table B ). At lower CH 4 concentrations (below 0.005 µM), we observed only minor differences in CH 4 consumption between temperatures. Consequently, the growth efficiency (number of cell divisions per CH 4 molecules oxidized, modelled with four-parameter logistics curves) at CH 4 saturation was much lower at 8 and 15 °C than at 21 and 27 °C (Fig. , SD B: Table B ). However, growth efficiency increased with decreasing CH 4 concentrations at all temperatures, with the largest increase at 8 °C, followed by 15, 27, and 21 °C, respectively. This observation, that at high CH 4 concentrations the highest growth efficiency was found at high temperatures (21 and 27 °C), while at low CH 4 concentrations the highest growth efficiency was found at low temperatures (8 and 15 °C), implies large shifts in cellular resource allocation and therefore CH 4 consumption due to changes in temperature and CH 4 concentration. This also disagrees with a proposed model in which fast and slow growth are suggested to be inefficient while the highest efficiency is obtained at intermediate growth rates . Methylobacter sp. G7 is closely related to M. tundripaludum SV96 T , pmoA gene phylogeny placing these strains within the same clade (Fig. S ). It originates from a biofilm inside the coalmine G7, close to Longyearbyen, Svalbard (78°13′00″N 15°38′00″E). Methylobacter sp. G7 has not been published as a novel species to date, but its gene identities to M. tundripaludum SV96 T (94.4% for pmoA and 98.9% for the 16 S rRNA gene) and different phospholipid fatty acid (PLFA) profile compared to M. tundripaludum SV96 T (SD B: Table B ) suggest that it may represent a novel Methylobacter species. The experiments with Methylobacter sp. G7 were set up the same way as for M. tundripaludum SV96 T ( n = 84). However, this strain did not grow at 27 °C, and therefore these experiments were carried out at 4, 8, 15, and 21 °C. The highest growth rate for Methylobacter sp. G7 under CH 4 saturation was at 8 °C, followed by 4, 15, and 21 °C (Fig. , Fig. S , SD B: Table B ). The highest CH 4 oxidation rates, under CH 4 saturation, were at 21 and 8 °C, followed by 15 and 4 °C (Fig. , Fig. S , SD B: Table B ). This indicates the highest growth efficiency at 4 °C, followed by 8, 15, and 21 °C, when CH 4 concentrations are saturated (Fig. , Fig. S , SD B: Table B ), opposite to that of M. tundripaludum SV96 T (Fig. ). However, like M. tundripaludum SV96 T , the growth efficiency of Methylobacter sp. G7 increased substantially with decreasing CH 4 concentrations. Methylobacter luteus is a mesophilic species that clusters separately from the two other strains (Fig. S ) and is found in, for example, freshwater, marine sediments, sewage, and landfills . M. luteus ACM 3304 T has a different PLFA profile than the two other strains (Supplementary dataset B, Table B ). The experiments ( n = 84) with M. luteus ACM 3304 T (originally isolated from sewage ) demonstrated the highest cell division and CH 4 oxidation rates under CH 4 saturation at 27 °C, followed by 21, 15, and 8 °C (Fig. G and , Fig. S , SD B: Tables B and B ). The highest growth efficiency under CH 4 saturation was observed at 15 °C (Fig. , Fig. S , SD B: Table B ), followed by 21, 8, and 27 °C. We observed only small changes in growth efficiency with changing CH 4 concentration in M. luteus ACM 3304 T , and apart from the large increase in growth efficiency at 15 °C, we saw little change with temperature, confirming that the thermal acclimation strategy of M. luteus ACM 3304 T is very different from the two other strains. However, a large influence of temperature on CH 4 consumption per cell was common to all three strains (Fig. ). By comparing the growth efficiencies of the three Methylobacter strains we can demonstrate how thermal acclimation can affect CH 4 uptake rates. For example, we see that substantial additional quantities of CH 4 are consumed by M. tundripaludum SV96 T at 15 and 8 °C to support growth at higher CH 4 concentrations, relative to 21 and 27 °C. The same is seen at 21 °C relative to the other temperatures for Methylobacter sp. G7 (Fig. ). Maximizing growth rates can be physiologically inefficient, for example in cases where more proteins are required to speed up the rates , but this strategy can be beneficial when sufficient resources are available. On the other hand, M. tundripaludum SV96 T and Methylobacter sp. G7 have both evolved acclimation strategies where the most efficient growth occurs at low temperatures and low CH 4 concentrations (Fig. ). Efficient growth means lower CH 4 oxidation rates, and thus our results suggest that transitions from inefficient to efficient growth due to temperature changes in nature can lead to lower CH 4 oxidation rates, or vice versa. For M. luteus ACM 3304 T , the effect of temperature on CH 4 consumption is especially evident in the much more efficient growth at 15 °C relative to the other temperatures. This implies that a drop in temperature from 25 to 15 °C would reduce the size of a M. luteus ACM 3304 T -dominated CH 4 sink considerably, on a per cell basis, not because the strain cannot function at the temperature, but because it is growing more efficiently. This further indicates that the competitiveness of a strain does not necessarily correlate with its CH 4 consumption. Thus, while CH 4 production rates are often expected to increase with increasing temperatures , our data demonstrate that CH 4 oxidation rates might not always correlate, for physiological reasons, possibly explaining some of the divergent observations of temperature effects on CH 4 oxidation in environmental studies (e.g., ).
M. tundripaludum SV96 T We directed our attention toward M. tundripaludum SV96 T for investigating in more depth the cellular mechanisms that are involved in thermal acclimation. For many decades, observations have been made of microorganisms, including algae and bacteria, that change size in response to temperature (e.g., ). For the temperatures 8 and 20 °C, we tested whether this was also the case for M. tundripaludum SV96 T . We observed slight, but significant, decreases of width (5% on average, n = 222–263) and length (10% on average, n = 222–263) with increasing temperature (Fig. ). However, we did not see consistent differences in optical density of the cultures despite these cell size differences, possibly due to the combination of low percentage change and technical variability in optical density measurements. We then prepared for running a transcriptomics experiment at 8, 15, 21, and 27 °C and saturating CH 4 concentrations (>0.1 mM dissolved CH 4 ) during exponential growth. A growth experiment (Fig. S , SD B, Table B ) was first carried out to test whether the same temperature response could be observed under these conditions as in the growth kinetics experiment (Fig. and Fig. S ). The results of the growth experiment showed no significant differences between the growth rates at 8 and 27 °C (0.035 and 0.038 cell divisions cell −1 h −1 , respectively), and no significant differences between the rates at 15 and 21 °C (0.085 and 0.095 cell divisions cell −1 h −1 , respectively). While these rates were lower than indicated by the model predictions from the growth kinetics experiment (Growth V max(app) : 0.0699 and 0.0612 cell divisions cell −1 h −1 at 8 °C and 27 °C, and 0.1128 and 0.1044 cell divisions cell −1 h −1 at 15 and 21 °C, respectively), the ratios between the rates at different temperatures were similar. This confirmed that temperature responses were consistent and thus comparable between experiments, so we proceeded to extract total nucleic acids from the cells incubated for transcriptomics. Prior to RNA purification we measured the quantities of RNA and DNA. The DNA quantities per 10 8 cells did not significantly differ between temperatures ( p > 0.08) while the RNA quantities were significantly different between all temperatures ( p < 0.05), except 21 and 15 °C ( p = 0.06), with highest RNA content at 15 and 21 °C, followed by 27 and 8 °C (Fig. , SD B: Table B ). Correspondingly, the RNA to DNA ratio in M. tundripaludum SV96 T varied between temperatures, with mean of 2.3, 3.7, 3.6 and 2.8, at 8, 15, 21, and 27 °C, respectively (Fig. , SD B: Table B ) and RNA:DNA ratios were significantly lower at 8 and 27 °C compared to 15 and 21 °C ( p < 0.05). Changes in cellular total RNA reflect changes in ribosomal RNA (rRNA) as most bacterial RNA is rRNA (82–90%) . Ribosomal RNA can constitute up to 20% of cell dry weight and ribosomal proteins make up 20–40% of total proteins . Thus, tuning of the cellular ribosome concentration is a key factor in optimizing growth rates in E. coli . Our results suggest that M. tundripaludum SV96 T reduces its cellular rRNA concentration at the higher (27 °C) and lower (8 °C) end of the temperature range that allows growth, while maintaining similar concentrations in the optimum range (15 and 21 °C). Similarly, in E. coli , cellular ribosome concentrations were suggested to remain constant at temperatures around the growth optimum but decline at higher and lower temperatures as some of the resources were diverted toward other physiological processes, including stress responses, that are needed to maintain high growth rates under sub-optimal conditions .
To identify which physiological components, other than rRNA, are adjusted for thermal acclimation in M. tundripaludum SV96 T , we sequenced three mRNA enriched transcriptomes from each of the temperatures (Fig. S ), resulting in a total of 12 mRNA libraries. The mRNA fraction ranged from 30 to 99% mRNA after rRNA depletion, leaving between 2.2 million and 15.3 million mRNA reads aligned to the M. tundripaludum SV96 T genome (SD B: Table B ). We observed upregulation (upregulation or increased gene expression referring to significantly increased transcript numbers) of genes encoding pMMO and methanol dehydrogenase at 15, 21, and 27 °C compared to 8 °C (Fig. , SD C: Table C ). The high oxidation rates (V max(app) ) at 8 °C (Fig. S ) despite low expression of pMMO (Fig. ) and fewer ribosomes per cell at this temperature (Fig. ) relative to 21 and 27 °C (Fig. S ), suggest that M. tundripaludum SV96 T may carry a low temperature-adapted pMMO. The combination of high pMMO expression (Fig. ) and high cellular ribosome content (Fig. ) at 15 °C corresponded to the highest cellular CH 4 oxidation rate (V max(app) = 0.36 µmol CH 4 10 8 cells −1 h −1 ) (Fig. S ). However, at 21 °C, similarly high ribosome content and pMMO expression as at 15 °C resulted in a much lower cellular V max(app) (Fig. S ), also supporting the proposition of a low temperature-adapted pMMO. Perhaps an inefficient pMMO at high temperature is one of the reasons for the overall higher pMMO gene expression at temperatures above 8 °C. We observed upregulation at 15 °C of genes in the NADH-generating tetrahydromethanopterin (H 4 MPT) pathway, used for C 1 transfer during formaldehyde oxidation to CO 2 (Fig. ). This may suggest higher rates of energy conservation. However, there was no consistent upregulation of oxidative phosphorylation at 15 °C (Fig. ). Rather, we observed increased expression, at different temperatures, of genes for different enzyme systems with the same functional role (e.g., the NUO and NQR versions of the NADH dehydrogenase complex I). Isozymes, enzymes with different amino acid sequences that catalyze the same reaction, have previously been suggested as an important part of microbial thermal acclimation . However, the discrete functional roles of different functionally analogous but non-homologous enzyme systems with roles in oxidative phosphorylation are generally not well understood . At 15 °C, we also observed upregulation of a large proportion of the genes leading from CH 4 oxidation through the Ribulose monophosphate pathway (RuMP) fructose bisphosphate branch 1 (FBP 1) toward nucleotide biosynthesis (Fig. ). Considering that nucleotides are the main building blocks for RNA, this matched the high cellular RNA content at 15 °C (Fig. ).
Nucleotide biosynthesis pathways receive the precursors 2-Oxo-glutarate and D-ribulose-5P from the tricarboxylic acid cycle and RuMP, respectively. We observed upregulation at 15 °C for 30 out of 46 pathway steps needed for nucleotide biosynthesis from these two precursors, matching the upregulation of RuMP toward nucleotide biosynthesis (Fig. S , SD C: Table C ). Among the genes encoding the DNA-directed RNA polymerase, only the β subunit, rpoB (α; rpoA , β ’ ; rpoC , β; rpoB ) was upregulated at 15 °C (Fig. S , SD C: Table C ). The genes for three out of four enzymes that compose the RNA degradosome; RNase E, Enolase, and Pnpase (polynucleotide phosphorylase); were also upregulated at 15 °C (Fig. S ). Only one of the genes encoding an ATP dependent RNA helicase ( rhlB ), was upregulated at 15 °C (although not relative to 8 °C), rhlE was not. Furthermore, genes encoding enzymes for protein folding ( groE , groL , and dnaK ) were upregulated at 15 °C (Fig. S ). These increased investments into transcription, RNA degradation, and protein folding at 15 °C matched an overall upregulation of genes for the small subunit (SSU) and large subunit (LSU) ribosomal proteins, with 15 out of 29 LSU and 14 out of 19 SSU proteins being significantly higher expressed at 15 °C (Fig. , SD C: Table C ). This was further supported by the upregulation of genes for many individual amino acid biosynthesis pathway steps (43 out of 79) at 15 °C relative to the other temperatures (Fig. S , SD C: Table C ). These findings are in accordance with previous observations of transcriptional adjustment of genes for translation, amino acid biosynthesis, nucleotide metabolism, and protein metabolism in E. coli exposed to changing temperatures . The upregulation of these genes in M. tundripaludum SV96 T was observed exclusively at 15 °C (Figs. , S , S , S ), while similar cellular rRNA concentrations were found at 15 and 21 °C (Fig. ). Thus, it seems that to maintain a similar cellular rRNA concentration and similar growth rate at 15 and 21 °C, the cell must tune its gene expression differently. In this context, it is important to note that the overall upregulation of protein biosynthesis machinery genes at 15 °C also does not imply that we find the highest protein synthesis rates at 15 °C. The reason is that all synthesis rates, including the synthesis of ribosomes and proteins, are directly influenced by temperature . However, it does imply higher protein biosynthesis rates at 15 °C, relative to the rates that would have been obtained at 15 °C without upregulation. Such catalytic compensation comes at a cost. Perhaps this is one of the reasons for the lower growth efficiency under CH 4 saturation at 15 °C, relative to 21 °C (Figs. and S ). Maybe the similar cellular rRNA concentrations and growth rates at 15 and 21 °C was only possible due to the increased cellular CH 4 consumption and larger investment into protein biosynthesis at 15 °C. Catalytic compensation at low temperature was also previously observed, for example, in the upregulation of ATP synthesis and CO 2 fixation via RuBisCO in bacteria and plants, respectively , and suggested as part of the bacterial response to soil warming .
Several bacteria are known to store superfluous carbon and energy as, for example, glycogen or PHA , and temperature effects on storage have been observed previously . In searching the transcriptome of M. tundripaludum SV96 T for carbon storage pathway genes we identified all steps leading from CH 4 oxidation, via the initial steps of RuMP-FBP (Fig. ), to glycogen synthesis (Fig. , SD C: Table C ). Starting from formaldehyde, the genes for six out of the eight steps needed to produce glycogen were upregulated at 8 °C and 15 °C (Fig. ). However, for the final step from amylose to glycogen, catalyzed by a 1,4-alpha-glucan branching enzyme (EC: 2.4.1.18), we counted two gene copies. These copies were expressed in opposite patterns in relation to temperature (Fig. ), suggesting this as an important regulatory step for glycogen synthesis. This also made us question whether the expression patterns really meant that glycogen synthesis is upregulated at low temperature. Thus, in a follow-up experiment, cells were harvested in exponential phase (the same physiological state of cells harvested for transcriptomics) for measurement of cellular glycogen. We could show that cells had accumulated highest concentrations of glycogen at 21 °C, followed by 27, 15, and 8 °C (Fig. , SD B: Table B ). Thus, our measurements matched the transcriptional pattern of only one of the gene copies encoding a 1,4-alpha-glucan branching enzyme, not the overall expression pattern of the genes in the pathway (Fig. ).
After having observed the adjustment of several central cellular mechanisms to different temperatures we surveyed those functions most directly related to growth. FtsZ is a key cytoskeleton protein involved in bacterial cell division, forming the Z-ring, a constricting structure at the division site . The relative gene expression pattern for FtsZ and related proteins reflected the growth rates of the cultures at different temperatures with similar expression levels at 15 and 21 °C, and lower expression at 8 and 27 °C (Fig. S , SD C: Table C ). This further corresponded to the upregulation of genes for cell wall synthesis at both 15 and 21 °C (Fig. S , SD C: Table C ). At 27 °C we observed a strong upregulation of genes for exopolysaccharides (Fig. S , SD C: Table C ). As M. tundripaludum SV96 T does not grow above 30 °C , these responses might reflect adjustments needed for survival at high temperatures. Increased expression of exopolysaccharides is often associated with sub-optimal conditions, including temperatures outside the growth optimum range and can have a role as cryoprotectant in microorganisms . Exopolysaccharides are also thought to protect cells from high temperatures by strengthening their structural integrity . On the contrary, desaturation of fatty acids to increase membrane fluidity is a common response in bacteria exposed to low temperature, for example, . Correspondingly, we observed significant upregulation of fatty acid desaturase ( desA ) at 8 °C and 15 °C, relative to 21 and 27 °C (Fig. S , SD C: Table C ).
In this study, we have demonstrated different thermal acclimation strategies in three members of the widespread methanotroph genus Methylobacter . We have also shown that the effect of temperature on growth and CH 4 consumption depends on the CH 4 concentration, and that CH 4 oxidation rates do not correlate with temperature and growth. This means that higher temperatures can, in some instances, lead to more efficiently growing methanotrophs that oxidize less CH 4 per cell over time. If the growth of such methanotrophs should be further inhibited by for example nutrient or oxygen limitations, increasing temperatures could lead to lower CH 4 oxidation rates in some ecosystems, despite the microorganisms themselves being better adapted to the higher temperature. These counter-intuitive insights may explain why temperature effects on CH 4 oxidation rates in nature vary so much. Furthermore, we have provided insights into the physiological adjustments that underlie thermal acclimation of CH 4 oxidation and growth, focusing on M. tundripaludum SV96 T . These physiological adjustments include changes in cell sizes, the protein biosynthesis machinery, glycogen storage, cell division, and exopolysaccharide expression. Considering the scale of temperature change globally, including daily, seasonal, and long-term climate change, thermal acclimation of methanotrophs might exert considerable influence on global CH 4 cycling. We show that the temperature effect on CH 4 consumption is directly related to how different strains adjust their physiology during thermal acclimation. Furthermore, large differences in thermal acclimation strategies and temperature effects on CH 4 consumption between strains suggest that also community shifts could have a substantial impact on the size of the biological CH 4 sink.
Cultures During stock culture maintenance, pre-incubations, and experiments, M. tundripaludum SV96 T , Methylobacter sp. G7 (unpublished), and Methylobacter luteus ACM 3304 T were cultivated in nitrate minimal salt (NMS) medium at pH 6.8 (see for trace element solution). See (SI) methods section, subsection “Cultures” for more information. Acclimation During acclimation, cells were incubated in NMS medium under ~80% air and ~20% CH 4 (100% CH 4 was injected into the air-containing bottle headspace) at 8, 15, 21, and 27 °C ( M. tundripaludum SV96 T and M. luteus ACM 3304 T ), or 4, 8, 15 and 21 °C ( Methylobacter sp. G7) (Fig. S ). The cultures were always allowed acclimation for the time needed to surpass 10 generations (7–12 days). During acclimation and experiments, cultures were shaken in a horizontal position, at 50 rpm. Some control experiments were carried out at 150 rpm to test the effect on mass transfer limitations. See SI methods, subsection “Acclimation” for more information. Shaking speed and mass transfer limitations Strains that consume gases can experience mass transfer limitations due to low substrate solubility. See SI methods, subsection “Shaking speed and mass transfer limitations.” for more information on how we accounted for mass transfer limitation. Cell growth experiments From samples of acclimated exponential phase M. tundripaludum SV96 T cultures (75 ml) with densities of ~1–3 × 10 7 cells mL −1 22 mL were transferred to three 125 mL glass bottles per temperature. Bottles were then prepared with headspaces of 80% air and 20% CH 4 and a pressure of 1.3 atm. This resulted in 0.36–0.51 mM dissolved CH 4 , depending on the temperature, above the threshold of CH 4 saturation for M. tundripaludum SV96 T at this density (0.1 mM CH 4 ; Fig. S ). The bottles were incubated at 8, 15, 21, and 27 °C and 50 rpm (Fig. S ) or 150 rpm (Fig. S ) for up to 36 h. See SD B: Table B , B for incubation times and sampling time points for each experiment. At time zero and respective sampling intervals, 300 µL of cell suspension (in duplicate) was transferred to a Nunclon Delta Surface plate (Thermo Scientific, Waltham, MA, USA) and the optical density was measured (Spectra Max 250 microplate reader, Molecular Devices, San José, CA, USA) at 600 nm (diluted NMS medium as blank). For optical density measurements (OD 600 ), blanks were subtracted from measurements. CH 4 oxidation kinetics and growth kinetics Subsamples of acclimated exponential-phase cultures with densities of 5 × 10 7 cells mL −1 were aliquoted (21.6 mL) in 125 mL glass bottles for measurement of CH 4 oxidation and growth rates at seven different CH 4 concentrations and four different temperatures (8, 15, 21, and 27 °C for M. luteus ACM 3304 T and M. tundripaludum SV96 T , and 4, 8, 15, and 21 °C for Methylobacter sp. G7). We used three aliquots for each of the seven CH 4 concentrations, for a total of 21 bottles per temperature. With four temperatures, that amounted to a total of 84 bottles per strain (Fig. S ). For each temperature, two negative controls (medium without cells) in 125 mL glass bottles were prepared per CH 4 concentration (56 negative controls in total per strain). None of the negative controls indicated non-biological CH 4 consumption or leakage. To create the seven different CH 4 concentrations, seven volumes (200 µL; 600 µL; 1,5 mL; 3 mL; 6 mL; 12 mL; 15 mL) of 100% CH 4 were injected with a plastic syringe (BD Plastipak, Franklin Lakes, NJ, USA) with a sterile 0.5 × 16 mm needle (BD Microlance, Franklin Lakes, NJ, USA) from a multi-layer polypropylene gas bag (RESTEK, Bad Homburd vor der Höhe, Germany). Final headspace concentrations of CH 4 ranged between 1000 and 130,000 p.p.m.v. while dissolved concentrations of CH 4 ranged between 1.5 and 260 µM depending also on temperature. The gas pressures in the bottles were then adjusted to a total headspace pressure of ~1.3 atm at 20 °C by injecting additional volumes of air. See SI methods, subsection “CH 4 oxidation kinetics and growth kinetics” for more information. Glycogen quantification The cells were acclimated and cultivated under 20% CH 4 in air as described above. Five replicate bottles with 20 mL culture and 20% CH 4 in air were incubated per temperature in darkness with 50 rpm shaking, five bottles with a starting cell concentration of ~1 × 10 7 cells mL −1 and five with a density of ~1 × 10 8 cells mL -1 . An incubation length of 48 h at 15 and 21 °C was sufficient to reach high-density exponential growth phase at 27 °C and 8 °C it took 60 and 90 h, respectively, to reach those same cell densities. During harvest, 2 mL of culture was sampled for glycogen measurements from the dense cultures and 10 mL from the dilute cultures. The samples were processed according to recommendations from the manufacturer (Glycogen assay kit, ab65620, Abcam, Cambridge, UK). Glycogen was quantified using a plate reader (GloMax Explorer, Promega, Madison, WI, USA). See SI methods, subsection “Glycogen quantification” for more information. OD600 to cell number conversion and cell sizes To normalize rates to cell numbers, standard curves correlating optical density (OD 600 ) to cell numbers were created for the three strains: M. tundripaludum SV96 T , M. luteus ACM 3304 T and Methylobacter sp. G7 For size estimation, cells were visualized using a Carl Zeiss AxioObserver Z1 with a 100x objective and Bright-Field and measured with the size estimation tool available in the AxioVision SE64 Rel 4.9.1 software. Cultures for size estimation were prepared by acclimation of cultures to 8 and 21 °C, as described above, followed by cultivation to reach exponential phase and harvest for size estimation. See SI methods, subsection “OD600 to cell number conversion and cell sizes” for more information. Calculations Specific growth rates were calculated as the slope of the natural logarithm of optical densities against time, during exponential growth. Mixing ratios of CH 4 and CO 2 were calculated by comparison to certified standards. Masses of headspace and dissolved CH 4 and total CO 2 at different temperatures were calculated from the mixing ratios of CH 4 and CO 2 using Henry’s Law, assuming an ideal state, knowing the ambient pressure, temperature, headspace volume of the bottle, headspace pressure, liquid volume, and respective temperature-dependent solubility constants of the gases. All calculations accounted for removal of gas and liquid for measurements. We calculated the CH 4 oxidation and CO 2 production rates from slopes of linear models fitted to the concentrations measured during experiments and adjusted to the number of cells in the culture. Growth efficiencies were calculated by multiplying the specific growth (cell divisions cell −1 h −1 ) rate by 10 8 and then dividing by CH 4 the oxidation rates (µmol CH 4 oxidized 10 8 cells −1 h −1 ) to give cell divisions per µmol CH 4 oxidized. The data used to estimate growth efficiency were the values predicted for each measurement time-point from the non-linear Michaelis-Menten regression models. For estimates of specific growth rates, CH 4 oxidation, and growth efficiency at CH 4 saturation, V max(app) values predicted from the Michaelis-Menten kinetics models were used (see below for V max(app) estimation). See SI methods, subsection “Calculations” for more information. Statistics for physiological data See SI methods, subsection “Statistics for physiological data” for a complete description of the methods used. RNA and DNA extraction, and sequencing Six cultures of M. tundripaludum SV96 T were acclimated at each of the temperatures, 8, 15, 21, or 27 °C, as described above. After acclimation, the cultures were cultivated to a sufficient cell density for extraction in exponential phase growth. Cultures at 8 and 27 °C were incubated for 36 h, and cultures at 15 and 21 °C for 16 h, prior to harvest for RNA and DNA extractions. See SI methods, subsection “RNA and DNA extraction, and sequencing” for more information. Computational analyses See SI methods, subsection “Computational analyses” for a complete description of the methods used to analyze the transcriptomes. Maps See SI methods, subsection “Maps” for a complete description of the methods used.
During stock culture maintenance, pre-incubations, and experiments, M. tundripaludum SV96 T , Methylobacter sp. G7 (unpublished), and Methylobacter luteus ACM 3304 T were cultivated in nitrate minimal salt (NMS) medium at pH 6.8 (see for trace element solution). See (SI) methods section, subsection “Cultures” for more information.
During acclimation, cells were incubated in NMS medium under ~80% air and ~20% CH 4 (100% CH 4 was injected into the air-containing bottle headspace) at 8, 15, 21, and 27 °C ( M. tundripaludum SV96 T and M. luteus ACM 3304 T ), or 4, 8, 15 and 21 °C ( Methylobacter sp. G7) (Fig. S ). The cultures were always allowed acclimation for the time needed to surpass 10 generations (7–12 days). During acclimation and experiments, cultures were shaken in a horizontal position, at 50 rpm. Some control experiments were carried out at 150 rpm to test the effect on mass transfer limitations. See SI methods, subsection “Acclimation” for more information.
Strains that consume gases can experience mass transfer limitations due to low substrate solubility. See SI methods, subsection “Shaking speed and mass transfer limitations.” for more information on how we accounted for mass transfer limitation.
From samples of acclimated exponential phase M. tundripaludum SV96 T cultures (75 ml) with densities of ~1–3 × 10 7 cells mL −1 22 mL were transferred to three 125 mL glass bottles per temperature. Bottles were then prepared with headspaces of 80% air and 20% CH 4 and a pressure of 1.3 atm. This resulted in 0.36–0.51 mM dissolved CH 4 , depending on the temperature, above the threshold of CH 4 saturation for M. tundripaludum SV96 T at this density (0.1 mM CH 4 ; Fig. S ). The bottles were incubated at 8, 15, 21, and 27 °C and 50 rpm (Fig. S ) or 150 rpm (Fig. S ) for up to 36 h. See SD B: Table B , B for incubation times and sampling time points for each experiment. At time zero and respective sampling intervals, 300 µL of cell suspension (in duplicate) was transferred to a Nunclon Delta Surface plate (Thermo Scientific, Waltham, MA, USA) and the optical density was measured (Spectra Max 250 microplate reader, Molecular Devices, San José, CA, USA) at 600 nm (diluted NMS medium as blank). For optical density measurements (OD 600 ), blanks were subtracted from measurements.
4 oxidation kinetics and growth kinetics Subsamples of acclimated exponential-phase cultures with densities of 5 × 10 7 cells mL −1 were aliquoted (21.6 mL) in 125 mL glass bottles for measurement of CH 4 oxidation and growth rates at seven different CH 4 concentrations and four different temperatures (8, 15, 21, and 27 °C for M. luteus ACM 3304 T and M. tundripaludum SV96 T , and 4, 8, 15, and 21 °C for Methylobacter sp. G7). We used three aliquots for each of the seven CH 4 concentrations, for a total of 21 bottles per temperature. With four temperatures, that amounted to a total of 84 bottles per strain (Fig. S ). For each temperature, two negative controls (medium without cells) in 125 mL glass bottles were prepared per CH 4 concentration (56 negative controls in total per strain). None of the negative controls indicated non-biological CH 4 consumption or leakage. To create the seven different CH 4 concentrations, seven volumes (200 µL; 600 µL; 1,5 mL; 3 mL; 6 mL; 12 mL; 15 mL) of 100% CH 4 were injected with a plastic syringe (BD Plastipak, Franklin Lakes, NJ, USA) with a sterile 0.5 × 16 mm needle (BD Microlance, Franklin Lakes, NJ, USA) from a multi-layer polypropylene gas bag (RESTEK, Bad Homburd vor der Höhe, Germany). Final headspace concentrations of CH 4 ranged between 1000 and 130,000 p.p.m.v. while dissolved concentrations of CH 4 ranged between 1.5 and 260 µM depending also on temperature. The gas pressures in the bottles were then adjusted to a total headspace pressure of ~1.3 atm at 20 °C by injecting additional volumes of air. See SI methods, subsection “CH 4 oxidation kinetics and growth kinetics” for more information.
The cells were acclimated and cultivated under 20% CH 4 in air as described above. Five replicate bottles with 20 mL culture and 20% CH 4 in air were incubated per temperature in darkness with 50 rpm shaking, five bottles with a starting cell concentration of ~1 × 10 7 cells mL −1 and five with a density of ~1 × 10 8 cells mL -1 . An incubation length of 48 h at 15 and 21 °C was sufficient to reach high-density exponential growth phase at 27 °C and 8 °C it took 60 and 90 h, respectively, to reach those same cell densities. During harvest, 2 mL of culture was sampled for glycogen measurements from the dense cultures and 10 mL from the dilute cultures. The samples were processed according to recommendations from the manufacturer (Glycogen assay kit, ab65620, Abcam, Cambridge, UK). Glycogen was quantified using a plate reader (GloMax Explorer, Promega, Madison, WI, USA). See SI methods, subsection “Glycogen quantification” for more information.
To normalize rates to cell numbers, standard curves correlating optical density (OD 600 ) to cell numbers were created for the three strains: M. tundripaludum SV96 T , M. luteus ACM 3304 T and Methylobacter sp. G7 For size estimation, cells were visualized using a Carl Zeiss AxioObserver Z1 with a 100x objective and Bright-Field and measured with the size estimation tool available in the AxioVision SE64 Rel 4.9.1 software. Cultures for size estimation were prepared by acclimation of cultures to 8 and 21 °C, as described above, followed by cultivation to reach exponential phase and harvest for size estimation. See SI methods, subsection “OD600 to cell number conversion and cell sizes” for more information.
Specific growth rates were calculated as the slope of the natural logarithm of optical densities against time, during exponential growth. Mixing ratios of CH 4 and CO 2 were calculated by comparison to certified standards. Masses of headspace and dissolved CH 4 and total CO 2 at different temperatures were calculated from the mixing ratios of CH 4 and CO 2 using Henry’s Law, assuming an ideal state, knowing the ambient pressure, temperature, headspace volume of the bottle, headspace pressure, liquid volume, and respective temperature-dependent solubility constants of the gases. All calculations accounted for removal of gas and liquid for measurements. We calculated the CH 4 oxidation and CO 2 production rates from slopes of linear models fitted to the concentrations measured during experiments and adjusted to the number of cells in the culture. Growth efficiencies were calculated by multiplying the specific growth (cell divisions cell −1 h −1 ) rate by 10 8 and then dividing by CH 4 the oxidation rates (µmol CH 4 oxidized 10 8 cells −1 h −1 ) to give cell divisions per µmol CH 4 oxidized. The data used to estimate growth efficiency were the values predicted for each measurement time-point from the non-linear Michaelis-Menten regression models. For estimates of specific growth rates, CH 4 oxidation, and growth efficiency at CH 4 saturation, V max(app) values predicted from the Michaelis-Menten kinetics models were used (see below for V max(app) estimation). See SI methods, subsection “Calculations” for more information.
See SI methods, subsection “Statistics for physiological data” for a complete description of the methods used.
Six cultures of M. tundripaludum SV96 T were acclimated at each of the temperatures, 8, 15, 21, or 27 °C, as described above. After acclimation, the cultures were cultivated to a sufficient cell density for extraction in exponential phase growth. Cultures at 8 and 27 °C were incubated for 36 h, and cultures at 15 and 21 °C for 16 h, prior to harvest for RNA and DNA extractions. See SI methods, subsection “RNA and DNA extraction, and sequencing” for more information.
See SI methods, subsection “Computational analyses” for a complete description of the methods used to analyze the transcriptomes.
See SI methods, subsection “Maps” for a complete description of the methods used.
Supplementary Information Supplementary dataset A Supplementary dataset B Supplementary dataset C
|
Life history strategies among soil bacteria—dichotomy for few, continuum for many | 7637dd7f-4f0a-473e-9b4e-49ea116e659f | 10030646 | Microbiology[mh] | The concept of copiotrophy and oligotrophy in microbial communities offers the potential for an organizing principle to describe the complexity of microbial systems. As such, it has been discussed for decades in relation to marine environments and soils . The framework is a direct descendent of r-K selection theory, which has long been applied to larger organisms . If successful, life history strategy frameworks should support inferences about processes from patterns in taxonomic composition. The copiotroph-oligotroph framework posits that microorganisms, facing strong selective pressure from their environment, adapt strategies defined by two endpoints along a continuum: either growing and reproducing quickly in the presence of abundant nutrients (copiotrophs), or specializing in resource-poor niches to escape from competition (oligotrophs) . Evidence for this framework in soils was first presented from greater relative abundance of some bacterial phyla in response to sucrose addition, indicating copiotrophic strategies, whereas other bacterial phyla were either unresponsive or responded negatively, suggesting oligotrophic strategies . An expansion of growth-trait associated strategies centers around the trade-off between maintenance energy, growth efficiency (i.e., yield), and growth rate or in the investment in resource acquisition . The copiotroph-oligotroph framework is commonly used for the interpretation of 16S rRNA gene bacterial community data [ – ]. Fierer et al. explicitly emphasized continuous and taxon-specific behavior, “These results do not suggest that every member of the Acidobacteria, β-Proteobacteria, and Bacteroidetes phyla are distinctly copiotrophic or oligotrophic” . Thus, the proposed framework in microbial ecology describes a continuum of nutrient responses. Viewed as a continuum, the copiotroph-oligotroph hypothesis holds that the traits associated with these two life history strategies are negatively correlated—not mutually exclusive—and that a continuous range of responses is expected. Additionally, factors beyond the physiological potential of the microorganism or available resources can influence population growth. Factors such as viral predation or soil structure are relevant to understanding microbal behavior in the soil habitat but may lead to a higher frequency of intermediate responses to nutrient addition (Fig. ). Such an outcome would raise serious objections to the usefulness of a dichotomous categorization scheme as a tool to understand and predict microbial communities in soil. It is also possible that the pressure to select for either copiotrophic or oligotrophic strategies is strong in soils but that the current practice of assignment of bacterial phyla to one strategy or the other is inaccurate (Fig. ) or imprecise (Fig. ). For example, classifying individual phyla as oligotrophic or copiotrophic may not be correct or useful given the behavior of the majority of their constituent taxa (i.e., categorization is inaccurate). Further, taxa within a given phylum may show a mix of response types (i.e., categorization is imprecise) according to previous publications . For example, if we assume that microorganisms must invest in one or the other growth strategy exclusively, we would expect to observe a bimodal distribution of growth responses in copiotrophic and a bimodal distribution in oligotrophic taxa. Although microbial ecologists typically organize their conclusions at the phylum level – it is well understood that bacterial lineages within phyla can have distinct metabolic and ecological roles and that finer taxonomic resolution may be necessary for assigning strategies accurately [ , – ]. This would indicate that current categorizations should be updated based on new methods for measuring taxon-specific growth rate or nutrient response, or that finer levels of taxonomic organization are more appropriate for making life history assignments as recommended by Ho and colleagues (based on references therein) . By contrast, if the life history strategy is coherent at the phylum level, and current classifications are correct, we expect to see a clear distinction in the distributions of growth response between copiotrophic and oligotrophic phyla (Fig. ). Here, we analyzed published quantitative stable isotope probing (qSIP) data [ , , ] on the growth of bacterial taxa within the complex and heterogenous soil environment. Because microbial activity is strongly dependent on resource stoichiometry [ – ], we analyzed data from two experimental treatments – labile carbon (C) or carbon + ammonium (N) addition. This design was meant to minimize N limitation in at least one treatment. We applied phylum-level assignments of life history strategy from the microbiological literature (Table ) and compared the expected responses of individual bacterial taxa to their growth responses measured by isotopic enrichment. Given previous findings of taxon-specific responses , we compared nutrient responses at different levels of taxonomic organization from the level of phylum to genus. Our hypotheses were (see Fig. ): H 0 , Null: No bimodality in nutrient response of taxa to nutrient addition (i.e., no clear selection for one strategy or the other exists, Fig. ) H 1 , misclassification: Bimodality in nutrient responses to nutrient addition within each life history category (i.e., phyla show inherent tendencies for one strategy or the other, but have been misclassified, Fig. ) H 2 , taxonomic resolution: Bimodality in nutrient responses to nutrient addition within each life history category (i.e., groups of taxa show inherent tendencies for one strategy or the other, but only at a finer taxonomic resolution than the phylum level, Fig. ) H 3 , phylum-level traits: Significant differences in nutrient response to nutrient addition between life history strategies (i.e., clear tendencies for one strategy or another; current phylum-level assignments of life history strategies are supported, Fig. )
Data were analyzed from samples collected, processed, and published previously [ , , ] and have been summarized here. The present analysis, which consisted of sequence data processing, the calculation of taxon-specific isotopic signatures, and subsequent analyses, reflects original work. Sample collection and isotope incubation To generate experimental data, three replicate soil samples were collected from the top 10 cm of plant-free patches in four ecosystems along the C. Hart Merriam elevation gradient in Northern Arizona. From low to high elevation, these sites are located in the following environments: desert grassland (GL; 1760 m), piñon-pine juniper woodland (PJ; 2020 m), ponderosa pine forest (PP; 2344 m), and mixed conifer forest (MC; 2620 m). Soil samples were air-dried for 24 h at room temperature, homogenized, and passed through a 2 mm sieve before being stored at 4 °C for another 24 h. This produced three distinct but homogenous soil samples from each of the four ecosystems that were subject to experimental treatments. Three treatments were applied to bring soils to 70% water-holding capacity: water alone (control), water with glucose (C treatment; 1000 µg C g −1 dry soil), or water with glucose and a nitrogen source (CN treatment; [NH 4 ] 2 SO 4 at 100 µg N g −1 dry soil). To track growth through isotope assimilation, both 18 O-enriched water (97 atom %) and 13 C-enriched glucose (99 atom %) were used. In all treatments isotopically heavy samples were paired with matching “light” samples that received water with a natural abundance isotope signatures. For 18 O incubations, this design resulted in three soil samples per ecosystem per treatment (across four ecosystems and three treatments, n = 36) while 13 C incubations were limited to only C and CN treatments ( n = 24). Previous analyses suggest that three replicates is sufficient to detect growth of 10 atom % 18 O in microbial DNA with a power of 0.6 and a growth of 5 atom % 18 O with a power of 0.3 (12 and 6 atom % respectively for 13 C) . All soils were incubated in the dark for one week. Following incubation, soils were frozen at −80 °C for one week prior to DNA extraction. Quantitative stable isotope probing The procedure of qSIP (quantitative stable isotope probing) is described here but has been applied to these samples as previously published [ , , ]. DNA extraction was performed on soils using a DNeasy PowerSoil HTP 96 Kit (MoBio Laboratories, Carlsbad, CA, USA) and following manufacturer’s protocol. Briefly, 0.25 g of soils from each sample were carefully added to deep, 96-well plates containing zirconium dioxide beads and a cell lysis solution with sodium dodecyl sulfate (SDS) and shaken for 20 min. Following cell lysis, supernatant was collected and centrifuged three times in fresh 96-well plates with reagents separating DNA from non-DNA organic and inorganic materials. Lastly, DNA samples were collected on silica filter plates, rinsed with ethanol and eluted into 100 µL of a 10 mM Tris buffer in clean 96-well plates. To quantify the degree of 18 O or 13 C isotope incorporation into bacterial DNA (excess atom fraction or EAF), the qSIP protocol was used, though modified slightly as reported previously [ , , ]. Briefly, microbial growth was quantified as the change in DNA buoyant density due to incorporation of the 18 O or 13 C isotopes through the method of density fractionation by adding 1 µg of DNA to 2.6 mL of saturated CsCl solution in combination with a gradient buffer (200 mM Tris, 200 mM KCL, 2 mM EDTA) in a 3.3 mL OptiSeal ultracentrifuge tube (Beckman Coulter, Fullerton, CA, USA). The solution was centrifuged to produce a gradient of increasingly labeled (heavier) DNA in an Optima Max bench top ultracentrifuge (Beckman Coulter, Brea, CA, USA) with a Beckman TLN-100 rotor (127,000 × g for 72 h) at 18 °C. Each post-incubation sample was thus converted from a continuous gradient into approximately 20 fractions (150 µL) using a modified fraction recovery system (Beckman Coulter). The density of each fraction was measured with a Reichart AR200 digital refractometer (Reichert Analytical Instruments, Depew, NY, USA). Fractions with densities between 1.640 and 1.735 g cm −3 were retained as densities outside this range generally did not contain DNA. In all retained fractions, DNA was cleaned and purified using isopropanol precipitation and the abundance of bacterial 16S rRNA gene copies was quantified with qPCR using primers specific to bacterial 16S rRNA genes ( Eub 515F: AAT GAT ACG GCG ACC ACC GAG TGC CAG CMG CCG CGG TAA, 806R: CAA GCA GAA GAC GGC ATA CGA GGA CTA CVS GGG TAT CTA AT). Triplicate reactions were 8 µL consisting of 0.2 mM of each primer, 0.01 U µL −1 Phusion HotStart II Polymerase (Thermo Fisher Scientific, Waltham, MA), 1× Phusion HF buffer (Thermo Fisher Scientific), 3.0 mM MgCl 2 , 6% glycerol, and 200 µL of dNTPs. Reactions were performed on a CFX384 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA) under the following cycling conditions: 95 °C at 1 min and 44 cycles at 95 °C (30 s), 64.5 °C (30 s), and 72 °C (1 min). Separate from qPCR, retained sample-fractions were subject to a similar amplification step of the 16S rRNA gene V4 region (515F: GTG YCA GCM GCC GCG GTA A, 806R: GGA CTA CNV GGG TWT CTA AT) in preparation for sequencing with the same reaction mix but differing cycle conditions – 95 °C for 2 min followed by 15 cycles at 95 °C (30 s), 55 °C (30 s), and 60 °C (4 min). The resulting 16S rRNA gene V4 amplicons were sequenced on a MiSeq sequencing platform (Illumina, Inc., San Diego, CA, USA). DNA sequence data and sample metadata have been deposited in the NCBI Sequence Read Archive under the project ID PRJNA521534. Sequence processing and qSIP analysis Independently from previous publications, we processed raw sequence data of forward and reverse reads (FASTQ) within the QIIME2 environment (release 2018.6) and denoised sequences within QIIME2 using the DADA2 pipeline . We clustered the remaining sequences into amplicon sequence variants (ASVs, at 100% sequence identity) against the SILVA 138 database using a pre-trained open-reference Naïve Bayes feature classifier . We removed samples with less than 3000 sequence reads, non-bacterial lineages, and global singletons and doubletons. We converted ASV sequencing abundances in each fraction to the number of 16S rRNA gene copies per gram dry soil based on qPCR abundances and the known amount of dry soil equivalent added to the initial extraction. This allowed us to express absolute population densities, rather than relative abundances. Across all replicates, we identified 114 543 unique bacterial ASVs. We calculated the 18 O and 13 C excess atom fraction (EAF) for each bacterial ASV using R version 4.0.3 and data.table with custom scripts available at https://www.github.com/bramstone/ . Negative enrichment values were corrected using previously published methods . ASVs that appeared in less than two of the three replicates of an ecosystem-treatment combination ( n = 3) and less than three density fractions within those two replicates were removed to avoid assigning spurious estimates of isotope enrichment to infrequent taxa. Any ASVs filtered out of one ecosystem-treatment group were allowed to be present in another if they met the frequency threshold. Applying these filtering criteria, we limited our analysis towards 3759 unique bacterial ASVs which accounted for a small proportion of the total diversity but represented 68.0% of all sequence reads, and encompassed most major bacterial groups (Supplementary Fig. ). Analysis of life history strategies and nutrient response All statistical tests were conducted in R version 4.0.3 . We assessed the ability of phylum-level assignment of life history strategy to predict growth in response to C and N addition, as proxied by the incorporation of heavy isotope during DNA replication . Phylum-level assignments (Table ) were based on the most frequently observed behavior of lineages with a representative phylum (or subphylum) as compiled previously . We averaged 18 O EAF values of bacterial taxa for each treatment and ecosystem and then subtracted the values in control soils from values in C-amended soils to determine C response (∆ 18 O EAF C ) and from the 18 O EAF of bacteria in CN-amended soils to determine C and N response (Δ 18 O EAF CN ). Because an ASV must have a measurable EAF in both the control and treatment for a valid Δ 18 O EAF to be calculated, we were only able to resolve the nutrient response for 2044 bacterial ASVs – 1906 in response to C addition and 1427 in response to CN addition. We used Gaussian finite mixture modeling, as implemented by the mclust R package , to demarcate plausible multi-isotopic signatures for oligotrophs and copiotrophs. For each treatment, we calculated average per-taxon 13 C and 18 O EAF values. To compare both isotopes directly, we divided 18 O EAF values by 0.6 based on the estimate that this value (designated as µ ) represents the fraction of oxygen atoms in DNA derived from the 18 O-water, rather than from 16 O within available C sources . Two mixture components, corresponding to oligotrophic and copiotrophic growth modes, were defined using the Mclust function using ellipsoids of equal volume and shape. We observed several microorganisms with high 18 O enrichment but comparatively low 13 C enrichment, potentially indicating growth following the depletion of the added glucose, and that were reasonably clustered as oligotrophs in our mixture model. We tested how frequently mixture model clustering of each microorganism’s growth (based on average 18 O– 13 C EAF in a treatment) could predict its growth across replicates ( n = 12 in each treatment—although individual). We applied the treatment-level mixture models defined above to the per-taxon isotope values in each replicate, recording when a microorganism’s life history strategy in a replicate agreed with the treatment-level cluster, and when it didn’t. We used exact binomial tests to test whether the number of “successes” (defined as a microorganism being grouped in the same life history category as its treatment-level cluster) was statistically significant. To account for type I error across all individual tests (one per ASV per treatment), we adjusted P values in each treatment using the false-discovery rate (FDR) method . To determine the extent that life history categorizations may be appropriately applied at finer levels of taxonomic resolution, we constructed several hierarchical linear models using the lmer function in the nlme package version 3.1-149 . To condense growth information from both isotopes into a single analysis, 18 O and 13 C EAF values were combined into a single variable using principal components analysis separately for each treatment. Across the C and CN treatments, the first principal component (PC1) was able to explain – respectively – 86% and 91% of joint variation of 18 O and 13 C EAF values. In all cases, we applied PC1 as the response variable and treated taxonomy and ecosystem as random model terms to limit the potential of pseudo-replication to bias significance values. We used likelihood ratio analysis and Akaike information criterion (AIC) values to compare models where life history strategy was determined based on observed nutrient responses at different taxonomic levels (Eq. ) against a model with the same random terms but without any life history strategy data (Eq. ). Separate models were applied to each treatment. To reduce model overfitting, we removed families represented by fewer than three bacterial ASVs as well as phyla represented by only one order. In addition, we removed bacterial ASVs with unknown taxonomic assignments (following Morrissey et al. ). This limited our analysis to 1 049 ASVs in the C amendment and 984 in the CN amendment. 1 [12pt]{minimal}
$${{{{{}}}}}{1}_{{18{{{{{}}}}} - 13{{{{{}}}}}}} {{{{{}}}}} + 1|{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}$$ PC 1 18 O − 13 C ~ strategy + 1 ∣ phylum / class / order / family / genus / eco 2 [12pt]{minimal}
$${{{{{}}}}}{1}_{{18{{{{{}}}}} - 13{{{{{}}}}}}} 1 + 1|{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}$$ PC 1 18 O − 13 C ~ 1 + 1 ∣ phylum / class / order / family / genus / eco Here, life history strategy was defined at each taxonomic level using the mixture models above and based on the mean 18 O and 13 C EAF values of each bacterial lineage (Supplemental Fig. ). We compared these models with the no-strategy model (Eq. ) directly using likelihood ratio testing.
To generate experimental data, three replicate soil samples were collected from the top 10 cm of plant-free patches in four ecosystems along the C. Hart Merriam elevation gradient in Northern Arizona. From low to high elevation, these sites are located in the following environments: desert grassland (GL; 1760 m), piñon-pine juniper woodland (PJ; 2020 m), ponderosa pine forest (PP; 2344 m), and mixed conifer forest (MC; 2620 m). Soil samples were air-dried for 24 h at room temperature, homogenized, and passed through a 2 mm sieve before being stored at 4 °C for another 24 h. This produced three distinct but homogenous soil samples from each of the four ecosystems that were subject to experimental treatments. Three treatments were applied to bring soils to 70% water-holding capacity: water alone (control), water with glucose (C treatment; 1000 µg C g −1 dry soil), or water with glucose and a nitrogen source (CN treatment; [NH 4 ] 2 SO 4 at 100 µg N g −1 dry soil). To track growth through isotope assimilation, both 18 O-enriched water (97 atom %) and 13 C-enriched glucose (99 atom %) were used. In all treatments isotopically heavy samples were paired with matching “light” samples that received water with a natural abundance isotope signatures. For 18 O incubations, this design resulted in three soil samples per ecosystem per treatment (across four ecosystems and three treatments, n = 36) while 13 C incubations were limited to only C and CN treatments ( n = 24). Previous analyses suggest that three replicates is sufficient to detect growth of 10 atom % 18 O in microbial DNA with a power of 0.6 and a growth of 5 atom % 18 O with a power of 0.3 (12 and 6 atom % respectively for 13 C) . All soils were incubated in the dark for one week. Following incubation, soils were frozen at −80 °C for one week prior to DNA extraction.
The procedure of qSIP (quantitative stable isotope probing) is described here but has been applied to these samples as previously published [ , , ]. DNA extraction was performed on soils using a DNeasy PowerSoil HTP 96 Kit (MoBio Laboratories, Carlsbad, CA, USA) and following manufacturer’s protocol. Briefly, 0.25 g of soils from each sample were carefully added to deep, 96-well plates containing zirconium dioxide beads and a cell lysis solution with sodium dodecyl sulfate (SDS) and shaken for 20 min. Following cell lysis, supernatant was collected and centrifuged three times in fresh 96-well plates with reagents separating DNA from non-DNA organic and inorganic materials. Lastly, DNA samples were collected on silica filter plates, rinsed with ethanol and eluted into 100 µL of a 10 mM Tris buffer in clean 96-well plates. To quantify the degree of 18 O or 13 C isotope incorporation into bacterial DNA (excess atom fraction or EAF), the qSIP protocol was used, though modified slightly as reported previously [ , , ]. Briefly, microbial growth was quantified as the change in DNA buoyant density due to incorporation of the 18 O or 13 C isotopes through the method of density fractionation by adding 1 µg of DNA to 2.6 mL of saturated CsCl solution in combination with a gradient buffer (200 mM Tris, 200 mM KCL, 2 mM EDTA) in a 3.3 mL OptiSeal ultracentrifuge tube (Beckman Coulter, Fullerton, CA, USA). The solution was centrifuged to produce a gradient of increasingly labeled (heavier) DNA in an Optima Max bench top ultracentrifuge (Beckman Coulter, Brea, CA, USA) with a Beckman TLN-100 rotor (127,000 × g for 72 h) at 18 °C. Each post-incubation sample was thus converted from a continuous gradient into approximately 20 fractions (150 µL) using a modified fraction recovery system (Beckman Coulter). The density of each fraction was measured with a Reichart AR200 digital refractometer (Reichert Analytical Instruments, Depew, NY, USA). Fractions with densities between 1.640 and 1.735 g cm −3 were retained as densities outside this range generally did not contain DNA. In all retained fractions, DNA was cleaned and purified using isopropanol precipitation and the abundance of bacterial 16S rRNA gene copies was quantified with qPCR using primers specific to bacterial 16S rRNA genes ( Eub 515F: AAT GAT ACG GCG ACC ACC GAG TGC CAG CMG CCG CGG TAA, 806R: CAA GCA GAA GAC GGC ATA CGA GGA CTA CVS GGG TAT CTA AT). Triplicate reactions were 8 µL consisting of 0.2 mM of each primer, 0.01 U µL −1 Phusion HotStart II Polymerase (Thermo Fisher Scientific, Waltham, MA), 1× Phusion HF buffer (Thermo Fisher Scientific), 3.0 mM MgCl 2 , 6% glycerol, and 200 µL of dNTPs. Reactions were performed on a CFX384 Touch Real-Time PCR Detection System (Bio-Rad, Hercules, CA, USA) under the following cycling conditions: 95 °C at 1 min and 44 cycles at 95 °C (30 s), 64.5 °C (30 s), and 72 °C (1 min). Separate from qPCR, retained sample-fractions were subject to a similar amplification step of the 16S rRNA gene V4 region (515F: GTG YCA GCM GCC GCG GTA A, 806R: GGA CTA CNV GGG TWT CTA AT) in preparation for sequencing with the same reaction mix but differing cycle conditions – 95 °C for 2 min followed by 15 cycles at 95 °C (30 s), 55 °C (30 s), and 60 °C (4 min). The resulting 16S rRNA gene V4 amplicons were sequenced on a MiSeq sequencing platform (Illumina, Inc., San Diego, CA, USA). DNA sequence data and sample metadata have been deposited in the NCBI Sequence Read Archive under the project ID PRJNA521534.
Independently from previous publications, we processed raw sequence data of forward and reverse reads (FASTQ) within the QIIME2 environment (release 2018.6) and denoised sequences within QIIME2 using the DADA2 pipeline . We clustered the remaining sequences into amplicon sequence variants (ASVs, at 100% sequence identity) against the SILVA 138 database using a pre-trained open-reference Naïve Bayes feature classifier . We removed samples with less than 3000 sequence reads, non-bacterial lineages, and global singletons and doubletons. We converted ASV sequencing abundances in each fraction to the number of 16S rRNA gene copies per gram dry soil based on qPCR abundances and the known amount of dry soil equivalent added to the initial extraction. This allowed us to express absolute population densities, rather than relative abundances. Across all replicates, we identified 114 543 unique bacterial ASVs. We calculated the 18 O and 13 C excess atom fraction (EAF) for each bacterial ASV using R version 4.0.3 and data.table with custom scripts available at https://www.github.com/bramstone/ . Negative enrichment values were corrected using previously published methods . ASVs that appeared in less than two of the three replicates of an ecosystem-treatment combination ( n = 3) and less than three density fractions within those two replicates were removed to avoid assigning spurious estimates of isotope enrichment to infrequent taxa. Any ASVs filtered out of one ecosystem-treatment group were allowed to be present in another if they met the frequency threshold. Applying these filtering criteria, we limited our analysis towards 3759 unique bacterial ASVs which accounted for a small proportion of the total diversity but represented 68.0% of all sequence reads, and encompassed most major bacterial groups (Supplementary Fig. ).
All statistical tests were conducted in R version 4.0.3 . We assessed the ability of phylum-level assignment of life history strategy to predict growth in response to C and N addition, as proxied by the incorporation of heavy isotope during DNA replication . Phylum-level assignments (Table ) were based on the most frequently observed behavior of lineages with a representative phylum (or subphylum) as compiled previously . We averaged 18 O EAF values of bacterial taxa for each treatment and ecosystem and then subtracted the values in control soils from values in C-amended soils to determine C response (∆ 18 O EAF C ) and from the 18 O EAF of bacteria in CN-amended soils to determine C and N response (Δ 18 O EAF CN ). Because an ASV must have a measurable EAF in both the control and treatment for a valid Δ 18 O EAF to be calculated, we were only able to resolve the nutrient response for 2044 bacterial ASVs – 1906 in response to C addition and 1427 in response to CN addition. We used Gaussian finite mixture modeling, as implemented by the mclust R package , to demarcate plausible multi-isotopic signatures for oligotrophs and copiotrophs. For each treatment, we calculated average per-taxon 13 C and 18 O EAF values. To compare both isotopes directly, we divided 18 O EAF values by 0.6 based on the estimate that this value (designated as µ ) represents the fraction of oxygen atoms in DNA derived from the 18 O-water, rather than from 16 O within available C sources . Two mixture components, corresponding to oligotrophic and copiotrophic growth modes, were defined using the Mclust function using ellipsoids of equal volume and shape. We observed several microorganisms with high 18 O enrichment but comparatively low 13 C enrichment, potentially indicating growth following the depletion of the added glucose, and that were reasonably clustered as oligotrophs in our mixture model. We tested how frequently mixture model clustering of each microorganism’s growth (based on average 18 O– 13 C EAF in a treatment) could predict its growth across replicates ( n = 12 in each treatment—although individual). We applied the treatment-level mixture models defined above to the per-taxon isotope values in each replicate, recording when a microorganism’s life history strategy in a replicate agreed with the treatment-level cluster, and when it didn’t. We used exact binomial tests to test whether the number of “successes” (defined as a microorganism being grouped in the same life history category as its treatment-level cluster) was statistically significant. To account for type I error across all individual tests (one per ASV per treatment), we adjusted P values in each treatment using the false-discovery rate (FDR) method . To determine the extent that life history categorizations may be appropriately applied at finer levels of taxonomic resolution, we constructed several hierarchical linear models using the lmer function in the nlme package version 3.1-149 . To condense growth information from both isotopes into a single analysis, 18 O and 13 C EAF values were combined into a single variable using principal components analysis separately for each treatment. Across the C and CN treatments, the first principal component (PC1) was able to explain – respectively – 86% and 91% of joint variation of 18 O and 13 C EAF values. In all cases, we applied PC1 as the response variable and treated taxonomy and ecosystem as random model terms to limit the potential of pseudo-replication to bias significance values. We used likelihood ratio analysis and Akaike information criterion (AIC) values to compare models where life history strategy was determined based on observed nutrient responses at different taxonomic levels (Eq. ) against a model with the same random terms but without any life history strategy data (Eq. ). Separate models were applied to each treatment. To reduce model overfitting, we removed families represented by fewer than three bacterial ASVs as well as phyla represented by only one order. In addition, we removed bacterial ASVs with unknown taxonomic assignments (following Morrissey et al. ). This limited our analysis to 1 049 ASVs in the C amendment and 984 in the CN amendment. 1 [12pt]{minimal}
$${{{{{}}}}}{1}_{{18{{{{{}}}}} - 13{{{{{}}}}}}} {{{{{}}}}} + 1|{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}$$ PC 1 18 O − 13 C ~ strategy + 1 ∣ phylum / class / order / family / genus / eco 2 [12pt]{minimal}
$${{{{{}}}}}{1}_{{18{{{{{}}}}} - 13{{{{{}}}}}}} 1 + 1|{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}/{{{{{}}}}}$$ PC 1 18 O − 13 C ~ 1 + 1 ∣ phylum / class / order / family / genus / eco Here, life history strategy was defined at each taxonomic level using the mixture models above and based on the mean 18 O and 13 C EAF values of each bacterial lineage (Supplemental Fig. ). We compared these models with the no-strategy model (Eq. ) directly using likelihood ratio testing.
Bacteria with strongly positive short-term nutrient response represent a small proportion of diversity within a limited number of phyla When comparing the difference in isotope assimilation of bacterial taxa in response to nutrients, we observed substantial overlap between the response of expected oligotrophs and expected copiotrophs and little bimodal tendency either across all phyla or within phyla (Fig. ), aligning with hypothesis H 0 (Fig. ). Accounting for shared taxonomy and differences across sites using hierarchical linear models, expected life history strategies (copiotrophic, oligotrophic, or undefined; Table ) were a non-significant predictor of individual bacterial responses to nutrients regardless of treatment or isotopic tracer (Δ 18 O C: F 2 , 8 = 0.94, P = 0.43; Δ 18 O CN: F 2 , 8 = 1.80, P = 0.23; 13 C: F 2 , 8 = 1.49, P = 0.28; 13 CN: F 2 , 8 = 2.81, P = 0.12). However, we did observe ASVs with strong positive responses to C and CN addition, despite the prevailing unimodal pattern (Fig. B, D); and these ASVs tended to come from lineages with expectations for copiotrophic growth . Bacterial ASVs with strongly positive nutrient response were constituents of the Gammaproteobacteria, Alphaproteobacteria, Actinobacteria, Firmicutes, and Bacteroidetes although they made up small proportions of each phylum (with the exception of the Firmicutes) (Fig. ). The differential response of ASVs within some phyla suggests support for hypothesis H3 (Fig. ). Mixture models produce plausible delinations of life history strategy, but identify few consistently copiotrophic taxa Bivariate Gaussian finite mixture modeling of joint 18 O- 13 C growth signatures of bacterial ASVs produced clusters with similar configurations in both treatments (Fig. ). Slow-growing (i.e., oligotrophic) ASVs had mean enrichment values under 0.15 ( [12pt]{minimal}
$$ x$$ x ¯ 18O = 0.12, [12pt]{minimal}
$$ x$$ x ¯ 13C = 0.07) while fast-growing (i.e., copiotrophic) ASVs had mean enrichment values greater than 0.3 ( [12pt]{minimal}
$$ x$$ x ¯ 18O = 0.32, [12pt]{minimal}
$$ x$$ x ¯ 13C = 0.38). In both the C and CN treatments, most bacterial ASVs (>90%) were clustered into the oligotrophic growth category defined by the mixture model (as based on their treatment-averaged 13 C and 18 O enrichment values) (Fig. ). Bacteria generally behaved consistently across replicates, but few ASVs could be identified as soley copiotrophic or oligotrophic with a high degree of statistical confidence. Nevertheless, assignment of bacterial ASVs into different life history clusters, based on groupings from mixture models (Fig. ), provided a clearer demarcation of behavior than expectations of life history strategy from the literature (Fig. ). Although, the vast majority of ASVs either clustered into some mix of copiotrophic and oligotrophic responses, or occurred too infrequently to assign a single life history strategy that was statistically significant. Per-replicate behavior significantly matched with treatment-level expectations for only 28 ASVs based on exact binomial tests (Supplemental Table ). Of those, only three ASVs could be significantly grouped into the copiotrophic cluster (Firmicutes: Paenibacillus and an unclassified genus within the order Bacillales, Actinobacteria: an unclassified genus within the Micrococcaceae) while the remaining occurred in the oligotrophic cluster. Life history strategy at fine taxonomic levels is necessary to accurately describe taxon-specific nutrient response Comparison of hierarchical linear models with life history categorizations at different taxonomic resolution indicated that finer levels were more predictive of nutrient response behavior. Assignment at the phylum and class levels, based on multi-isotopic mixture modeling clusters (Fig. ), produced nearly identical models that were both significantly better than site and taxonomic information alone (likelihood ratio tests; C response: L = 12.5, P < 0.001; CN response: L = 10.0, P = 0.0015), providing evidence against hypothesis H 0 (Table ). In both the phylum-level and class-level models, the Firmicutes and Bacilli (as a class within the Firmicutes) were the only respective lineages designated as copiotrophic, despite many ASVs showing strongly positive enrichment (Fig. ). At finer taxonomic resolution, the number of lineages designated as copiotrophic broadened, and models were stronger predictors of bacterial growth (Table ) (Fig. ). The strongest improvement was at the genus level (likelihood ratio tests; C response: order L = 39.7, P < 0.001; family L = 54.7, P < 0.001; genus L = 105.0, P < 0.001; CN response: order L = 39.3, P < 0.001; family L = 64.3, P < 0.001; genus L = 168.8, P < 0.001). This provides strong support for hypothesis H 2 (Fig. ) in that life history assignments of bacterial genera may be useful in predicting nutrient response (Supplemental Figs. , ) (Supplemental Data ).
When comparing the difference in isotope assimilation of bacterial taxa in response to nutrients, we observed substantial overlap between the response of expected oligotrophs and expected copiotrophs and little bimodal tendency either across all phyla or within phyla (Fig. ), aligning with hypothesis H 0 (Fig. ). Accounting for shared taxonomy and differences across sites using hierarchical linear models, expected life history strategies (copiotrophic, oligotrophic, or undefined; Table ) were a non-significant predictor of individual bacterial responses to nutrients regardless of treatment or isotopic tracer (Δ 18 O C: F 2 , 8 = 0.94, P = 0.43; Δ 18 O CN: F 2 , 8 = 1.80, P = 0.23; 13 C: F 2 , 8 = 1.49, P = 0.28; 13 CN: F 2 , 8 = 2.81, P = 0.12). However, we did observe ASVs with strong positive responses to C and CN addition, despite the prevailing unimodal pattern (Fig. B, D); and these ASVs tended to come from lineages with expectations for copiotrophic growth . Bacterial ASVs with strongly positive nutrient response were constituents of the Gammaproteobacteria, Alphaproteobacteria, Actinobacteria, Firmicutes, and Bacteroidetes although they made up small proportions of each phylum (with the exception of the Firmicutes) (Fig. ). The differential response of ASVs within some phyla suggests support for hypothesis H3 (Fig. ).
Bivariate Gaussian finite mixture modeling of joint 18 O- 13 C growth signatures of bacterial ASVs produced clusters with similar configurations in both treatments (Fig. ). Slow-growing (i.e., oligotrophic) ASVs had mean enrichment values under 0.15 ( [12pt]{minimal}
$$ x$$ x ¯ 18O = 0.12, [12pt]{minimal}
$$ x$$ x ¯ 13C = 0.07) while fast-growing (i.e., copiotrophic) ASVs had mean enrichment values greater than 0.3 ( [12pt]{minimal}
$$ x$$ x ¯ 18O = 0.32, [12pt]{minimal}
$$ x$$ x ¯ 13C = 0.38). In both the C and CN treatments, most bacterial ASVs (>90%) were clustered into the oligotrophic growth category defined by the mixture model (as based on their treatment-averaged 13 C and 18 O enrichment values) (Fig. ). Bacteria generally behaved consistently across replicates, but few ASVs could be identified as soley copiotrophic or oligotrophic with a high degree of statistical confidence. Nevertheless, assignment of bacterial ASVs into different life history clusters, based on groupings from mixture models (Fig. ), provided a clearer demarcation of behavior than expectations of life history strategy from the literature (Fig. ). Although, the vast majority of ASVs either clustered into some mix of copiotrophic and oligotrophic responses, or occurred too infrequently to assign a single life history strategy that was statistically significant. Per-replicate behavior significantly matched with treatment-level expectations for only 28 ASVs based on exact binomial tests (Supplemental Table ). Of those, only three ASVs could be significantly grouped into the copiotrophic cluster (Firmicutes: Paenibacillus and an unclassified genus within the order Bacillales, Actinobacteria: an unclassified genus within the Micrococcaceae) while the remaining occurred in the oligotrophic cluster.
Comparison of hierarchical linear models with life history categorizations at different taxonomic resolution indicated that finer levels were more predictive of nutrient response behavior. Assignment at the phylum and class levels, based on multi-isotopic mixture modeling clusters (Fig. ), produced nearly identical models that were both significantly better than site and taxonomic information alone (likelihood ratio tests; C response: L = 12.5, P < 0.001; CN response: L = 10.0, P = 0.0015), providing evidence against hypothesis H 0 (Table ). In both the phylum-level and class-level models, the Firmicutes and Bacilli (as a class within the Firmicutes) were the only respective lineages designated as copiotrophic, despite many ASVs showing strongly positive enrichment (Fig. ). At finer taxonomic resolution, the number of lineages designated as copiotrophic broadened, and models were stronger predictors of bacterial growth (Table ) (Fig. ). The strongest improvement was at the genus level (likelihood ratio tests; C response: order L = 39.7, P < 0.001; family L = 54.7, P < 0.001; genus L = 105.0, P < 0.001; CN response: order L = 39.3, P < 0.001; family L = 64.3, P < 0.001; genus L = 168.8, P < 0.001). This provides strong support for hypothesis H 2 (Fig. ) in that life history assignments of bacterial genera may be useful in predicting nutrient response (Supplemental Figs. , ) (Supplemental Data ).
Our results indicate that microbial life history strategies, as currently conceptualized, do not provide a strong predictive framework on the behavior and activity of most microorganisms in the soil environment. Rather, under the conditions of this experiment, microorganisms exhibited a continuous distribution from copiotrophic to oligotrophic strategies as represented from high to low growth rates, with most microorganisms showing low to intermediate growth rates. While this suggests support for our null hypothesis, H 0 , we found evidence for bimodality when considering the growth rates of a small subset of ASVs in the community which in several cases were also highly abundant. With the exception of the Firmicutes, we observed little evidence distinguishing any bacterial phylum as strongly copiotrophic or oligotrophic (as expected under hypotheses H 1 and H 3 ), indicating that assumptions about the behavior of any particular taxon cannot be made on the basis of its representative phylum. Fierer et al. note as much in their seminal work . Copiotrophic and oligotrophic modes are roughly analogous to nutrient acquisition or stress tolerance strategies (respectively) within the yield-acquisition-stress tolerance framework . The assumptions of the YAS framework indicate that more complex substrates may be better suited to differentiating nutrient acquisitive microorganisms (copiotrophs) from others. While the labile nutrients supplied in this study, glucose and ammonium sulfate, were intended to serve as a proxy for plant root exudates in a priming experiment , they did not truly represent the diversity and complexity of native substrates that would be expected in a copiotrophic soil environment. Thus, the ability of this study to address broad hypotheses about life history strategies across the bacterial tree of life may be limited. We saw significant improvement in hierarchical linear models of nutrient response when life history strategy was estimated at finer levels of taxonomic resolution (e.g., family and especially genus) which indicates strong support for hypothesis H 2 . Therefore, while we refute the continued use of categorical assumptions of oligotrophic or copiotrophic life histories for bacterial phyla, our in situ findings suggest that such representations could be useful if made at the genus level (although perhaps only in the context of artificial resource amendment) – in agreement with previously reported conclusions . Despite the best performance of the genus-level models to estimate nutrient responses, we had difficulty confidently characterizing the growth of individual ASVs; few could be consistently labeled as copiotrophs because many grew both quickly and slowly across the replicates in our experiment. Such indeterminancy is partially due to the low sample size of our experiment, but also likely stems from the inherent stochasticity of the soil environment. This context-dependency of bacterial responses (either by nutrient complexity and character or by local ecosystem characteristics) is another argument against categorical application of life history strategy at a broad taxonomic level. For example, it will be difficult to predict the growth of a “copiotroph” if its behavior depends on a complex arrangement of soil characteristics, nutrient availability, and local biotic interactions rather than more relatively static traits such as 16S rRNA gene copy number or genome size. The soils used in the current experiment were subject to considerable disturbance including physical disruption, dry-down, and sudden wet-up, inducing a strong pattern of microbial turnover, activity, and respiration from new organic matter made available to the soil community [ , – ]. As such, we employed 13 C-glucose additions to track its utilization specifically. We found high shared variation between 18 O and 13 C EAF values, suggesting that most microorganisms utilized the added glucose. Further, 18 O and 13 C EAF values covaried more strongly in the CN treatment, suggesting that N limitation may have limited glucose uptake in the C treatment. Thus, microorganisms with high 18 O signatures but low 13 C signatures may be those with high N demand who prioritized the decomposition of native soil organic matter to meet their needs. However, the presence of microorganisms with high 18 O but low 13 C signatures in the CN treatment suggests that N limitation alone may not explain why some microorganisms did not utilize the added glucose. Besides mixing, successional dynamics across the incubation may also explain differences in isotopic signatures of bacteria. Thus, we took this into account in our mixture model specifications such that these microorganisms (high 18 O but low 13 C EAF values) were clustered as oligotrophic, based on the possibility that they grew after the depletion of added glucose. Our results do suggest that – in our aerobic mineral soils at least – the potential for quick growth in response to labile nutrients exists within a small portion of the bacterial community and that this potential seems to be phylum-specific. Among these phyla, however, it is more accurate to understand nutrient response as a continuum rather than a dichotomous classification. If classification is necessary for statistical or narrative purposes, we recommend to restrict life history designations to the family or genus level. These findings (produced by qSIP) were measured by within the context of microbial community interactions which is an important line of inference to better understand microbial trait adaptations. In keeping with other ecological frameworks (e.g., C-S-R and Y-A-S ), stress treatments are a priority for future studies in order to understand the diversity of stress tolerance strategies and their effect on growth. The utilization of both simple and complex nutrient sources across the community (as well as from both plant and microbial origin) will also be a key point of inquiry, and designs that explore this difference will refine our thinking of microbial ecology in the soil realm (e.g., Dang et al. ). Lastly, the relatively short timescales inherent to nutrient pulse-type experiments mean that such incubations must be placed into longer-term studies strategically. For example, repeated or long-term amendments with both qSIP and complimentary 16S rRNA gene surveys can show how well short-term growth rates relate to stable community adaptation. Plant ecologists have embraced trait-based approaches, such as the application of leaf economic spectrum as an important predictor of global carbon flux within a larger framework of interrelated trait dimensions and trade-offs . Correspondingly, future trait explorations in microbial ecology should also be paired with measures of nutrient and energy fluxes to link community composition with ecosystem dynamics.
Supplemental Information Bacterial Life History Strategy and Taxonomy Experimental data
|
Correlations between care users’ and the healthcare inspectorate’s ratings of the quality of care in long-term care homes | 96428199-77c9-4b56-90d4-e75a3686aa58 | 10030658 | Patient-Centered Care[mh] | Healthcare inspectorates value the importance of care users’ experiences of the quality of care provided in long-term care (LTC) homes, however, they struggle to address these experiences in regulatory practice. Earlier studies showed weak to low-to-medium positive correlations between care users’ and inspectorates’ ratings of the quality of care provided by LTC homes.
Care users’ ratings of LTC homes are weakly related to the inspectorates’ ratings of person-centred care, but not to the ratings of other aspects of the quality of care.
We recommend to investigate approaches to intensify or innovate the involvement of care users’ experiences in regulatory practice to do them justice.
Increased emphasis is being placed on person-centredness as a quality requirement for long-term care (LTC), such as the care provided by nursing homes and other LTC homes. The importance of person-centred care, that is, care that fulfils the needs of the person in accordance with their preferences, has been widely discussed. The WHO states that LTC should be person-centred care that is consistent with peoples’ basic rights, fundamental freedoms and human dignity. Users of LTC must be invited and empowered to participate in decisions about their care, and their wishes should be protected and their dignity and autonomy promoted. Person-centred care is also an important issue for healthcare inspectorates when they assess the quality of care in LTC homes. Healthcare inspectorates are entrusted with the task of supervising healthcare by making sure that care providers comply with the relevant legal and field standards through site visits, reports of serious incidents and analysis of relevant information. As an acknowledgement of the importance of person-centred care, regulators have attempted to incorporate care users’ experiences into the inspection process by collecting information from care users, including people living in a nursing home setting. Braithwaite and Makkai showed that—in general—an inspection process is both reliable and practical, regardless of the severity of the needs of the users in care homes. The involvement of care users and family members in the inspection process may contribute to justice and empowerment among care users and their families and improve the quality and legitimacy of regulation. Although care users’ experiences are valued by regulators during assessments of the quality of LTC provided in care homes, research shows that it is not easy to collect data on care users’ experiences. For example, the Dutch Health and Youth Care Inspectorate introduced ‘Mystery Guests’ as an equivalent of the ‘mystery shopper’ in the consumer sector to assess the quality of LTC. However, a pilot study revealed that inspectors did not use the information gathered by the mystery guests, as the mystery guests evaluated the quality of care and reported their findings in a manner that did not align with the practices of the inspectorate. In a subsequent study, experts by experience were selected and trained to assess the quality of care provided in LTC homes. Experts by experience are lay people whose experiences lie between the world of care users and the inspectors of LTC homes, who either have personal experience with LTC homes or experience as an informal carer of users of an LTC. However, the results of the study showed that the added value of the experiences reported by these experts was limited, as the factual information produced during the inspectorate’s assessments was repeatedly valued as being more legitimate. In other words, as a cultural effect, the inspectors valued their professional knowledge more highly than the practical wisdom of the experts by experience. As a result, the experts by experience were unable to contribute their own experiences during the assessment by the inspectorate, except when the experts by experience established contact and spoke with care users. The Care Quality Commission (CQC), the regulator of health and social care in England, also included experts by experience (either service users or lay people) in their inspection teams. A review of policy documents and interviews with CQC staff and patient and public representatives showed that the CQC tried to include the users’ experiences within their inspections and ratings. However, the process by which the users’ experiences were incorporated was not transparent and the CQC stopped engaging with the experts by experience at that time. Nevertheless, regulators in several countries, including The Netherlands and the UK, continue to improve their approaches to include care user information and involvement in their inspection processes and assessments ; this raises the question of whether the quality of nursing home care as assessed by regulators reflects the quality of care perceived by the care users. Care users’ ratings collected from publicly available online review websites have previously been used to explore the usability and relationship of care users’ feedback with inspection outcomes. These websites were originally intended to help future care users and their representatives compare care providers and make informed decisions when selecting a specific care organisation. Research in the USA has shown that Yelp care users’ five-star ratings of nursing home care were significantly higher than the Nursing Home Compare inspection ratings. However, another study from the USA showed that the aggregate scores of users’ five-star ratings collected from four social media sites (Facebook, Yelp, Google Consumer Reviews and caring.com) correlated positively to Nursing Home Compare inspection ratings. Since care users’ experiences feature prominently on the agenda of regulators in many countries, this study aimed to explore whether care users’ and the inspectorate’s ratings of the quality of nursing home care in The Netherlands are related, and if so, to what extent. Due to the focus of the Dutch Inspectorate to involve care users and family members in regulatory practice, we hypothesise that both the inspectorate’s and care users’ ratings will show at least some overlap, particularly with respect to person-centred care.
Selection of LTC homes This study investigated the quality of care provided by a wide range of LTC homes in The Netherlands, here called LTC homes, for which public care users’ ratings of the quality of care were available and ratings of the Dutch Health and Youth Care Inspectorate (the inspectorate) (N=200; see below for more details on the selection process). In The Netherlands, there are different types of LTC homes, ranging from small-scale living arrangements, such as care farms or group living homes, to traditional large-scale nursing homes. The Care Needs Assessment Centre decides whether people are entitled to care in an LTC home as stipulated in the LTC-act (Wet Langdurige Zorg: WLZ). One of the requirements is that the person needs constant care or supervision. Patient and public involvement Patient or public were not directly involved in this study. However, we did use publicly available care users' ratings of the quality of care provided in LTC homes and investigated the correlations with the inspectorate’s ratings as described below. Care users’ ratings Care users’ ratings of the quality of care provided in LTC homes were extracted from a public Dutch online rating site managed by The Netherlands Patient Federation. Care users or their representatives—relatives or other close ones—can rate the quality of care on a voluntary basis by scoring six items on a scale from 1 to 10, where 10 is the best score (see for a description of the items). For the main analyses, we calculated the mean care users’ ratings for the six items as well as the total care users’ ratings for each LTC home. To ensure the relevance of the care users’ ratings, we only used data on care users’ ratings published within the 2 years prior to the rating of each LTC home by the inspectorate. The ratings are published anonymously on the ZorgkaartNederland website. ZorgkaartNederland does not collect any data on the characteristics of the people who provide the ratings. In total, care users’ ratings were available for 2152 LTC homes. Ratings of the inspectorate Inspectors visit LTC homes and use a framework to rate adherence to the standards for quality of nursing home care. The framework consists of eight items covering three themes: ‘attention to person-centred care’, ‘working towards sufficient and competent care staff’ and ‘focusing on quality and safety’ (see ). The items are rated by at least two inspectors on a four-point scale: ‘does not meet the standard’, ‘largely does not meet the standard’, ‘largely meets the standard’ or ‘meets the standard’. Inspectors base their ratings on at least three different sources per item, including conversations with care staff, care users or family members; observations of care; and findings from care user-files and other documents. For the analyses in this study, we calculated the aggregated rating per theme for each LTC home. We first dichotomised the scores for the items as ‘does not (or does not largely) meet the standard’ (0) and ‘(largely) meets the standard’ and then summed the scores per theme and for the three themes for each LTC home. For 301 LTC homes, a regular rating conducted by the inspectorate between January 2017 and March 2019 was available. To guarantee the independence of the observations, we only included the first LTC home rated for care organisations of which multiple LTC homes were visited. In addition, we did not include follow-up ratings by the inspectorate for LTC homes that did not meet the standard during an earlier assessment, or ratings based on assessments focused on a specific theme. In total, 228 of the 301 LTC homes with ratings of the inspectorate were eligible for this study. Thirty-seven (12.2%) of the 301 inspectorate’s ratings of LTC homes were excluded because some items were not scored; for example, due to a lack of time during the assessment or a lack of information required for formal assessment of an item. Of the remaining 264 LTC homes, 36 LTC homes belonged to care organisations for which another LTC home was already included in this study. To guarantee the independence of the observations, we only included the first LTC home assessed if multiple homes run by the same care provider were visited. Since no care user ratings were available for 28 of these 228 LTC homes, we used data of 200 LTC homes for which care user ratings were available in the 2 years prior to an assessment by the inspectorate. The number of available care users’ ratings per item ranged from 1 to 225 per LTC home over a period of 2 years (M=31; SD=33). This range of ratings can be explained by the different number of residents living in LTC homes. The included LTC homes had 6–350 residents (M=89; SD=57) and belonged to organisations with 1–40 LTC homes in total (M=6; SD=6). Analysis First, we conducted Mann-Whitney U tests to compare the ratings of the LTC homes that were assessed by the inspectorate and LTC homes that were not assessed in order to investigate the generalisability of our results to the group of LTC homes that was not assessed. Since the LTC homes that were not assessed lacked a rating date, we included all care users’ ratings from 2 years prior to the date of the first rating by the inspectorate included in this study up to the date of the last rating. For these analyses, we only included the first LTC home that received a rating of care users in each care organisation. Next, principal component analysis (PCA) was used to check the distribution of the scores for the items assessed under the three themes that the inspectorate aims to rate. In PCA, items that share the most common explained variance cluster together. Oblimin rotation was applied as we expected that the different themes would correlate with each other. To investigate our hypothesis, whether care users’ and the inspectorate’s ratings of quality of care of LTC homes show at least some overlap, we calculated the Spearman rank correlations between the overall mean care users’ ratings and the inspectorate’s scores for each theme. We included all LTC homes, also those with only a few ratings, to ensure the inclusion of small LTC homes with few residents. However, it allows extreme ratings that would otherwise have averaged out. Therefore, we reanalysed these correlations for a subgroup of the LTC homes, only including LTC homes with at least 30 care users’ ratings to investigate to what extend the results of the full sample are influenced by extreme values. To correct for multiple comparisons, we corrected the p values for each analysis using the false discovery rate (FDR) procedure to reduce the type I error. P values were considered significant if their corrected value did not exceed 0.05. We performed post-hoc analyses to obtain more insight into the items that explain the significant correlation, that is, the extent to which care users’ ratings on the six separate items correlated with the theme ‘attention to person-centred care’. As before, we calculated the Spearman’s rank correlations to investigate the extent to which care users’ ratings on the six separate items correlated with the individual themes. Conversely, we also examined the correlations between the scores for the separate items in the significantly correlated themes and the total mean care users’ ratings. All analyses were carried out using SPSS IBM V.24. The FDR correction was performed using R V.3.6.1.
This study investigated the quality of care provided by a wide range of LTC homes in The Netherlands, here called LTC homes, for which public care users’ ratings of the quality of care were available and ratings of the Dutch Health and Youth Care Inspectorate (the inspectorate) (N=200; see below for more details on the selection process). In The Netherlands, there are different types of LTC homes, ranging from small-scale living arrangements, such as care farms or group living homes, to traditional large-scale nursing homes. The Care Needs Assessment Centre decides whether people are entitled to care in an LTC home as stipulated in the LTC-act (Wet Langdurige Zorg: WLZ). One of the requirements is that the person needs constant care or supervision.
Patient or public were not directly involved in this study. However, we did use publicly available care users' ratings of the quality of care provided in LTC homes and investigated the correlations with the inspectorate’s ratings as described below.
Care users’ ratings of the quality of care provided in LTC homes were extracted from a public Dutch online rating site managed by The Netherlands Patient Federation. Care users or their representatives—relatives or other close ones—can rate the quality of care on a voluntary basis by scoring six items on a scale from 1 to 10, where 10 is the best score (see for a description of the items). For the main analyses, we calculated the mean care users’ ratings for the six items as well as the total care users’ ratings for each LTC home. To ensure the relevance of the care users’ ratings, we only used data on care users’ ratings published within the 2 years prior to the rating of each LTC home by the inspectorate. The ratings are published anonymously on the ZorgkaartNederland website. ZorgkaartNederland does not collect any data on the characteristics of the people who provide the ratings. In total, care users’ ratings were available for 2152 LTC homes.
Inspectors visit LTC homes and use a framework to rate adherence to the standards for quality of nursing home care. The framework consists of eight items covering three themes: ‘attention to person-centred care’, ‘working towards sufficient and competent care staff’ and ‘focusing on quality and safety’ (see ). The items are rated by at least two inspectors on a four-point scale: ‘does not meet the standard’, ‘largely does not meet the standard’, ‘largely meets the standard’ or ‘meets the standard’. Inspectors base their ratings on at least three different sources per item, including conversations with care staff, care users or family members; observations of care; and findings from care user-files and other documents. For the analyses in this study, we calculated the aggregated rating per theme for each LTC home. We first dichotomised the scores for the items as ‘does not (or does not largely) meet the standard’ (0) and ‘(largely) meets the standard’ and then summed the scores per theme and for the three themes for each LTC home. For 301 LTC homes, a regular rating conducted by the inspectorate between January 2017 and March 2019 was available. To guarantee the independence of the observations, we only included the first LTC home rated for care organisations of which multiple LTC homes were visited. In addition, we did not include follow-up ratings by the inspectorate for LTC homes that did not meet the standard during an earlier assessment, or ratings based on assessments focused on a specific theme. In total, 228 of the 301 LTC homes with ratings of the inspectorate were eligible for this study. Thirty-seven (12.2%) of the 301 inspectorate’s ratings of LTC homes were excluded because some items were not scored; for example, due to a lack of time during the assessment or a lack of information required for formal assessment of an item. Of the remaining 264 LTC homes, 36 LTC homes belonged to care organisations for which another LTC home was already included in this study. To guarantee the independence of the observations, we only included the first LTC home assessed if multiple homes run by the same care provider were visited. Since no care user ratings were available for 28 of these 228 LTC homes, we used data of 200 LTC homes for which care user ratings were available in the 2 years prior to an assessment by the inspectorate. The number of available care users’ ratings per item ranged from 1 to 225 per LTC home over a period of 2 years (M=31; SD=33). This range of ratings can be explained by the different number of residents living in LTC homes. The included LTC homes had 6–350 residents (M=89; SD=57) and belonged to organisations with 1–40 LTC homes in total (M=6; SD=6).
First, we conducted Mann-Whitney U tests to compare the ratings of the LTC homes that were assessed by the inspectorate and LTC homes that were not assessed in order to investigate the generalisability of our results to the group of LTC homes that was not assessed. Since the LTC homes that were not assessed lacked a rating date, we included all care users’ ratings from 2 years prior to the date of the first rating by the inspectorate included in this study up to the date of the last rating. For these analyses, we only included the first LTC home that received a rating of care users in each care organisation. Next, principal component analysis (PCA) was used to check the distribution of the scores for the items assessed under the three themes that the inspectorate aims to rate. In PCA, items that share the most common explained variance cluster together. Oblimin rotation was applied as we expected that the different themes would correlate with each other. To investigate our hypothesis, whether care users’ and the inspectorate’s ratings of quality of care of LTC homes show at least some overlap, we calculated the Spearman rank correlations between the overall mean care users’ ratings and the inspectorate’s scores for each theme. We included all LTC homes, also those with only a few ratings, to ensure the inclusion of small LTC homes with few residents. However, it allows extreme ratings that would otherwise have averaged out. Therefore, we reanalysed these correlations for a subgroup of the LTC homes, only including LTC homes with at least 30 care users’ ratings to investigate to what extend the results of the full sample are influenced by extreme values. To correct for multiple comparisons, we corrected the p values for each analysis using the false discovery rate (FDR) procedure to reduce the type I error. P values were considered significant if their corrected value did not exceed 0.05. We performed post-hoc analyses to obtain more insight into the items that explain the significant correlation, that is, the extent to which care users’ ratings on the six separate items correlated with the theme ‘attention to person-centred care’. As before, we calculated the Spearman’s rank correlations to investigate the extent to which care users’ ratings on the six separate items correlated with the individual themes. Conversely, we also examined the correlations between the scores for the separate items in the significantly correlated themes and the total mean care users’ ratings. All analyses were carried out using SPSS IBM V.24. The FDR correction was performed using R V.3.6.1.
Comparison of care users’ ratings for included and excluded LTC homes The mean total score for the care users’ ratings of the quality of care was 7.9 (N=200, SD=1.1; ). Mann-Whitney U tests revealed a weak, but significant, difference between the care users’ ratings for the LTC homes assessed by the inspectorate (M=7.9, N=200) and the LTC homes not rated by the inspectorate (M=8.3, N=219, U=16 310.5, p<0.001). PCA of the ratings of the inspectorate With respect to the ratings by the inspectorate, we first conducted PCA to investigate the consistency of the inspectorate’s framework across the three themes and the eight underlying items. The PCA showed that the three-dimensional solution was consistent with the themes as conceptualised in the original framework, except for item 6 ‘The care organisation ensures that sufficient expert caregivers are available, geared to the older people’s present and current care needs’ (see ). This item had the highest loading on the theme ‘focusing on quality and safety’, but was originally conceptualised as part of the theme ‘working towards sufficient and competent care staff’. The three principal components explained 40.7%, 16.0% and 11.3% of the variance, respectively. Based on the face validity of these results, we used this new structure instead of the original; thus, item 6 was included in the theme ‘focusing on quality and safety’ (see ). The aggregated scores on the three themes of the framework used by the inspectorate indicated that the mean aggregated ratings for ‘attention to person-centred care’ are higher than the mean aggregated ratings for the two other themes ( ). Correlations between care users’ and the inspectorate’s ratings We observed a weak, significant positive correlation between the mean care users’ ratings and the aggregated ratings by the inspectorate for the theme ‘attention to person-centred care’ (r=0.26, N=200, p adj <0.01). Thus, higher ratings of care users were related with higher ratings of the inspectorate of the theme ‘Attention to person-centred care’. No other correlations were significant (p adj >0.05; ). Similar results were obtained when only LTC homes with more than 30 ratings were included (N=75; 32.8%). However, after FDR correction, the correlation between care users’ ratings and the inspectorate’s ratings of the theme ‘attention to person-centred care’ was no longer significant. Since there was a significant correlation between the total care users’ ratings and the scores on the inspectorate’s theme ‘attention to person-centred care’, we next carried out a post-hoc analysis (see ). We found weak, significant correlations between the care users’ ratings for each of the six individual items and the total scores on inspectorate’s theme ‘attention to person-centred care’ (r=0.21–0.30, p adj <0.01). We also analysed the correlation between the mean total care users’ ratings and the scores for the three individual items of the inspectorate’s theme ‘attention to person-centred care’. Weak, positive correlations were observed between the care user’s ratings and each individual item (r=0.16–0.24, p adj <0.5).
The mean total score for the care users’ ratings of the quality of care was 7.9 (N=200, SD=1.1; ). Mann-Whitney U tests revealed a weak, but significant, difference between the care users’ ratings for the LTC homes assessed by the inspectorate (M=7.9, N=200) and the LTC homes not rated by the inspectorate (M=8.3, N=219, U=16 310.5, p<0.001).
With respect to the ratings by the inspectorate, we first conducted PCA to investigate the consistency of the inspectorate’s framework across the three themes and the eight underlying items. The PCA showed that the three-dimensional solution was consistent with the themes as conceptualised in the original framework, except for item 6 ‘The care organisation ensures that sufficient expert caregivers are available, geared to the older people’s present and current care needs’ (see ). This item had the highest loading on the theme ‘focusing on quality and safety’, but was originally conceptualised as part of the theme ‘working towards sufficient and competent care staff’. The three principal components explained 40.7%, 16.0% and 11.3% of the variance, respectively. Based on the face validity of these results, we used this new structure instead of the original; thus, item 6 was included in the theme ‘focusing on quality and safety’ (see ). The aggregated scores on the three themes of the framework used by the inspectorate indicated that the mean aggregated ratings for ‘attention to person-centred care’ are higher than the mean aggregated ratings for the two other themes ( ).
We observed a weak, significant positive correlation between the mean care users’ ratings and the aggregated ratings by the inspectorate for the theme ‘attention to person-centred care’ (r=0.26, N=200, p adj <0.01). Thus, higher ratings of care users were related with higher ratings of the inspectorate of the theme ‘Attention to person-centred care’. No other correlations were significant (p adj >0.05; ). Similar results were obtained when only LTC homes with more than 30 ratings were included (N=75; 32.8%). However, after FDR correction, the correlation between care users’ ratings and the inspectorate’s ratings of the theme ‘attention to person-centred care’ was no longer significant. Since there was a significant correlation between the total care users’ ratings and the scores on the inspectorate’s theme ‘attention to person-centred care’, we next carried out a post-hoc analysis (see ). We found weak, significant correlations between the care users’ ratings for each of the six individual items and the total scores on inspectorate’s theme ‘attention to person-centred care’ (r=0.21–0.30, p adj <0.01). We also analysed the correlation between the mean total care users’ ratings and the scores for the three individual items of the inspectorate’s theme ‘attention to person-centred care’. Weak, positive correlations were observed between the care user’s ratings and each individual item (r=0.16–0.24, p adj <0.5).
Key findings This study explored the relationships between care users’ ratings of the quality of care provided in LTC homes in The Netherlands and ratings of the Dutch Health and Youth Care Inspectorate. Ratings of both care users and the inspectorate were available for 200 LTC homes. We observed a weak, positive correlation between care users’ and the inspectorate’s ratings of the quality of care, but only for the inspectorate’s ratings on the theme ‘attention to person-centred care’. This finding is consistent with our hypothesis that there would be some overlap, particularly with respect to person-centred care. We did not find any relationship between care users’ and the inspectorate’s ratings for the two other themes assessed during inspection visits: ‘working towards sufficient and competent care staff’ and ‘focusing on quality and safety’. We observed a non-significant relationship of a similar size after excluding LTC homes with less than 30 care users’ ratings. This finding shows that although the power was too low to find a significant correlation of this size for this subsample, a similar relationship seems present. The correlation we found for the full sample was thus not solely caused by extreme values for LTC homes with only a few ratings. Post-hoc analyses revealed weak to low-to-medium correlations between the individual items of care in the user ratings and the inspectorate’s scores for the theme ‘attention to person-centred care’. Care users’ rating of the items ‘Alignment of care to own life’ showed the highest correlation with the inspectorate’s scores on the theme ‘attention to person-centred care’. Conversely, weak but significant correlations were observed between the total care users’ ratings and the three individual items of the theme ‘attention to person-centred care’. Reflection on findings Our findings are in line with earlier studies conducted in the USA and Great Britain, which reported discrepancies and weak to low-to-medium correlations between care users’ ratings of the care provided by nursing homes on social media or online review sites and regulators’ ratings The weak correlations might be partly explained by the duration of the measurements, since the care users’ ratings covered a period of 2 years prior to the ratings of the inspectorate. We cannot rule out the impact of this time period on the quality of care. However, analysis of a shorter time period, for example, 1 year was not considered suitable, as the number of available care users’ ratings would have been substantially lower. Nevertheless, it seems unlikely that the duration of measurement alone is responsible for the weak relationship between the care users’ and inspectorate’s ratings. There is no evidence that gaming, whereby ratings are artificially manipulated, may play a significant role in the outcomes of this study. The patient rating website for the quality of care in The Netherlands is an independent website run by a not-for profit organisation, the Patient Federation, that verifies the quality of the ratings that are posted. Only ratings that comply with the code of conduct are posted online. About 90% of the submitted ratings are approved immediately. To prevent posting of improper (ie, junk or spam) reviews, the editorial office examines all reviews before publishing. In addition, the email addresses and IP addresses of the reviewers are also checked to prevent improper use and duplicates. This increases the quality of the ratings that have been posted. The small size of the correlation found in this study between care users’ ratings of the quality of care in LTC homes and the inspectorates’ ratings of person-centred care at least raises the question whether the inspectorate takes sufficient account of the care users’ experiences during their assessment of the quality of LTC. This finding suggests there is room for the inspectorate to intensify or innovate approaches to address care users’ experiences in their assessments of the quality of care to do them justice. Strengths and limitations of the study One methodological strength of this study is the sample size of 200 LTC homes. All LTC homes were from different providers, and the sample represents about 40% of all care providers with one or more LTC homes in The Netherlands known to the inspectorate. However, the LTC homes that were assessed by the inspectorate were not selected at random; therefore, the sample is not necessarily representative. The selection process for inspection also includes LTC providers where risks were foreseen, which may have contributed to the slightly lower care users’ ratings for the LTC homes that were assessed by the inspectorate compared with the LTC homes that were not assessed. The fact that the ratings of LTC homes were obtained from the independent, non-commercial ZorgkaartNederland website is also a strength of this study. The threshold to post a review is low; in other words, anyone can post a review. Although this could potentially lead to misuse of the platform, the ZorgkaartNederland editorial office verifies the legitimacy of each review before publishing the reviews online, which increases the validity of the ratings. However, it is a limitation that the care users’ ratings are published anonymously on the ZorgkaartNederland website. ZorgkaartNederland does not ask for or publish any characteristics from people who place a rating. Therefore, analyses of the relationships between the ratings of care users (or their representatives) and any of their personal characteristics were not possible. Another strength of this study is that we could relate the inspectorate’s ratings for different aspects (themes) of the quality of care, to care users’ ratings and examine the differences in the strengths of these relationships. In contrast, earlier studies only described general ratings of the quality of care by inspectorates. A last point to be addressed is that all three authors of this study are paid staff of the Dutch Health and Youth Care Inspectorate. However, in this respect, it is important to mention that all the data used in this study are publicly available, the methods used to analyse the data are transparent and the interpretation of the findings highlight the need for reflection on the own assessment process of the Dutch Health and Youth Care Inspectorate. Recommendations for further research Although the Dutch Inspectorate values and attempts to incorporate the experiences of care users and their family members in their assessments, this study shows that their ratings of LTC homes are only weakly related to the experiences of care users based on ratings from the Dutch patient rating site ZorgkaartNederland. Integration of the experiences of care users in the inspection process is challenging, as previous research has shown. Although a variety of approaches are currently used to promote patient and family involvement in healthcare regulation, the extent to which these approaches are incorporated into regulatory procedures is unclear. It might be fruitful to intensify some of these approaches; for example, by systematically collecting data on care users’ experiences during inspection visits and investigate how they can be embodied in regulatory practice. In addition, new regulatory approaches may be needed to do justice to care users’ experiences. For example, regulators may need to move from an outcome-based approach to a process-based and reflexive regulatory approach, in which regulators and service providers organise themselves around care users. This could allow regulators to assess the presence and quality of the processes that allow organisations to provide person-centred care. Further research is needed to develop and substantiate such regulatory innovations with evidence.
This study explored the relationships between care users’ ratings of the quality of care provided in LTC homes in The Netherlands and ratings of the Dutch Health and Youth Care Inspectorate. Ratings of both care users and the inspectorate were available for 200 LTC homes. We observed a weak, positive correlation between care users’ and the inspectorate’s ratings of the quality of care, but only for the inspectorate’s ratings on the theme ‘attention to person-centred care’. This finding is consistent with our hypothesis that there would be some overlap, particularly with respect to person-centred care. We did not find any relationship between care users’ and the inspectorate’s ratings for the two other themes assessed during inspection visits: ‘working towards sufficient and competent care staff’ and ‘focusing on quality and safety’. We observed a non-significant relationship of a similar size after excluding LTC homes with less than 30 care users’ ratings. This finding shows that although the power was too low to find a significant correlation of this size for this subsample, a similar relationship seems present. The correlation we found for the full sample was thus not solely caused by extreme values for LTC homes with only a few ratings. Post-hoc analyses revealed weak to low-to-medium correlations between the individual items of care in the user ratings and the inspectorate’s scores for the theme ‘attention to person-centred care’. Care users’ rating of the items ‘Alignment of care to own life’ showed the highest correlation with the inspectorate’s scores on the theme ‘attention to person-centred care’. Conversely, weak but significant correlations were observed between the total care users’ ratings and the three individual items of the theme ‘attention to person-centred care’.
Our findings are in line with earlier studies conducted in the USA and Great Britain, which reported discrepancies and weak to low-to-medium correlations between care users’ ratings of the care provided by nursing homes on social media or online review sites and regulators’ ratings The weak correlations might be partly explained by the duration of the measurements, since the care users’ ratings covered a period of 2 years prior to the ratings of the inspectorate. We cannot rule out the impact of this time period on the quality of care. However, analysis of a shorter time period, for example, 1 year was not considered suitable, as the number of available care users’ ratings would have been substantially lower. Nevertheless, it seems unlikely that the duration of measurement alone is responsible for the weak relationship between the care users’ and inspectorate’s ratings. There is no evidence that gaming, whereby ratings are artificially manipulated, may play a significant role in the outcomes of this study. The patient rating website for the quality of care in The Netherlands is an independent website run by a not-for profit organisation, the Patient Federation, that verifies the quality of the ratings that are posted. Only ratings that comply with the code of conduct are posted online. About 90% of the submitted ratings are approved immediately. To prevent posting of improper (ie, junk or spam) reviews, the editorial office examines all reviews before publishing. In addition, the email addresses and IP addresses of the reviewers are also checked to prevent improper use and duplicates. This increases the quality of the ratings that have been posted. The small size of the correlation found in this study between care users’ ratings of the quality of care in LTC homes and the inspectorates’ ratings of person-centred care at least raises the question whether the inspectorate takes sufficient account of the care users’ experiences during their assessment of the quality of LTC. This finding suggests there is room for the inspectorate to intensify or innovate approaches to address care users’ experiences in their assessments of the quality of care to do them justice.
One methodological strength of this study is the sample size of 200 LTC homes. All LTC homes were from different providers, and the sample represents about 40% of all care providers with one or more LTC homes in The Netherlands known to the inspectorate. However, the LTC homes that were assessed by the inspectorate were not selected at random; therefore, the sample is not necessarily representative. The selection process for inspection also includes LTC providers where risks were foreseen, which may have contributed to the slightly lower care users’ ratings for the LTC homes that were assessed by the inspectorate compared with the LTC homes that were not assessed. The fact that the ratings of LTC homes were obtained from the independent, non-commercial ZorgkaartNederland website is also a strength of this study. The threshold to post a review is low; in other words, anyone can post a review. Although this could potentially lead to misuse of the platform, the ZorgkaartNederland editorial office verifies the legitimacy of each review before publishing the reviews online, which increases the validity of the ratings. However, it is a limitation that the care users’ ratings are published anonymously on the ZorgkaartNederland website. ZorgkaartNederland does not ask for or publish any characteristics from people who place a rating. Therefore, analyses of the relationships between the ratings of care users (or their representatives) and any of their personal characteristics were not possible. Another strength of this study is that we could relate the inspectorate’s ratings for different aspects (themes) of the quality of care, to care users’ ratings and examine the differences in the strengths of these relationships. In contrast, earlier studies only described general ratings of the quality of care by inspectorates. A last point to be addressed is that all three authors of this study are paid staff of the Dutch Health and Youth Care Inspectorate. However, in this respect, it is important to mention that all the data used in this study are publicly available, the methods used to analyse the data are transparent and the interpretation of the findings highlight the need for reflection on the own assessment process of the Dutch Health and Youth Care Inspectorate.
Although the Dutch Inspectorate values and attempts to incorporate the experiences of care users and their family members in their assessments, this study shows that their ratings of LTC homes are only weakly related to the experiences of care users based on ratings from the Dutch patient rating site ZorgkaartNederland. Integration of the experiences of care users in the inspection process is challenging, as previous research has shown. Although a variety of approaches are currently used to promote patient and family involvement in healthcare regulation, the extent to which these approaches are incorporated into regulatory procedures is unclear. It might be fruitful to intensify some of these approaches; for example, by systematically collecting data on care users’ experiences during inspection visits and investigate how they can be embodied in regulatory practice. In addition, new regulatory approaches may be needed to do justice to care users’ experiences. For example, regulators may need to move from an outcome-based approach to a process-based and reflexive regulatory approach, in which regulators and service providers organise themselves around care users. This could allow regulators to assess the presence and quality of the processes that allow organisations to provide person-centred care. Further research is needed to develop and substantiate such regulatory innovations with evidence.
This study observed a weak correlation between care users’ ratings of the quality of care they receive in LTC homes in The Netherlands and the inspectorate’s ratings of the provision of person-centred care in these homes. No correlations were found between care users’ ratings and inspectorate’s ratings of other aspects of the quality of care in LTC homes. Further research is needed to investigate how healthcare regulators can do justice to care users’ experiences during their assessment of the quality of LTC, particularly person-centred care.
|
Novel patient-centred outcome in cancer care, days at home: a scoping review protocol | 02fcc239-1655-4156-ae9a-a455544c1912 | 10030791 | Patient-Centered Care[mh] | Patient-centred care is important and desired by patients and providers. It is a holistic approach to healthcare that seeks to treat patients as unique individuals by incorporating patient and family values, perspectives and expectations into a shared decision-making model. Patient-centred care was initially introduced over 50 years ago as a contentious shift in philosophy of care. Today, there is clear consensus across major medical organisations and institutions on its utility and value in reaching the goal of high-quality care. Accordingly, the WHO has released a global strategy towards improved healthcare by focusing on ways to implement patient-centred care. The value of patient-centred care as a philosophy is no longer disputed, but challenges remain in ensuring care is delivered with this framework in mind. In oncology, the value of providing patient-centred care resides in tailoring care to individual goals and needs that are known to vary according to demographics, tumour biology, values and preferences. In addition, many patients with cancer prefer being active participants in shared decision making. To improve care provided to patients with cancer by leveraging principles of patient-centred care, clinicians need accurate metrics that are meaningful to both patients and providers to support counselling, decision-making and preparedness for treatments. The patient-centred outcome, days at home (DAH), is meaningful to patients and accurately measured, and has construct validity for many provider-important and patient-important outcomes. DAH assesses the time a patient spends alive and out of hospitals and healthcare institutions and is measured in units of days, a metric that is simple and understood by patients and providers. DAH relates to the construct of time toxicity, a concept emerging in oncology from recognition of the significant time spent attending treatments, appointments or managing toxicities from therapy. The outcome DAH was first used in 2005 in a randomised control trial in the field of cardiology, being introduced only recently into the oncology literature. Since its first use, DAH has been shown to have construct validity with many traditional outcome measures such as morbidity and mortality, along with other patient-centred outcomes. Given that measurement of DAH is feasible using various data sources, from trials to administrative health data, it has potential to be a meaningful research, policy and clinical tool in cancer care. Several terms have been applied to the construct captured by DAH in the literature including but not limited to DAH, time at home, days spent at home, days alive and out of hospital, days alive and at home, home time and healthy DAH. Accordingly, definitions have been inconsistent. In addition, some researchers opt to measure its inverse, which has been termed institution days or time toxicity. The heterogeneity in both terms used and definitions point to the need for a better understanding of its use to date to inform future applications. A preliminary search for existing reviews on the construct captured by DAH was performed on 25 October 2022 on MEDLINE with no prior reviews retrieved. Therefore, we aim to perform a scoping review to consolidate the current information on the patient-centred outcome DAH in cancer care, review outcome definitions and provide an overview of how it has been used in the oncology literature to date. This work is conducted with a view to guide future use of the outcome DAH in patient-centred cancer research, clinical decision making and policy, such as quality monitoring and resource allocation.
This scoping review protocol has been designed with joint guidance from the JBI Manual for Evidence Synthesis and the expanded framework from Arksey and O’Malley. Reporting of the scoping review will adhere to the Preferred Reporting Items for Systematic reviews and Meta-Analyses Extension for Scoping Reviews. This review has been registered with Open Science Framework ( https://osf.io/xzpj7 ). Objectives The scoping review will answer the following research questions: How has the construct ‘DAH’ been termed and defined as an outcome in cancer care? How has the construct DAH been used as an outcome measure in cancer care? In what context has DAH been validated in cancer care? Eligibility criteria See for inclusion and exclusion criteria. The relatively recent introduction of this construct into the scientific literature necessitates broad inclusion criteria to have wide capture of what has been reported on to date. Population The population of interest includes studies reporting on adults (≥18 years) with a current or past cancer diagnosis. All cancer types will be included. Concept The outcome DAH is inconsistently termed and defined in the literature. For the review, DAH has been defined as a composite outcome, recording the time a patient spends alive and out of hospitals and healthcare institutions. Healthcare institutions include oncology centres, emergency departments, rehabilitation centres, long-term care homes and other healthcare facilities. The use of the term DAH in this review refers to all variables measuring the same construct, acknowledging the variability in terms and definitions used. We will also include studies measuring the inverse of DAH, such as institution days, which speaks to the same underlying construct. DAH may be operationalised as a continuous variable or as a proportion of a predefined number of days. The definition used for this review seeks to encompass all applications of a construct that uses hospital and healthcare institution days as a patient-centred metric to reflect the overall treatment burden on patients. Context To capture literature studying the outcome DAH in cancer care; studies evaluating any interventions in the diagnosis, management and follow-up of cancer care will be included. This will include all forms of systemic therapy, radiation therapy and surgery for the purposes of cancer treatment. Study details A broad list of study designs will be included. We will include randomised and non-randomised trials, prospective and retrospective cohort studies, case–control studies, cross-sectional studies, quasi-experimental studies and case series of 10 or more participants. There will be no publishing year restrictions. Grey literature will be excluded. All geographical regions and languages will be included; however, the literature search will only be performed in English. Search strategy and information sources The search strategy was developed for MEDLINE through consultation with a librarian at the University of Toronto. The search was adopted with aid from a librarian for use in two additional databases: Embase and Scopus. Additional text words may be added to the search in an iterative manner as reviewers explore the evidence base. A list of text words were chosen based on a preliminary literature review. Terms analogous to DAH were included to ensure broad capture. See for the full search strategy for MEDLINE. Search strategies for additional databases are provided in the . The search is proposed to be completed between 1 January 2023 and 1 March 2023. 10.1136/bmjopen-2022-071201.supp1 Supplementary data Study selection Study selection will follow guidelines as set out by the JBI Manual for Evidence Synthesis and the expanded Arksey and O’Malley framework. An initial pilot testing phase of eligibility criteria will be completed on a random sample of 25 titles and abstracts by two independent reviewers. The team of reviewers will review selection results, discuss discrepancies and modify eligibility criteria as necessary. Management of search results will be completed via Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia). Once inter-rater reliability of 75% or greater is achieved, study selection will begin. The first stage of study selection will involve title and abstract examination based on inclusion and exclusion criteria by two independent reviewers. The second stage will involve full-text review by two independent reviewers. Disagreements during both stages will be addressed through consensus or by consultation with the research team. Inclusion and exclusion criteria will be reviewed and may be modified following the pilot testing phase and iteratively throughout the search during research team meetings. See for study selection flowchart. Data extraction Data extraction tables have been designed by the research team with guidance from the JBI Manual for Evidence Synthesis and Arksey and O’Malley with research objectives in mind. See for preliminary extraction tables. Data extraction will follow an iterative process as outlined by expert guidelines, with updates to tables as deemed necessary by the research team. As a pilot step, two independent researchers will extract data from the first 10 studies into preliminary tables, as recommended by expert guidance. Results from pilot data extraction will be reviewed by the research team, and changes to the tables will be made as necessary, guided by research objectives. Data analysis In keeping with scoping review objectives and methodology, data analysis will include a descriptive numerical analysis followed by a thematic analysis. A descriptive numerical summary using tables and charts will describe proportions of study characteristics as guided by study objectives and data extraction. Key study characteristics will include study design, aims and validation. The nature and distribution of studies assessing the outcome DAH will provide insight into its scope and use in the cancer literature to date. Potential implications on research, policy or clinical use will be discussed. Patient and public involvement Patient and service user engagement is known to enhance the relevance, validity, quality and applicability of research. Patients and service users are essential partners in research, not only providing unique insight into lived experience of illness but also helping to determine research plans and outputs. Following the patient and service user engagement framework, we have engaged patients, service users, healthcare professionals (HCPs) and health decision makers to obtain additional sources of information, perspectives and applicability to the study. Two patient partners with lived experience of cancer (EK and JD) are members of the research team and have been involved from inception and will participate in all parts of the study to ensure clinical relevance. Consultations with stakeholders will also be used to share preliminary findings, validate and identify any gaps in our findings, and to inform future research efforts. Ethics and dissemination This scoping review protocol outlines a method to systematically search and map the literature on DAH for cancer care. Since this review will only include published data, ethics approval will not be sought. This scoping review will constitute the first stage in developing clinical tools to integrate DAH information into cancer care. Terminology, definitions and measures of DAH will inform the building of predictive tools and decision aids for personalised cancer care delivery. This is necessary to create tools that go beyond typical prognostication and provide patients with information on outcomes that are most relevant to their experience to support decision making and preparedness for treatment. Moreover, the information generated by this review can also be used by health systems, patient organisations, researchers and HCPs to plan cancer care delivery, clinical trial design and conduct, and assessment of health services.
The scoping review will answer the following research questions: How has the construct ‘DAH’ been termed and defined as an outcome in cancer care? How has the construct DAH been used as an outcome measure in cancer care? In what context has DAH been validated in cancer care?
See for inclusion and exclusion criteria. The relatively recent introduction of this construct into the scientific literature necessitates broad inclusion criteria to have wide capture of what has been reported on to date. Population The population of interest includes studies reporting on adults (≥18 years) with a current or past cancer diagnosis. All cancer types will be included. Concept The outcome DAH is inconsistently termed and defined in the literature. For the review, DAH has been defined as a composite outcome, recording the time a patient spends alive and out of hospitals and healthcare institutions. Healthcare institutions include oncology centres, emergency departments, rehabilitation centres, long-term care homes and other healthcare facilities. The use of the term DAH in this review refers to all variables measuring the same construct, acknowledging the variability in terms and definitions used. We will also include studies measuring the inverse of DAH, such as institution days, which speaks to the same underlying construct. DAH may be operationalised as a continuous variable or as a proportion of a predefined number of days. The definition used for this review seeks to encompass all applications of a construct that uses hospital and healthcare institution days as a patient-centred metric to reflect the overall treatment burden on patients. Context To capture literature studying the outcome DAH in cancer care; studies evaluating any interventions in the diagnosis, management and follow-up of cancer care will be included. This will include all forms of systemic therapy, radiation therapy and surgery for the purposes of cancer treatment. Study details A broad list of study designs will be included. We will include randomised and non-randomised trials, prospective and retrospective cohort studies, case–control studies, cross-sectional studies, quasi-experimental studies and case series of 10 or more participants. There will be no publishing year restrictions. Grey literature will be excluded. All geographical regions and languages will be included; however, the literature search will only be performed in English.
The population of interest includes studies reporting on adults (≥18 years) with a current or past cancer diagnosis. All cancer types will be included.
The outcome DAH is inconsistently termed and defined in the literature. For the review, DAH has been defined as a composite outcome, recording the time a patient spends alive and out of hospitals and healthcare institutions. Healthcare institutions include oncology centres, emergency departments, rehabilitation centres, long-term care homes and other healthcare facilities. The use of the term DAH in this review refers to all variables measuring the same construct, acknowledging the variability in terms and definitions used. We will also include studies measuring the inverse of DAH, such as institution days, which speaks to the same underlying construct. DAH may be operationalised as a continuous variable or as a proportion of a predefined number of days. The definition used for this review seeks to encompass all applications of a construct that uses hospital and healthcare institution days as a patient-centred metric to reflect the overall treatment burden on patients.
To capture literature studying the outcome DAH in cancer care; studies evaluating any interventions in the diagnosis, management and follow-up of cancer care will be included. This will include all forms of systemic therapy, radiation therapy and surgery for the purposes of cancer treatment.
A broad list of study designs will be included. We will include randomised and non-randomised trials, prospective and retrospective cohort studies, case–control studies, cross-sectional studies, quasi-experimental studies and case series of 10 or more participants. There will be no publishing year restrictions. Grey literature will be excluded. All geographical regions and languages will be included; however, the literature search will only be performed in English.
The search strategy was developed for MEDLINE through consultation with a librarian at the University of Toronto. The search was adopted with aid from a librarian for use in two additional databases: Embase and Scopus. Additional text words may be added to the search in an iterative manner as reviewers explore the evidence base. A list of text words were chosen based on a preliminary literature review. Terms analogous to DAH were included to ensure broad capture. See for the full search strategy for MEDLINE. Search strategies for additional databases are provided in the . The search is proposed to be completed between 1 January 2023 and 1 March 2023. 10.1136/bmjopen-2022-071201.supp1 Supplementary data
Study selection will follow guidelines as set out by the JBI Manual for Evidence Synthesis and the expanded Arksey and O’Malley framework. An initial pilot testing phase of eligibility criteria will be completed on a random sample of 25 titles and abstracts by two independent reviewers. The team of reviewers will review selection results, discuss discrepancies and modify eligibility criteria as necessary. Management of search results will be completed via Covidence systematic review software (Veritas Health Innovation, Melbourne, Australia). Once inter-rater reliability of 75% or greater is achieved, study selection will begin. The first stage of study selection will involve title and abstract examination based on inclusion and exclusion criteria by two independent reviewers. The second stage will involve full-text review by two independent reviewers. Disagreements during both stages will be addressed through consensus or by consultation with the research team. Inclusion and exclusion criteria will be reviewed and may be modified following the pilot testing phase and iteratively throughout the search during research team meetings. See for study selection flowchart.
Data extraction tables have been designed by the research team with guidance from the JBI Manual for Evidence Synthesis and Arksey and O’Malley with research objectives in mind. See for preliminary extraction tables. Data extraction will follow an iterative process as outlined by expert guidelines, with updates to tables as deemed necessary by the research team. As a pilot step, two independent researchers will extract data from the first 10 studies into preliminary tables, as recommended by expert guidance. Results from pilot data extraction will be reviewed by the research team, and changes to the tables will be made as necessary, guided by research objectives.
In keeping with scoping review objectives and methodology, data analysis will include a descriptive numerical analysis followed by a thematic analysis. A descriptive numerical summary using tables and charts will describe proportions of study characteristics as guided by study objectives and data extraction. Key study characteristics will include study design, aims and validation. The nature and distribution of studies assessing the outcome DAH will provide insight into its scope and use in the cancer literature to date. Potential implications on research, policy or clinical use will be discussed.
Patient and service user engagement is known to enhance the relevance, validity, quality and applicability of research. Patients and service users are essential partners in research, not only providing unique insight into lived experience of illness but also helping to determine research plans and outputs. Following the patient and service user engagement framework, we have engaged patients, service users, healthcare professionals (HCPs) and health decision makers to obtain additional sources of information, perspectives and applicability to the study. Two patient partners with lived experience of cancer (EK and JD) are members of the research team and have been involved from inception and will participate in all parts of the study to ensure clinical relevance. Consultations with stakeholders will also be used to share preliminary findings, validate and identify any gaps in our findings, and to inform future research efforts.
This scoping review protocol outlines a method to systematically search and map the literature on DAH for cancer care. Since this review will only include published data, ethics approval will not be sought. This scoping review will constitute the first stage in developing clinical tools to integrate DAH information into cancer care. Terminology, definitions and measures of DAH will inform the building of predictive tools and decision aids for personalised cancer care delivery. This is necessary to create tools that go beyond typical prognostication and provide patients with information on outcomes that are most relevant to their experience to support decision making and preparedness for treatment. Moreover, the information generated by this review can also be used by health systems, patient organisations, researchers and HCPs to plan cancer care delivery, clinical trial design and conduct, and assessment of health services.
Reviewer comments Author's
manuscript
|
Diversity and bioactive potential of | e019c2bb-7678-476a-bdd2-97e31941d99d | 10031196 | Microbiology[mh] | Natural products play a key role in drug discovery, especially when it comes to the treatment of infectious diseases and cancers (Newman and Cragg ). However, the rapid spread of bacterial infections has been observed in recent decades, mainly due to the rapid emergence and spread of multidrug-resistant (MDR) pathogens. One of the main reasons of this is the spread of MDR pathogens has demonstrated the lack of effective antibiotics to counteract and control them (Murray et al. ). As a consequence, there is now a general consensus that the discovery of new antibiotics would represent the best solution in the fight against antibiotic resistance among microorganisms. The Actinomycetia are a class of bacteria known to be widely distributed in environments characterised by a high G + C content in the genomes. However, the main feature of the Actinomycetia that has attracted the attention of researchers is their ability to produce biologically active compounds, principally antibiotics. In fact, they produce more than 70% of all known antibiotics of natural origin, with the Streptomyces genus producing approximately 55% of all antibiotics (Demain and Adrio ; Hutchings et al. ). Unfortunately, following the ‘golden age’ of antibiotic discovery, which lasted from the 1940s to the 1960s, when the most common compounds were discovered and the producer strains were isolated, the discovery and clinical introduction of new antibiotics slowed considerably. The difficulty in introducing antibiotics can be explained by the significant rediscovery of previously known antibiotics, predominantly among the streptomycetes (Katz and Baltz ). One strategy with which to discover new bioactive compounds, including antibiotics, involves exploring new and understudied environments. Today, there are increasing reports of the discovery of new antibiotic producers in previously understudied environments. In particular, actinomycete strains from marine habitats (Jagannathan et al. ), deserts (Rateb et al. ), caves (Farda et al. ), Antarctica (Tistechok et al. ) and other difficult to access or unexplored environments are being very actively studied. Moreover, the isolation of (rare) non-streptomycete genera from these environments increases the likelihood of discovering new biologically active compounds, as they are known to produce a number of antibiotics (Subramani and Sipkema ). Such studies confirm that much of the globe has been understudied in terms of antibiotic producer screening. Thus, continued research in under- and unexplored areas may lead to the discovery of important new antibiotics. The diversity of the microbiome of the Crimean Peninsula, which is characterised by great variability with regard to its microclimatic conditions, has not been practically studied. Our previous studies have demonstrate successfully investigated certain streptomycete strains isolated from the rhizosphere of this peninsula in terms of their potential as producers of biologically active compounds (Tistechok et al. , ), including previously unknown ones (Raju et al. , , , ). However, the diversity and biosynthetic potential of actinomycetes as producers of bioactive compounds from the rhizosphere soil of Juniperus excelsa M. Bieb. (also known as Greek juniper) in this region have not been described yet. Thus, in this study, we investigate the diversity of juniper rhizosphere actinomycetes and evaluate their potential for producing antimicrobial compounds.
Actinomycete strains and growth conditions The 372 actinomycete-like strains examined in this study were isolated in 2008 from rhizosphere soil of J. excelsa collected from Kischka Mountain on the Crimean Peninsula in Ukraine (Global Positioning System (GPS) coordinates: N 44° 24′ 02.07″ E 33° 59′ 32.96″). Three methods were used to isolate the actinomycetes, namely, (i) the direct inoculation of the aqueous-soil suspension, (ii) the pre-treatment of a soil sample with 1.5% aqueous phenol solution and (iii) the heating of the soil sample at 100 °C for 60 min. The obtained aqueous-soil suspension was plated on six different isolation media and then cultured for 14 days at 29 °C (Tistechok et al. ). The isolated strains were deposited in the Microbial Culture Collection of Antibiotic Producers of Ivan Franko National University of Lviv. Oatmeal medium (OM) (Tistechok et al. ) was used for the cultivation of the isolated strains. Liquid tryptic soy broth (TSB) (Sigma-Aldrich, St. Louis, MO, USA) was used for the cultivation of the actinomycete strains required for the total DNA extraction. In addition, SG medium (20 g/L glucose, 10 g/L soy peptone and 2 g/L CaCO 3 ; pH 7.2) and DNPM medium (40 g/L dextrin, 7.5 g/L soy peptone, 5 g/L baker’s yeast and 21 g/L 3–[N-morpholino]propanesulfonic acid; pH 7.2) were used to produce the secondary metabolites. Molecular identification of the actinomycete strains The actinomycete-like strains for the total DNA extraction were cultivated in TSB medium for 3–5 days at a temperature of 28 °C and a shaking rate of 180 rpm. The total DNA extraction procedure utilised in this study has previously been described (Kieser et al. ). The 16S rRNA gene was amplified by means of a polymerase chain reaction (PCR) using the universal primers 8F (5′-AGAGTTTGATYMTGGCTCAG-3′) and 1510R (5′-TACGGYTACCTTGTTACGACTT-3′). In a total volume of 50 μL, the PCR mixture consisted of 5 μL of 10 × PCR buffer, 1.0 μL of deoxynucleoside triphosphates (10.0 mmol/L each), 0.5 μL of each primer (100 pmol/L), 0.5 μL of DNA polymerase (1 U/μL), 2.5 μL of dimethyl sulfoxide, 2.0 μL of DNA template and 38.0 μL of Milli-Q-grade water. The PCR parameters were as follows: initial denaturation at 95 °C for 5 min, followed by 30 cycles of denaturation at 95 °C for 30 s, annealing of the primers at 53 °C for 30 s, extension at 72 °C for 90 s and a final extension at 72 °C for 10 min. The PCR products were visualised on 1% agarose gel. A QIAquick Gel Extraction Kit (Qiagen, Venlo, Netherlands) was used for the purification. The sequencing of the purified samples (5 μL) was performed by GENEWIZ from Azenta Life Sciences (Leipzig, Germany) in accordance with the manufacturer’s recommendations. The obtained 16S rRNA gene sequences were analysed using Geneious version 9.1.2 software (Kearse et al. ). The taxonomic affiliations of the isolated actinomycete strains were determined using Ribosomal Database Project Release 11 ( https://rdp.cme.msu.edu/index.jsp ). The 16S rRNA gene sequences of the isolated actinomycete strains were deposited in the National Center for Biotechnology Information (NCBI)’s GenBank (accession numbers available in Table S1). A phylogenetic tree was constructed using the 16S rRNA gene sequences of the isolated actinomycete strains and their closest neighbours. The closest related species to the 16S rRNA were identified on the basis of the Basic Local Alignment Search Tool data available within the NCBI’s database ( https://blast.ncbi.nlm.nih.gov/Blast.cgi ) and then obtained from the GenBank. Moreover, the Multiple Sequence Comparison by Log-Expectation alignment tool was used to align the sequences (Edgar ). The phylogenetic tree was constructed based on this alignment using the neighbour-joining method with 1000 bootstraps according to the Kimura two-parameter method. This was performed using Molecular Evolutionary Genetics Analysis version 11 software (Tamura et al. ). Screening for antimicrobial activity The primary screening for antimicrobial activity was performed for all of the isolated actinomycetes by means of the spot inoculation method (Tistechok et al. ). More specifically, the actinomycete strains were spot inoculated on OM medium at seven strains per plate around the perimeter and then incubated at 28 °C for 6 days. The antimicrobial activity was tested against the following microorganisms: Staphylococcus aureus ATCC 25,923, Bacillus subtilis ATCC 31,324, Escherichia coli ATCC 25,922, Pseudomonas aeruginosa ATCC 9027, Klebsiella pneumonia ATCC 13,883, Proteus vulgaris ATCC 29,905 and Candida albicans ATCC 885–653. The bacterial strains were grown on Luria agar (10 g/L tryptone, 10 g/L NaCl, 5 g/L yeast extract and 15 g/L agar), whereas the yeast was grown on Sabouraud dextrose agar (Pronadisa (Conga), Madrid, Spain). The freshly grown test microbial strains were inoculated into 15 mL of the appropriate medium and then incubated in the proper conditions at a shaking rate of 180 rpm overnight. Next, the OM plates containing the 6-day-old actinomycete strains were covered with 5 mL of soft agar (0.7% w/v agar) that had previously been inoculated using one of the overnight cultures. The antimicrobial activity was observed on the basis of the appearance of zones revealing the growth inhibition of the test cultures. The antimicrobial activity index (AAI) was determined by measuring the ratio of the diameter of the inhibition zone to the diameter of the colony. The formula for determining the AAI was as follows: [12pt]{minimal}
$$=\;\;()}{\;()}$$ AAI = inhibition zone diameter ( mm ) colony diameter ( mm ) Secondary metabolite extraction from the Streptomyces sp. Je 1–651 strain For the secondary metabolite extraction, the Streptomyces sp. Je 1–651 strain was grown in 15 mL of TSB in a 100-mL flask for 2 days, before 1 mL of pre-culture was inoculated into 100 mL of production SG and DNPM media in a 500-mL flask. The Je 1–561 strain was grown for 7 days at 28 °C and 180 rpm in an Infors Multitron Incubation Shaker (Infors AG, Basel, Switzerland). The secondary metabolites of the Je 1–651 strain were extracted from the culture supernatant using an equal amount of ethyl acetate and acetone:methanol (1:1) mixture from the culture biomass. The obtained extracts were evaporated using an IKA RV-8 Rotary Evaporator (IKA, Staufen, Germany) at 40 °C and then dissolved in methanol. Liquid chromatography–mass spectrometry (LC–MS) and dereplication analysis The secondary metabolite extracts were analysed using a Dionex Ultimate 3000 UPLC System (Thermo Fisher Scientific, Waltham, MA, USA) coupled to a photodiode array (PDA) detector using a 100 mm ACQUITY UPLC BEH C 18 1.7 μm column (Waters Corporation, Milford, MA, USA). A linear gradient (5 to 95%) of a water solution containing 0.1% (volume/volume ( v/v )) formic acid (solvent A) and an acetonitrile solution containing 0.1% ( v/v ) formic acid (solvent B) as the mobile phase was used to separate the extracts at a flow rate of 0.6 mL/min for 18 min. The mass detection and analysis were performed with a Thermo LTQ Orbitrap XL (Thermo Fisher Scientific, Waltham, MA, USA) mass spectrometer using the positive mode of ionisation and a detection range of 200–2000 m/z. The gathered data were analysed using Thermo Xcalibur version 3.0 software. The monoisotopic masses were compared using the Dictionary of Natural Products (DNP) database (Buckingham ) with the following parameters: exact molecular mass, absorption spectrum, source of isolation, fragmentation and physical characteristics (Running ). Compounds were considered similar when the difference between their exact masses was less than 5 ppm and the absorption spectra were identical.
The 372 actinomycete-like strains examined in this study were isolated in 2008 from rhizosphere soil of J. excelsa collected from Kischka Mountain on the Crimean Peninsula in Ukraine (Global Positioning System (GPS) coordinates: N 44° 24′ 02.07″ E 33° 59′ 32.96″). Three methods were used to isolate the actinomycetes, namely, (i) the direct inoculation of the aqueous-soil suspension, (ii) the pre-treatment of a soil sample with 1.5% aqueous phenol solution and (iii) the heating of the soil sample at 100 °C for 60 min. The obtained aqueous-soil suspension was plated on six different isolation media and then cultured for 14 days at 29 °C (Tistechok et al. ). The isolated strains were deposited in the Microbial Culture Collection of Antibiotic Producers of Ivan Franko National University of Lviv. Oatmeal medium (OM) (Tistechok et al. ) was used for the cultivation of the isolated strains. Liquid tryptic soy broth (TSB) (Sigma-Aldrich, St. Louis, MO, USA) was used for the cultivation of the actinomycete strains required for the total DNA extraction. In addition, SG medium (20 g/L glucose, 10 g/L soy peptone and 2 g/L CaCO 3 ; pH 7.2) and DNPM medium (40 g/L dextrin, 7.5 g/L soy peptone, 5 g/L baker’s yeast and 21 g/L 3–[N-morpholino]propanesulfonic acid; pH 7.2) were used to produce the secondary metabolites.
The actinomycete-like strains for the total DNA extraction were cultivated in TSB medium for 3–5 days at a temperature of 28 °C and a shaking rate of 180 rpm. The total DNA extraction procedure utilised in this study has previously been described (Kieser et al. ). The 16S rRNA gene was amplified by means of a polymerase chain reaction (PCR) using the universal primers 8F (5′-AGAGTTTGATYMTGGCTCAG-3′) and 1510R (5′-TACGGYTACCTTGTTACGACTT-3′). In a total volume of 50 μL, the PCR mixture consisted of 5 μL of 10 × PCR buffer, 1.0 μL of deoxynucleoside triphosphates (10.0 mmol/L each), 0.5 μL of each primer (100 pmol/L), 0.5 μL of DNA polymerase (1 U/μL), 2.5 μL of dimethyl sulfoxide, 2.0 μL of DNA template and 38.0 μL of Milli-Q-grade water. The PCR parameters were as follows: initial denaturation at 95 °C for 5 min, followed by 30 cycles of denaturation at 95 °C for 30 s, annealing of the primers at 53 °C for 30 s, extension at 72 °C for 90 s and a final extension at 72 °C for 10 min. The PCR products were visualised on 1% agarose gel. A QIAquick Gel Extraction Kit (Qiagen, Venlo, Netherlands) was used for the purification. The sequencing of the purified samples (5 μL) was performed by GENEWIZ from Azenta Life Sciences (Leipzig, Germany) in accordance with the manufacturer’s recommendations. The obtained 16S rRNA gene sequences were analysed using Geneious version 9.1.2 software (Kearse et al. ). The taxonomic affiliations of the isolated actinomycete strains were determined using Ribosomal Database Project Release 11 ( https://rdp.cme.msu.edu/index.jsp ). The 16S rRNA gene sequences of the isolated actinomycete strains were deposited in the National Center for Biotechnology Information (NCBI)’s GenBank (accession numbers available in Table S1). A phylogenetic tree was constructed using the 16S rRNA gene sequences of the isolated actinomycete strains and their closest neighbours. The closest related species to the 16S rRNA were identified on the basis of the Basic Local Alignment Search Tool data available within the NCBI’s database ( https://blast.ncbi.nlm.nih.gov/Blast.cgi ) and then obtained from the GenBank. Moreover, the Multiple Sequence Comparison by Log-Expectation alignment tool was used to align the sequences (Edgar ). The phylogenetic tree was constructed based on this alignment using the neighbour-joining method with 1000 bootstraps according to the Kimura two-parameter method. This was performed using Molecular Evolutionary Genetics Analysis version 11 software (Tamura et al. ).
The primary screening for antimicrobial activity was performed for all of the isolated actinomycetes by means of the spot inoculation method (Tistechok et al. ). More specifically, the actinomycete strains were spot inoculated on OM medium at seven strains per plate around the perimeter and then incubated at 28 °C for 6 days. The antimicrobial activity was tested against the following microorganisms: Staphylococcus aureus ATCC 25,923, Bacillus subtilis ATCC 31,324, Escherichia coli ATCC 25,922, Pseudomonas aeruginosa ATCC 9027, Klebsiella pneumonia ATCC 13,883, Proteus vulgaris ATCC 29,905 and Candida albicans ATCC 885–653. The bacterial strains were grown on Luria agar (10 g/L tryptone, 10 g/L NaCl, 5 g/L yeast extract and 15 g/L agar), whereas the yeast was grown on Sabouraud dextrose agar (Pronadisa (Conga), Madrid, Spain). The freshly grown test microbial strains were inoculated into 15 mL of the appropriate medium and then incubated in the proper conditions at a shaking rate of 180 rpm overnight. Next, the OM plates containing the 6-day-old actinomycete strains were covered with 5 mL of soft agar (0.7% w/v agar) that had previously been inoculated using one of the overnight cultures. The antimicrobial activity was observed on the basis of the appearance of zones revealing the growth inhibition of the test cultures. The antimicrobial activity index (AAI) was determined by measuring the ratio of the diameter of the inhibition zone to the diameter of the colony. The formula for determining the AAI was as follows: [12pt]{minimal}
$$=\;\;()}{\;()}$$ AAI = inhibition zone diameter ( mm ) colony diameter ( mm )
For the secondary metabolite extraction, the Streptomyces sp. Je 1–651 strain was grown in 15 mL of TSB in a 100-mL flask for 2 days, before 1 mL of pre-culture was inoculated into 100 mL of production SG and DNPM media in a 500-mL flask. The Je 1–561 strain was grown for 7 days at 28 °C and 180 rpm in an Infors Multitron Incubation Shaker (Infors AG, Basel, Switzerland). The secondary metabolites of the Je 1–651 strain were extracted from the culture supernatant using an equal amount of ethyl acetate and acetone:methanol (1:1) mixture from the culture biomass. The obtained extracts were evaporated using an IKA RV-8 Rotary Evaporator (IKA, Staufen, Germany) at 40 °C and then dissolved in methanol.
The secondary metabolite extracts were analysed using a Dionex Ultimate 3000 UPLC System (Thermo Fisher Scientific, Waltham, MA, USA) coupled to a photodiode array (PDA) detector using a 100 mm ACQUITY UPLC BEH C 18 1.7 μm column (Waters Corporation, Milford, MA, USA). A linear gradient (5 to 95%) of a water solution containing 0.1% (volume/volume ( v/v )) formic acid (solvent A) and an acetonitrile solution containing 0.1% ( v/v ) formic acid (solvent B) as the mobile phase was used to separate the extracts at a flow rate of 0.6 mL/min for 18 min. The mass detection and analysis were performed with a Thermo LTQ Orbitrap XL (Thermo Fisher Scientific, Waltham, MA, USA) mass spectrometer using the positive mode of ionisation and a detection range of 200–2000 m/z. The gathered data were analysed using Thermo Xcalibur version 3.0 software. The monoisotopic masses were compared using the Dictionary of Natural Products (DNP) database (Buckingham ) with the following parameters: exact molecular mass, absorption spectrum, source of isolation, fragmentation and physical characteristics (Running ). Compounds were considered similar when the difference between their exact masses was less than 5 ppm and the absorption spectra were identical.
Phylogenetic characterisation of the actinomycete strains In our previous study, 372 actinomycete-like strains were isolated from the rhizosphere soil of J. excelsa . Most of the isolated strains exhibited the typical characteristics of actinomycetes, including slow growth, substrate and aerial mycelium formation, sporulation and pigment production. The taxonomic identification of these strains was evaluated on the basis of their 16S rRNA gene sequences (Table S1). The isolated actinomycete strains were distributed among seven families (Kribbellaceae, Micrococcaceae, Micromonosporaceae, Nocardiaceae, Promicromonosporaceae, Pseudonocardiaceae and Streptomycetaceae) and eleven genera ( Actinoplanes , Actinorectispora , Amycolatopsis , Kribbella , Micrococcus , Micromonospora , Nocardia , Promicromonospora , Rhodococcus , Saccharopolyspora and Streptomyces ). However, the majority of the isolates (350 or 94.08% of all the isolated strains) belonged to the genus Streptomyces . The evolutionary relationships among the isolated actinomycete strains are demonstrated in the phylogenetic tree of the 16S rRNA gene presented in Fig. . All of the isolated strains were grouped within the respective genus and formed close clades with the 16S rRNA gene of representatives of the respective genera (Figs. – ). Based on the evolutionary relationships of the 16S rRNA genes, the isolated strains of the genus Streptomyces were conditionally combined into seven groups. These groups formed clades consisting of both isolates and typical members of the genus Streptomyces and were named after them. Of the seven groups, the S. kanamyceticus group was the largest and included 103 isolated strains. There were also some strains among the isolates that were not combined into the formed groups. The non-streptomyce group included the group of isolates identified as non-streptomycetes and their closest neighbors, as shown in detail in Fig. S6. Analysis of the antimicrobial activity of the actinomycete strains All of the isolated actinomycete strains were tested with regard to their ability to produce antimicrobial metabolites against Gram-positive bacteria, Gram-negative bacteria and yeast using the spot inoculation technique. Among the 372 isolated actinomycete strains, 159 strains (42.74%) exhibited antimicrobial activity against at least one of the tested microbial strains. Most of the strains inhibited the growth of Gram-positive bacteria, including B. subtilis ATCC 31,324 (132 strains, 35.48%) and S. aureus ATCC 25,923 (73 strains, 19.62%). However, significantly fewer strains inhibited the growth of Gram-negative bacteria. Here, 19 strains (5.11%) produced antimicrobial compounds against E. coli ATCC 25,922, whereas 14 strains (3.76%) and 18 strains (4.84%) were active against K. pneumoniae ATCC 13,883 and P. vulgaris ATCC 29,905, respectively, and only seven strains (1.88%) exhibited antagonistic activity against P. aeruginosa ATCC 9027. In addition, 36 strains (9.67%) exhibited antimicrobial activity against C. albicans ATCC 885–653. Aside from ascertaining the ability of the isolated strains to inhibit the growth of a particular test culture, we also evaluated the level of antibiotic activity of these strains, which was calculated as the AAI. For this purpose, we grouped the different AAI into three categories: < 3 (low), 3–6 (medium) and > 6 (high). Most of the studied strains (from 1.08% of K. pneumoniae antagonists to 23.65% of B. subtilis antagonists) had a low AAI, whereas significantly fewer strains had a medium AAI (0.54–10.5%). The exception was the K. pneumoniae antagonists, among which 1.08% had an AAI < 3 while twice as many strains had a medium AAI. Only 2.71% of the strains had an AAI greater than 6, with no AAI greater than 6 being found for the E. coli , P. aeruginosa or C. albicans antagonists (Fig. ). About half of the strains of the genus Streptomyces exhibited antimicrobial activity. Among the strains from other, less numerous genera, only a few showed antimicrobial activity. In particular, two representatives of the genus Amycolatopsis (strains Je 1–447 and Je 1–666) and the Actinorectispora sp. strain Je 1–571 suppressed the growth of the Gram-positive bacteria B. subtilis and S. aureus . Representatives of other genera of the isolated actinomycetes showed no inhibitory activity against the utilised test cultures. Dereplication of the secondary metabolite profile of the Streptomyces sp. strain Je 1–651 After analysing the established collection of actinomycete strains, we identified several strains with a broad spectrum of antimicrobial activity, which captured both antibacterial and antifungal activities. In particular, the Streptomyces sp. Je 1–651 strain exhibited strong inhibitory activity (the AAI of this strain was at a medium level (3–6) or above) against all of the utilised microbial test cultures, except for P. aeruginosa . To better understand the nature of the compounds that may be responsible for the observed activity, we analysed the secondary metabolites produced by this strain. To accomplish this, we performed a dereplication analysis of the secondary metabolites of this strain within the DNP database (CRC Press). The ability of the strain Je 1–651 to produce secondary metabolites was studied by growing it in DNPM and SG liquid media. A total of 18 major secondary metabolite peaks were detected in the cultural liquid and biomass extracts of the Streptomyces sp. Je 1–651 cultivated in both media (Figs. , and Table ). Using the DNP database, seven of these peaks were identified as spiramycins. These peaks were present in all of the studied chromatograms, although their number and magnitude varied. In the crude biomass extract of the Je 1–651 strain, which was grown in SG medium, three large peaks were identified in addition to the spiramycins (Fig. b). These peaks were annotated as stambomycin A/B (retention time (tR) of 9.08; m/z 1376.9378 [M + H] + ), stambomycin C/D (tR of 8.7; m/z 1362.9238 [M + H] + ), stambomycin E (tR of 8.39; m/z 1348.9108 [M + H] + ) and stambomycin F (tR of 9.66; m/z 1390.9558 [M + H] + ) (Figs. b and ). In the crude extracts of the Je 1–651 strain grown on DNPM medium, seven peaks were identified that did not yield positive matches in the DNP database and so could not be annotated based on the available mass spectrometry data. The absence of matches in the database may indicate the novelty of the associated compounds. The characteristics of the identified peaks, which likely form unknown compounds, are shown in Table .
In our previous study, 372 actinomycete-like strains were isolated from the rhizosphere soil of J. excelsa . Most of the isolated strains exhibited the typical characteristics of actinomycetes, including slow growth, substrate and aerial mycelium formation, sporulation and pigment production. The taxonomic identification of these strains was evaluated on the basis of their 16S rRNA gene sequences (Table S1). The isolated actinomycete strains were distributed among seven families (Kribbellaceae, Micrococcaceae, Micromonosporaceae, Nocardiaceae, Promicromonosporaceae, Pseudonocardiaceae and Streptomycetaceae) and eleven genera ( Actinoplanes , Actinorectispora , Amycolatopsis , Kribbella , Micrococcus , Micromonospora , Nocardia , Promicromonospora , Rhodococcus , Saccharopolyspora and Streptomyces ). However, the majority of the isolates (350 or 94.08% of all the isolated strains) belonged to the genus Streptomyces . The evolutionary relationships among the isolated actinomycete strains are demonstrated in the phylogenetic tree of the 16S rRNA gene presented in Fig. . All of the isolated strains were grouped within the respective genus and formed close clades with the 16S rRNA gene of representatives of the respective genera (Figs. – ). Based on the evolutionary relationships of the 16S rRNA genes, the isolated strains of the genus Streptomyces were conditionally combined into seven groups. These groups formed clades consisting of both isolates and typical members of the genus Streptomyces and were named after them. Of the seven groups, the S. kanamyceticus group was the largest and included 103 isolated strains. There were also some strains among the isolates that were not combined into the formed groups. The non-streptomyce group included the group of isolates identified as non-streptomycetes and their closest neighbors, as shown in detail in Fig. S6.
All of the isolated actinomycete strains were tested with regard to their ability to produce antimicrobial metabolites against Gram-positive bacteria, Gram-negative bacteria and yeast using the spot inoculation technique. Among the 372 isolated actinomycete strains, 159 strains (42.74%) exhibited antimicrobial activity against at least one of the tested microbial strains. Most of the strains inhibited the growth of Gram-positive bacteria, including B. subtilis ATCC 31,324 (132 strains, 35.48%) and S. aureus ATCC 25,923 (73 strains, 19.62%). However, significantly fewer strains inhibited the growth of Gram-negative bacteria. Here, 19 strains (5.11%) produced antimicrobial compounds against E. coli ATCC 25,922, whereas 14 strains (3.76%) and 18 strains (4.84%) were active against K. pneumoniae ATCC 13,883 and P. vulgaris ATCC 29,905, respectively, and only seven strains (1.88%) exhibited antagonistic activity against P. aeruginosa ATCC 9027. In addition, 36 strains (9.67%) exhibited antimicrobial activity against C. albicans ATCC 885–653. Aside from ascertaining the ability of the isolated strains to inhibit the growth of a particular test culture, we also evaluated the level of antibiotic activity of these strains, which was calculated as the AAI. For this purpose, we grouped the different AAI into three categories: < 3 (low), 3–6 (medium) and > 6 (high). Most of the studied strains (from 1.08% of K. pneumoniae antagonists to 23.65% of B. subtilis antagonists) had a low AAI, whereas significantly fewer strains had a medium AAI (0.54–10.5%). The exception was the K. pneumoniae antagonists, among which 1.08% had an AAI < 3 while twice as many strains had a medium AAI. Only 2.71% of the strains had an AAI greater than 6, with no AAI greater than 6 being found for the E. coli , P. aeruginosa or C. albicans antagonists (Fig. ). About half of the strains of the genus Streptomyces exhibited antimicrobial activity. Among the strains from other, less numerous genera, only a few showed antimicrobial activity. In particular, two representatives of the genus Amycolatopsis (strains Je 1–447 and Je 1–666) and the Actinorectispora sp. strain Je 1–571 suppressed the growth of the Gram-positive bacteria B. subtilis and S. aureus . Representatives of other genera of the isolated actinomycetes showed no inhibitory activity against the utilised test cultures.
After analysing the established collection of actinomycete strains, we identified several strains with a broad spectrum of antimicrobial activity, which captured both antibacterial and antifungal activities. In particular, the Streptomyces sp. Je 1–651 strain exhibited strong inhibitory activity (the AAI of this strain was at a medium level (3–6) or above) against all of the utilised microbial test cultures, except for P. aeruginosa . To better understand the nature of the compounds that may be responsible for the observed activity, we analysed the secondary metabolites produced by this strain. To accomplish this, we performed a dereplication analysis of the secondary metabolites of this strain within the DNP database (CRC Press). The ability of the strain Je 1–651 to produce secondary metabolites was studied by growing it in DNPM and SG liquid media. A total of 18 major secondary metabolite peaks were detected in the cultural liquid and biomass extracts of the Streptomyces sp. Je 1–651 cultivated in both media (Figs. , and Table ). Using the DNP database, seven of these peaks were identified as spiramycins. These peaks were present in all of the studied chromatograms, although their number and magnitude varied. In the crude biomass extract of the Je 1–651 strain, which was grown in SG medium, three large peaks were identified in addition to the spiramycins (Fig. b). These peaks were annotated as stambomycin A/B (retention time (tR) of 9.08; m/z 1376.9378 [M + H] + ), stambomycin C/D (tR of 8.7; m/z 1362.9238 [M + H] + ), stambomycin E (tR of 8.39; m/z 1348.9108 [M + H] + ) and stambomycin F (tR of 9.66; m/z 1390.9558 [M + H] + ) (Figs. b and ). In the crude extracts of the Je 1–651 strain grown on DNPM medium, seven peaks were identified that did not yield positive matches in the DNP database and so could not be annotated based on the available mass spectrometry data. The absence of matches in the database may indicate the novelty of the associated compounds. The characteristics of the identified peaks, which likely form unknown compounds, are shown in Table .
The rapid increase in the number of MDR pathogens seen in recent decades has stimulated the development of new therapeutic agents to combat them. Moreover, the urgent need for new antibiotics in a clinical setting has led to a revival of screening from nature. Microbial natural products are recognised as being among the most important elements when it comes to drug discovery (Pham et al. ). In this study, we analysed the phylogenetic diversity and antibiotic properties of actinomycete isolates from the rhizosphere soil of J. excelsa . The medicinal properties of J. excelsa are a reason for our interest in the rhizosphere actinomycetes of this species. The traditional Persian medicine used drugs based on these plants (Sargin et al. ). Extracts from different parts of this plant contain numerous metabolites that have antibacterial, antifungal, antiparasitic and other activities (Almaarri et al. ; Moein et al. ). Our preliminary studies of the metabolic potential of individual isolates of the juniper rhizosphere enabled us to identify several new compounds. In particular, leopolylic acid A (Raju et al. ) which is considered as a potential protease inhibitor (M pro ) of SARS-CoV-2 (Mazzini et al. ), juniperolide A (Raju et al. ) and furaquinocins K and L (Tistechok et al. ). However, the phylogenetic diversity of actinomycetes of the Crimean Peninsula, in particular in the rhizosphere of J. excelsa , has not yet been described. Our taxonomic identification and phylogenetic analysis revealed among the assessed isolates representatives of seven families, including eleven genera of class Actinomycetia . Yet, Streptomyces was the most dominant genus among these isolates. This finding is unsurprising, as representatives of this genus are widely distributed in terrestrial ecosystems, especially in soil, where they play an important role in soil formation (Goodfellow and Williams ). Furthermore, as noted above, streptomycetes are a potential source of biologically active compounds, which may lead to further research into which new compounds may be identified. We also isolated several non-streptomycete strains, among which we found representatives of some rather rare genera. In particular, we obtained three isolates (Je 1–106, Je 1–148 and Je 1–339) of the genus Promicromonospora (19 species have been identified to date ( https://lpsn.dsmz.de/genus/Promicromonospora )) and strain Je 1–571 representing the genus Actinorectispora . This recently discovered genus of Actinomycetia currently has only two typical strains, namely, A. indica (Quadri et al. ) and A. metalli (Cao et al. ). Currently, there is no data on the range of secondary metabolites that can be produced by members of this genus. We also managed to identify members of the Amycolatopsis and Saccharopolyspora genera of the Pseudonocardiaceae, whose representatives are characterised by a great diversity of biosynthetic gene clusters and considered a promising source of new natural products (Gavriilidou et al. ). In future studies, a genome analysis of the actinomycete strains of these genera, as well as a chemical analysis of their metabolites, will allow us to reveal their biosynthetic potential. Furthermore, the presence of rare genera in the collection of strains isolated from the rhizosphere soil of J. excelsa , particularly the Actinorectispora genus, is promising and will contribute to further studies of the actinomycete diversity in this region. The phylogenetic diversity of the actinomycete strains isolated from the rhizosphere of J. excelsa may be a guarantee of the rich composition of the associated biologically active compounds, among which there is a high probability of new compounds being discovered. One of the simplest approaches to evaluating a large collection of strains in terms of their ability to produce antimicrobial substances involves studying their antagonistic properties against specific test cultures (Balouiri et al. ). The actinomycete strains that we tested represented antagonists that specifically inhibited either Gram-positive (e.g. Je 1–101, Je 1–253 and others), Gram-negative bacteria (e.g. Je 1–208, Je 1–392 and others) or yeasts (e.g. Je 1–22.2, Je 1–93 and others). At the same time, some strains inhibited most or all of the utilised test cultures (e.g. Je 1–620, Je 1–651 and others) (Table S1). As previously noted, most of the strains that exhibited antibiotic activity were streptomycetes. Moreover, only some strains of the family Pseudonocardiaceae inhibited the growth of Gram-positive bacteria. Clearly, the non-streptomycete strains require more special fermentation conditions when compared with the streptomycetes (Amin et al. ). The wide spectrum of antimicrobial activity observed in this study indicates prerequisites for a deeper investigation of individual isolated strains. As mentioned above, some of the studied strains demonstrated a wide range of antibiotic activity. Among them was the Je 1–651 strain, which exhibited the highest level of antimicrobial activity against the test cultures, except for P. aeruginosa . Additionally, we performed a dereplicative analysis of one promising strain that showed strong antimicrobial activity against the utilised test cultures. As a result, we were able to isolate the macrolide antibiotics spiramycin (Karray et al. ) and the 51-membered glycosylated macrolide stambomycin (Laureti et al. ). The dereplicative analysis allowed us to partially relate the observed biological activity of the Je 1–651 strain to specific compounds. We assume that the ability of the Streptomyces sp. Je 1–651 to inhibit the growth of Gram-positive bacteria, Gram-negative bacteria and yeast is due to its production of spiramycins, given their wide spectrum of activity (Labro ). It is also worth noting that spiramycins and stambomycins are produced by the S. ambofaciens strain (Aigle et al. ), which led us to assume that the Je 1–651 strain is perhaps phylogenetically related to the S. ambofaciens strains. A phylogenetic analysis based on the 16S rRNA gene sequence of the Je 1–651 strain showed similarity (100% identity) with type strains of S. ambofaciens . They also form a common clade on the phylogenetic tree (Fig. S8). Given the results of the phylogenetic analysis and secondary metabolite dereplication, we assume that Je 1–651 may be quite close to the S. ambofaciens strains. A detailed genomic analysis and comparison of these strains would improve our understanding of the similarities and differences between these strains. We do not exclude that the biosynthetic gene clusters of spiramycin and stambomycin may have differences, as was shown by the evolution of the antimycin biosynthetic gene cluster (Joynt and Seipke ). Such changes may provide selective advantages in the appropriate habitat. As the antibiotic spiramycin is in widespread use, further research on this strain is required, especially with respect to spiramycin production levels. In addition, the identification of new producers also has practical implications, as they may prove to be more convenient (e.g. fast growth, a higher production or are better suited for gene engineering manipulations) than currently utilised strains. Furthermore, the use of different production media allowed us to examine the potential of the Je 1–651 strain in greater depth. Aside from spiramycin and stambomycin, several secondary metabolic peaks were detected in the extract of the Je 1–651 strain that we were unable to identify by means of the dereplicative analysis. This may indicate the potential novelty of these compounds, the structures of which we will try to determine in future studies.
In the present study, we demonstrated the diversity and bioactive potential of actinomycetes previously isolated from the rhizosphere soil of J. excelsa . Among the isolated strains, 11 genera were identified, including the quite rare genus Actinorectispora . Screening the antimicrobial activity of the isolated strains revealed their potential as producers of biologically active compounds. Moreover, the metabolic profiling of the widely active Streptomyces sp. Je 1–651 strain led to the identification of the commercially used antibiotic spiramycin as well as several potential new compounds. Thus, the findings of this study support the notion that actinomycete strains from unique ecosystems may be a potential source of producers of both pharmaceutical and biotechnological interests.
Below is the link to the electronic supplementary material. Supplementary file1 (DOCX 439 KB)
|
Management of wounds with exposed bone structures using an induced‐membrane followed by polymethyl methacrylate and second‐stage skin grafting in the elderly with a 3‐year follow‐up | 4ae4747c-6c12-4d61-ab13-3b4ce37d93e0 | 10031252 | Debridement[mh] | INTRODUCTION Victims of high‐speed vehicular accidents often present with large degloving injuries, compound fractures and severe damage to soft tissues. , These wounds are frequently left open and require repeated debridement, resulting in large soft‐tissue defects. Traumatic wounds with exposed bone or tendon heal poorly unless aggressive debridement or free‐flap microsurgery is performed, which poses a substantial reconstructive challenge. Moreover, the use of these techniques is, in practice, frequently barred by the magnitude of the tissue defect and/or the paucity of a donor site. In addition, free‐flap operations are complex, technically demanding, costly and time‐consuming, with significant rates of complications, donor‐site morbidities and failure. , , Furthermore, the patient's general condition may be too poor to tolerate prolonged free flap surgery, and a donor site may not be available for an effective flap. Before the development of microsurgical techniques, the treatment of severe lower‐extremity wounds often consisted of skin grafting. In complex wounds, skin grafting is usually inadequate because of the difficulty of healing over exposed bone, leading to high rates of osteomyelitis and amputation. Chen et al developed a method for managing these wounds in stages, using an artificial dermis and skin grafting technique, and tested its feasibility in 17 wounds in 15 patients. The authors concluded that in the management of a non‐healing wound, specifically in the lower extremity, a staged approach using an artificial dermis followed by a skin graft is effective, especially in cases involving exposed bony structures. However, the disadvantages included a risk of immune rejection, prolonged hospitalisation and an increased surgical risk. Masquelet's induced membrane (IM) technique, which uses a two‐stage procedure to surgically reconstruct segmental bone defects, is relatively new. , In this approach, the second surgery is performed only after the polymethyl methacrylate (PMMA) spacer has elicited a local foreign‐body response and an autologous foreign‐body IM, the key component of the procedure, has formed around the spacer. The presence of pro‐angiogenic factors in association with the IM, including vascular endothelial growth factor (VEGF), angiotensin II and fibroblast growth factor 2, has been documented. , Recently, Tang et al reported elevated levels of the anti‐angiogenic Notch signalling components DLL4 and NOTCH1 in more mature membranes. Vessel density in the IM increases during the first few weeks to months after the first surgery, which is evidence that an IM can provide an effective and sufficient blood supply for soft‐tissue and bone reconstruction. Here we present a two‐stage strategy for wound management, comprising initial coverage of a fresh wound with PMMA, followed by grafting with an autologous split‐thickness skin graft under the IM. The experience gained from the treatment of 50 patients using this technique with a 3‐year follow‐up is discussed herein.
MATERIALS AND METHODS 2.1 Patient materials Fifty patients (35 males, 15 females) ranging in age from 60 to 78 years (mean: 68.54 ± 5.49 years) were treated from December 2016 to December 2019. All wounds were located in the lower extremities, and all involved exposure of the underlying bony or tendon structures. Car accidents were the most common cause of injuries (35 patients), followed by crush injuries (7 patients) and falls (8 patients). General information on the patients is presented in Table . All experimental procedures were approved by the ethics committee of our hospital and were conducted in accordance with the Declaration of Helsinki. The requirement for informed consent was waived as the data were analysed anonymously and personal identifiers were completely removed. 2.2 Surgical procedure Our reconstruction strategy is based on a two‐stage process. During the first stage, the entire defect area is thoroughly inspected and aggressively debrided. Surgical debridement is performed in all patients, in addition to the removal of devitalised tissues. For wounds without periosteal coverage, the bony or tendon surface is lightly abraded (Figure ). The exposed bone is decorticalised (Figure ), and fluid and tissue samples are sent for microscopy, culture, and antibiotic sensitivity testing, with or without histology as indicated. In patients with fractures, the appropriate fixation method is chosen. The suitability of the implant is evaluated and possibly revised. Internal or external methods of fixation are appropriate, as long as the construction is stable enough to minimise disturbance of the cement spacer and IM formation without risking infection. If only bone or tendon is exposed, external fixation (cast or brace) after surgery may suffice. The freshly treated wound is covered with PMMA bone cement (Figure ) loaded with antibiotics (vancomycin). The amount of cement used will depend on the size of the wound but should be larger than the actual wound area. The cement is anchored tightly before it solidifies by stapling along the wound edges. Antibiotics against most gram‐positive variants are administered intravenously to all patients for 3‐5 days postoperatively. The patient can then be discharged after wound covering. After 4‐6 weeks, during the second stage, the PMMA cement is removed and the wound base lightly abraded to determine the adequacy of revascularisation; light blood seepage is usually observed in a mature wound bed. At this stage, the IM should cover the exposed bone and tendon (Figure ), forming a mature IM. After confirmation that neither the IM nor the wound is infected, a 1 × 1 cm piece of IM is removed for laboratory testing. The remaining IM is removed as well and fixed in 10% paraformaldehyde for pathological examination. In this study, normal periosteal membrane (PM) was harvested from adult patients with fractures as a control. An autologous split‐thickness piece of skin (0.15‐0.20‐mm thick) is harvested using an electric dermatome to cover the resultant wound above the IM. The grafted wound is managed routinely, with the graft anchored to the wound base with sutures or staples and covered with a piece of petrolatum gauze. Finally, the surgical area is closed using pressure. After 6‐7 days, the wound is uncovered to ascertain the survival of the skin graft. The patient is observed for the next 6‐12 months to ensure proper recovery from the procedure and healing of the injuries. The antibiotic therapy was first managed empirically and then modified according to the results of an antibiogram from the intraoperative culture. All patients were regularly tracked by our multidisciplinary team until 36 months. 2.3 Histology and immunohistochemistry Histological sections of the IM and normal PM were stained with haematoxylin and eosin (H&E) and analysed by two clinical pathologists, who assessed IM vascularisation via immunohistochemistry (IHC) detection of VEGF, CD31 and CD34. Paraffin sections were prepared by sequential washes (100% xylene, 20 minutes thrice; 100% alcohol, 5 minutes; 95% alcohol, 5 minutes; 85% alcohol, 5 minutes; 75% alcohol, 5 minutes; distilled water, 5 minutes), followed by antigen retrieval and incubation first with the primary antibody (VEGF, CD31 or CD34) and then with the secondary antibody (anti‐rabbit). Finally, the IHC sections were stained with DAB and haematoxylin, dehydrated with alcohol and examined by microscopy at ×400 magnification (DMIL‐LED, LEICA, Wetzlar, Germany). Positive immunostaining of the cells was determined quantitatively. 2.4 Clinical observational indicators The primary outcomes were the healing rate. Healing was determined to be complete epithelialisation of the surgical wound at two consecutive clinic visits. Non‐healing was defined as no significant reduction in the wound size or no significant decrease or worsening of secretions, with no need for or refusal for major amputation. The secondary outcomes included the duration of healing (the number of weeks from the initial surgical intervention to the date of complete healing), frequency of debridement procedures, and reulceration (the appearance of a new ulcer on the same or contralateral limb during the follow‐up). The psychological status of the patients was determined using the self‐rating anxiety scale (SAS), which was administered at the last follow‐up. The SAS has a total possible score of 80 points. Patients with a score <50 are regarded as normal, those scoring 50‐59 as slightly anxious or depressed, those scoring 60‐69 points as moderately anxious or depressed, and those scoring >70 points as severely anxious or depressed. The functional recovery was determined 12 months after IM and autologous skin grafting by using the Lower Extremity Function Scale (LEFS) survey, including the determination of internal consistency, reliability, construct validity, sensitivity to change, and clinical application. The LEFS is a validated scoring system that can track changes in lower limb function on a numerical scale from 0 to 80. Patients were also monitored for the following complications: skin graft necrosis, infections, blisters, chronic ulcers and hypertrophic scarring at the donor site.
Patient materials Fifty patients (35 males, 15 females) ranging in age from 60 to 78 years (mean: 68.54 ± 5.49 years) were treated from December 2016 to December 2019. All wounds were located in the lower extremities, and all involved exposure of the underlying bony or tendon structures. Car accidents were the most common cause of injuries (35 patients), followed by crush injuries (7 patients) and falls (8 patients). General information on the patients is presented in Table . All experimental procedures were approved by the ethics committee of our hospital and were conducted in accordance with the Declaration of Helsinki. The requirement for informed consent was waived as the data were analysed anonymously and personal identifiers were completely removed.
Surgical procedure Our reconstruction strategy is based on a two‐stage process. During the first stage, the entire defect area is thoroughly inspected and aggressively debrided. Surgical debridement is performed in all patients, in addition to the removal of devitalised tissues. For wounds without periosteal coverage, the bony or tendon surface is lightly abraded (Figure ). The exposed bone is decorticalised (Figure ), and fluid and tissue samples are sent for microscopy, culture, and antibiotic sensitivity testing, with or without histology as indicated. In patients with fractures, the appropriate fixation method is chosen. The suitability of the implant is evaluated and possibly revised. Internal or external methods of fixation are appropriate, as long as the construction is stable enough to minimise disturbance of the cement spacer and IM formation without risking infection. If only bone or tendon is exposed, external fixation (cast or brace) after surgery may suffice. The freshly treated wound is covered with PMMA bone cement (Figure ) loaded with antibiotics (vancomycin). The amount of cement used will depend on the size of the wound but should be larger than the actual wound area. The cement is anchored tightly before it solidifies by stapling along the wound edges. Antibiotics against most gram‐positive variants are administered intravenously to all patients for 3‐5 days postoperatively. The patient can then be discharged after wound covering. After 4‐6 weeks, during the second stage, the PMMA cement is removed and the wound base lightly abraded to determine the adequacy of revascularisation; light blood seepage is usually observed in a mature wound bed. At this stage, the IM should cover the exposed bone and tendon (Figure ), forming a mature IM. After confirmation that neither the IM nor the wound is infected, a 1 × 1 cm piece of IM is removed for laboratory testing. The remaining IM is removed as well and fixed in 10% paraformaldehyde for pathological examination. In this study, normal periosteal membrane (PM) was harvested from adult patients with fractures as a control. An autologous split‐thickness piece of skin (0.15‐0.20‐mm thick) is harvested using an electric dermatome to cover the resultant wound above the IM. The grafted wound is managed routinely, with the graft anchored to the wound base with sutures or staples and covered with a piece of petrolatum gauze. Finally, the surgical area is closed using pressure. After 6‐7 days, the wound is uncovered to ascertain the survival of the skin graft. The patient is observed for the next 6‐12 months to ensure proper recovery from the procedure and healing of the injuries. The antibiotic therapy was first managed empirically and then modified according to the results of an antibiogram from the intraoperative culture. All patients were regularly tracked by our multidisciplinary team until 36 months.
Histology and immunohistochemistry Histological sections of the IM and normal PM were stained with haematoxylin and eosin (H&E) and analysed by two clinical pathologists, who assessed IM vascularisation via immunohistochemistry (IHC) detection of VEGF, CD31 and CD34. Paraffin sections were prepared by sequential washes (100% xylene, 20 minutes thrice; 100% alcohol, 5 minutes; 95% alcohol, 5 minutes; 85% alcohol, 5 minutes; 75% alcohol, 5 minutes; distilled water, 5 minutes), followed by antigen retrieval and incubation first with the primary antibody (VEGF, CD31 or CD34) and then with the secondary antibody (anti‐rabbit). Finally, the IHC sections were stained with DAB and haematoxylin, dehydrated with alcohol and examined by microscopy at ×400 magnification (DMIL‐LED, LEICA, Wetzlar, Germany). Positive immunostaining of the cells was determined quantitatively.
Clinical observational indicators The primary outcomes were the healing rate. Healing was determined to be complete epithelialisation of the surgical wound at two consecutive clinic visits. Non‐healing was defined as no significant reduction in the wound size or no significant decrease or worsening of secretions, with no need for or refusal for major amputation. The secondary outcomes included the duration of healing (the number of weeks from the initial surgical intervention to the date of complete healing), frequency of debridement procedures, and reulceration (the appearance of a new ulcer on the same or contralateral limb during the follow‐up). The psychological status of the patients was determined using the self‐rating anxiety scale (SAS), which was administered at the last follow‐up. The SAS has a total possible score of 80 points. Patients with a score <50 are regarded as normal, those scoring 50‐59 as slightly anxious or depressed, those scoring 60‐69 points as moderately anxious or depressed, and those scoring >70 points as severely anxious or depressed. The functional recovery was determined 12 months after IM and autologous skin grafting by using the Lower Extremity Function Scale (LEFS) survey, including the determination of internal consistency, reliability, construct validity, sensitivity to change, and clinical application. The LEFS is a validated scoring system that can track changes in lower limb function on a numerical scale from 0 to 80. Patients were also monitored for the following complications: skin graft necrosis, infections, blisters, chronic ulcers and hypertrophic scarring at the donor site.
RESULTS All wounds healed successfully. The smallest wound was approximately 2 × 3 cm and the largest, 12 × 6 cm (Table ). The reasons for not performing a flap are presented in Table . A mature IM developed in 48 of 50 patients, and all 48 were lightly revascularised 4‐6 weeks postoperatively. The wounds of the remaining 2 patients did not form a mature IM. In those cases, after a second complete debridement followed by bone decortication and replacement of the PMMA cement, a mature IM appeared after 4‐6 weeks. 3.1 Gross histology An intensely fibrous, cell‐rich, vascularised tissue was seen in the IM by H&E staining (Figure ), whereas in the normal PM, there were few vessel‐like structures. 3.2 Vascularisation Vascularisation in the IM and PM was assessed by IHC to determine the percentages of VEGF‐, CD31‐ and CD34‐positive cells, containing brown or yellow particles. As shown in Figure , the positive cells were mainly scattered and formed tube‐like vascular structures in the IM. The qualitative results indicated significantly higher numbers of VEGF‐, CD31‐ and CD34‐positive cells in the IM than in the PM. 3.3 Clinical outcomes At the final follow‐up, all wounds were well healed. The duration of healing in the group was 5.4 ± 1.32 months, with a mean number of debridement procedures of 1.92 ± 0.60. There were two patients with reulceration in the group, in whom the wound was located on the heel. Both of them healed after multiple dressing changes. According to the SAS, all patients had comparable levels of severe anxiety and depression preoperatively, with SAS scores ranging from 35 to 60 (mean 48.02 ± 8.12). Postoperatively, the LEFS score was 57.83 ± 11.89. 3.4 Case report 3.4.1 Case 1 A 65‐year‐old male suffered a machine‐related crush injury, resulting in a soft‐tissue defect of the lateral malleolus and heel, as well as an exposed calcaneus (Figure ). The involved heel area was approximately 10 × 5 cm. No fracture in either the calcaneus or ankle joint was seen by X‐ray (Figure ). An ultrasound of the blood vessels in the lower extremities revealed varicose veins with thrombosis as well as arteriosclerosis with atherosclerotic plaque. A pulmonary function test indicated that the patient's lung function was poor. Thus, in this patient, a free flap was not the best choice. Instead, after thorough debridement, the wound with exposed bone was covered with PMMA cement (Figure ). Five weeks later, the PMMA cement was removed, revealing an IM over the wound that was rich in blood vessels with freshly oozing blood and an elastic and moist surface (Figure ). An autologous partial‐thickness skin graft was used to cover the IM, followed by bandage compression (Figure ). At the 1‐month (Figure ) and 6‐month (Figure ) follow‐up examinations, good wound healing was observed. The patient was very satisfied with the outcome. 3.4.2 Case 2 A 72‐year‐old male suffered a skin contusion of the left ankle during a car accident (Figure ). The injured area also had extensive sulfuric acid burns. Emergency debridement of the wounds revealed two skin defects, one at the lateral malleolus (12 × 6 cm) that included an exposed tendon and another at the medial malleolus (6 × 5 cm) with exposed bone (Figure ). Due to the poor local conditions at the burn site and the patient's lower‐extremity arteriosclerosis, a pedicled or random flap was, in our opinion, contraindicated. Thus, the wound with the exposed bone was thoroughly debrided and then covered with PMMA cement (Figure ). Four weeks later, after removal of the bone cement, the IM was clearly visible (Figure ). An autologous partial‐thickness skin graft was then used to treat the newly freshened wound. Three months later, the patient's ankle had healed well, and the patient was satisfied with the outcome (Figure ). 3.4.3 Case 3 Following a tractor accident, a 60‐year‐old male suffered an extensive skin degloving injury of the right lower extremity, from the thigh down to the ankle, involving the deep fascia and muscle plane and exposing the tibia and fibula (Figure ), as well as an open fracture of the right ankle joint. During emergency surgery of the right lower limb, the necrotic tissue and skin were completely removed during thorough wound debridement. A split‐thickness skin piece harvested from the left thigh was used as the skin graft. The exposed bone in a skin defect (10 × 4 cm) on the lateral side of the knee was covered with PMMA cement (Figure ). Two weeks later, exudate was still present on the wound surface, indicating that the debridement had not been sufficiently thorough. Thus, a second debridement and replacement of the bone cement were performed (Figure ). Six weeks later, the cement was removed, revealing an IM rich in new blood vessels on the exposed bone surface (Figure ). An autologous partial‐thickness skin graft was used to cover the IM, followed by bandage compression. At the last follow‐up, the inner thigh skin graft area was still viable (Figure ). All wounds healed well after the skin graft (Figure ).
Gross histology An intensely fibrous, cell‐rich, vascularised tissue was seen in the IM by H&E staining (Figure ), whereas in the normal PM, there were few vessel‐like structures.
Vascularisation Vascularisation in the IM and PM was assessed by IHC to determine the percentages of VEGF‐, CD31‐ and CD34‐positive cells, containing brown or yellow particles. As shown in Figure , the positive cells were mainly scattered and formed tube‐like vascular structures in the IM. The qualitative results indicated significantly higher numbers of VEGF‐, CD31‐ and CD34‐positive cells in the IM than in the PM.
Clinical outcomes At the final follow‐up, all wounds were well healed. The duration of healing in the group was 5.4 ± 1.32 months, with a mean number of debridement procedures of 1.92 ± 0.60. There were two patients with reulceration in the group, in whom the wound was located on the heel. Both of them healed after multiple dressing changes. According to the SAS, all patients had comparable levels of severe anxiety and depression preoperatively, with SAS scores ranging from 35 to 60 (mean 48.02 ± 8.12). Postoperatively, the LEFS score was 57.83 ± 11.89.
Case report 3.4.1 Case 1 A 65‐year‐old male suffered a machine‐related crush injury, resulting in a soft‐tissue defect of the lateral malleolus and heel, as well as an exposed calcaneus (Figure ). The involved heel area was approximately 10 × 5 cm. No fracture in either the calcaneus or ankle joint was seen by X‐ray (Figure ). An ultrasound of the blood vessels in the lower extremities revealed varicose veins with thrombosis as well as arteriosclerosis with atherosclerotic plaque. A pulmonary function test indicated that the patient's lung function was poor. Thus, in this patient, a free flap was not the best choice. Instead, after thorough debridement, the wound with exposed bone was covered with PMMA cement (Figure ). Five weeks later, the PMMA cement was removed, revealing an IM over the wound that was rich in blood vessels with freshly oozing blood and an elastic and moist surface (Figure ). An autologous partial‐thickness skin graft was used to cover the IM, followed by bandage compression (Figure ). At the 1‐month (Figure ) and 6‐month (Figure ) follow‐up examinations, good wound healing was observed. The patient was very satisfied with the outcome. 3.4.2 Case 2 A 72‐year‐old male suffered a skin contusion of the left ankle during a car accident (Figure ). The injured area also had extensive sulfuric acid burns. Emergency debridement of the wounds revealed two skin defects, one at the lateral malleolus (12 × 6 cm) that included an exposed tendon and another at the medial malleolus (6 × 5 cm) with exposed bone (Figure ). Due to the poor local conditions at the burn site and the patient's lower‐extremity arteriosclerosis, a pedicled or random flap was, in our opinion, contraindicated. Thus, the wound with the exposed bone was thoroughly debrided and then covered with PMMA cement (Figure ). Four weeks later, after removal of the bone cement, the IM was clearly visible (Figure ). An autologous partial‐thickness skin graft was then used to treat the newly freshened wound. Three months later, the patient's ankle had healed well, and the patient was satisfied with the outcome (Figure ). 3.4.3 Case 3 Following a tractor accident, a 60‐year‐old male suffered an extensive skin degloving injury of the right lower extremity, from the thigh down to the ankle, involving the deep fascia and muscle plane and exposing the tibia and fibula (Figure ), as well as an open fracture of the right ankle joint. During emergency surgery of the right lower limb, the necrotic tissue and skin were completely removed during thorough wound debridement. A split‐thickness skin piece harvested from the left thigh was used as the skin graft. The exposed bone in a skin defect (10 × 4 cm) on the lateral side of the knee was covered with PMMA cement (Figure ). Two weeks later, exudate was still present on the wound surface, indicating that the debridement had not been sufficiently thorough. Thus, a second debridement and replacement of the bone cement were performed (Figure ). Six weeks later, the cement was removed, revealing an IM rich in new blood vessels on the exposed bone surface (Figure ). An autologous partial‐thickness skin graft was used to cover the IM, followed by bandage compression. At the last follow‐up, the inner thigh skin graft area was still viable (Figure ). All wounds healed well after the skin graft (Figure ).
Case 1 A 65‐year‐old male suffered a machine‐related crush injury, resulting in a soft‐tissue defect of the lateral malleolus and heel, as well as an exposed calcaneus (Figure ). The involved heel area was approximately 10 × 5 cm. No fracture in either the calcaneus or ankle joint was seen by X‐ray (Figure ). An ultrasound of the blood vessels in the lower extremities revealed varicose veins with thrombosis as well as arteriosclerosis with atherosclerotic plaque. A pulmonary function test indicated that the patient's lung function was poor. Thus, in this patient, a free flap was not the best choice. Instead, after thorough debridement, the wound with exposed bone was covered with PMMA cement (Figure ). Five weeks later, the PMMA cement was removed, revealing an IM over the wound that was rich in blood vessels with freshly oozing blood and an elastic and moist surface (Figure ). An autologous partial‐thickness skin graft was used to cover the IM, followed by bandage compression (Figure ). At the 1‐month (Figure ) and 6‐month (Figure ) follow‐up examinations, good wound healing was observed. The patient was very satisfied with the outcome.
Case 2 A 72‐year‐old male suffered a skin contusion of the left ankle during a car accident (Figure ). The injured area also had extensive sulfuric acid burns. Emergency debridement of the wounds revealed two skin defects, one at the lateral malleolus (12 × 6 cm) that included an exposed tendon and another at the medial malleolus (6 × 5 cm) with exposed bone (Figure ). Due to the poor local conditions at the burn site and the patient's lower‐extremity arteriosclerosis, a pedicled or random flap was, in our opinion, contraindicated. Thus, the wound with the exposed bone was thoroughly debrided and then covered with PMMA cement (Figure ). Four weeks later, after removal of the bone cement, the IM was clearly visible (Figure ). An autologous partial‐thickness skin graft was then used to treat the newly freshened wound. Three months later, the patient's ankle had healed well, and the patient was satisfied with the outcome (Figure ).
Case 3 Following a tractor accident, a 60‐year‐old male suffered an extensive skin degloving injury of the right lower extremity, from the thigh down to the ankle, involving the deep fascia and muscle plane and exposing the tibia and fibula (Figure ), as well as an open fracture of the right ankle joint. During emergency surgery of the right lower limb, the necrotic tissue and skin were completely removed during thorough wound debridement. A split‐thickness skin piece harvested from the left thigh was used as the skin graft. The exposed bone in a skin defect (10 × 4 cm) on the lateral side of the knee was covered with PMMA cement (Figure ). Two weeks later, exudate was still present on the wound surface, indicating that the debridement had not been sufficiently thorough. Thus, a second debridement and replacement of the bone cement were performed (Figure ). Six weeks later, the cement was removed, revealing an IM rich in new blood vessels on the exposed bone surface (Figure ). An autologous partial‐thickness skin graft was used to cover the IM, followed by bandage compression. At the last follow‐up, the inner thigh skin graft area was still viable (Figure ). All wounds healed well after the skin graft (Figure ).
DISCUSSION The treatment of non‐healing wounds with exposed bone or tendon is a surgical challenge, , although appropriate treatment has long been the focus of considerable research, and the goals (early irrigation and debridement followed by early soft‐tissue coverage) have remained essentially the same. Free tissue transfer allows the greatest surface coverage and the most flexibility in terms of flap placement; thus, it is still the gold standard for wound coverage. Moreover, good long‐term results with respect to limb or joint function have been reported. However, free‐tissue transfer is a lengthy and expensive procedure that requires specialised practitioners as well as postoperative monitoring in the intensive care unit. , Parrett et al reviewed the changing treatment protocols used by their institution to treat traumatic wounds with exposed bone or tendon. Among the changes in practice was a trend of less reconstruction, such that the use of free flaps has become less common, closure and skin graft procedures are intentionally delayed, and closures are frequently supported by the use of a vacuum‐assisted closure sponge. Less‐invasive treatments with equally efficacious results, such as artificial dermis and skin grafting and a skin graft with perifascial areolar tissue, are appealing but pose risks of immune rejection and prolonged hospitalisation. The use of a composite tissue transfer, while preferred, may not be possible because of the lack of a graft donor site and/or the patient's poor physical status. In their animal studies, Koga et al showed that the periosteum is an essential tissue component of a wound bed for an exposed bone requiring a skin graft. In the absence of a periosteal covering, the result will be poor and the application of a full‐thickness skin graft over the bony or tendon surface will be required. Neo‐vascularisation induced by decorticating the bone or drilling multiple holes into the medullary cavity, for instance, is unreliable and the extent of vascularised bed growth unpredictable, with an inadequate bed possibly causing severe disabilities due to the need to resect functional bone or tendon postoperatively. Although negative pressure wound therapy is a well‐established method to promote wound healing and its efficacy is well accepted by many clinicians, it is often used as a bridging device for the closure of a skin graft or skin flap. , Therefore, the formation of vascularised granulation tissue on the exposed areas may be slower, which can be a burden on patients. In this study, the angiogenic potential of the IM surrounding the PMMA cement was exploited in combination with skin grafting to repair exposed open wounds in patients with injuries resulting in exposed bone or tendon wounds with a periosteal defect. The outcomes ranged from good to excellent in nearly all patients. The advantage of our approach is that, because the wound surface is covered by bone cement, the patient can be discharged from the hospital before IM formation, thus reducing the length and cost of hospitalisation. The IM is an initially avascular bed that supports the survival of the skin graft, and it differs from the PM. VEGF and CD31, produced by platelets, stimulate the growth of vascular endothelial cells, thereby accelerating revascularisation and enhancing the blood supply to ischemic flaps, thus increasing their survival and, in turn, wound coverage. , IHC showed the greater vascularisation potential of the IM compared with the PM due to the higher‐level expression of VEGF, CD31 and CD34. IM vascularisation has also been demonstrated in other human and animal studies. , Sufficient debridement is a prerequisite for wound coverage, including radical excision of non‐viable tissues and those of doubtful viability. The infection of granulation tissue must be avoided as it can result in poor adhesion despite sufficient surgical debridement and other measures during surgery. Bone or tendon must be excised until healthy, bright, bleeding tissue is seen. An abundant blood supply to both the soft tissue and bone is an important factor in the formation of mature IM. In two of our patients, a second more thorough debridement and bone decortication were required for mature IM formation. Niikura et al reported histologically confirmed IM formation, with rich vascularity, slight inflammation, a foreign‐body reaction and fibrosis, in all of their patients, as well as less IM vascularisation in patients with than in those without free flap surgery. This observation provides further evidence of the importance of a rich blood supply in the soft tissues for the formation of a mature IM. However, the amount of time that should be allowed for IM formation is a matter of debate. Although basic research studies suggest 4‐8 weeks, , , due to patient‐related circumstances, the second surgery in the Masquelet technique is often substantially delayed, ranging from 4 to 96 weeks. A clear correlation between IM maturation and a worse outcome for bone reconstruction has yet to be demonstrated. Similar studies of the effect on wound coverage are also lacking. In our method, the PMMA spacer is removed 4‐6 weeks postoperatively, at which time the IM is mature, as shown by the survival of all subsequent skin grafts. Whether a shorter or longer waiting time results in a more mature IM is unknown, but emerging clinical evidence suggests that time alone is not a reliable measure of IM maturity. , The SAS score showed good psychological status in the patients after the repair, and good functional recovery was achieved by combining an IM with skin grafting. Similar to our technique, Shang et al used an artificial dermis combined with autologous split‐thickness skin grafting. A higher SAS score and greater functional recovery were obtained in those patients compared with patients who underwent autologous intermediate‐thickness skin grafting. The LEFS has been widely validated in the orthopaedic literature following hip and knee arthroplasty, as well as other reconstructive procedures. In addition, it boasts a high test‐retest and inter‐rater reliability; it is also more sensitive at detecting change than other tests of functional outcome, such as the SF‐36. Current literatures reported that the LEFS after lower extremity wound reconstruction ranged from 35 to 65. , Our reported LEFS was 57.83. The patient was mostly satisfied. This is a rare report of a method combining an IM with autologous split‐thickness skin grafting to cover a non‐healing wound with exposed bone or tendon. Liu et al had used this technology in the treatment of diabetic foot ulcer (DFU) and achieved good results. They provided preliminary information on IM formation followed by PMMA implantation in the management of DFUs when revascularization is not feasible. Nonetheless, free skin flaps remain the method of choice for these traumatic wounds, and the indications for our approach should be carefully considered based on the following: First, the patient's general condition is poor, and a long operation to achieve free skin flap transplantation would not be tolerated. Second, the condition of the flap donor site is too poor (infected or scarred) for a flap to be provided. Third, in addition to multiple injuries, the patient has diseases of other systems that delay wound treatment; in such cases, bone cement can be used to temporarily cover the wound. Fourth, the arterial condition of the flap graft area is a contraindication for the procedure. Likewise, this technique may not be applicable in weight‐bearing areas of the foot and in locations with high joint mobility. A limitation of our study was the small sample size. Thus, although the clinical outcome of our patients was good, the results remain to be confirmed in a larger series with a control group (visual analogue scale (VAS) or dressing group). In addition, our study was conducted at a single centre. Prospective multi‐centre investigations are needed to confirm the adaptability of our approach. The signalling pathways and mechanisms involved in IM maturation will be explored further via in vivo and vitro experiments.
CONCLUSION This study demonstrated a good‐to‐excellent clinical outcome and low morbidity rate using a novel approach for the management of non‐healing wounds involving the lower extremities, especially in cases in which bony structures were exposed in the elderly. The method is based on the use of an IM membrane and skin grafting.
This work is supported by a grant from Wenzhou Science and Technology Bureau Foundation (Grant No. Y20210046).
The authors declare no conflicts of interest.
|
Medical Residents, the Group and the Formation of Professional Identity During the COVID-19 Pandemic | eb87d865-eade-450e-9b12-39d57686857d | 10032050 | Gynaecology[mh] | Due to the growing complexity of health care, physician training has been modified to include the skills necessary for new and constantly changing scenarios. The coronavirus disease 2019 (COVID-19) pandemic represents one of the most important health crises in recent human history. In medical education, these changes have forced adaptations in curriculum design regarding new skills, interpersonal relationships, teaching strategies, and the need for community interventions. For physicians in training, such as medical residents, the pandemic has also led to direct changes in work and leisure routines, social interaction, and their own health care. In addition to the inevitable stress that arises from training in a medical specialty, these young physicians are overwhelmed with concerns about how to address the health care of the population and how to deal with the dangers of contagion, both for themselves and those close to them. In other words, in addition to resident burnout, which has already reached epidemic levels in recent years, the pandemic has added a further destabilizing factor. Recently, professional identity formation has been studied as a crucial element in the long trajectory of medical education. Thus, becoming a doctor in such a disruptive and uncertain context may deserve greater attention by researchers committed to medical education. This study is part of a broader, qualitative research project that seeks to analyze the formation of professional identity among gynecology-obstetrics residents in a public hospital in Brazil. Medical residents and preceptors in this program were interviewed, and their interviews were recorded, transcribed, and analyzed using NVivo software (QSR International, Doncaster, Australia). This study was approved by the institutional research ethics committee (protocol no. CAAE 27172919.6.0000.5327). After some initial interviews, it was clear that the pandemic had definitely affected this process. Following its spread from China, the COVID-19 pandemic hit Brazil in late February 2020, with the first case confirmed on February 26, and the first death recorded on March 17. Although there were enormous contrasts in the epidemiological characteristics of the disease's progress, all regions of the country were severely affected, and there were significant changes in the functioning of healthcare networks. In view of this situation, the National Medical Residency Commission (CNRM, an organ linked to the Ministry of Education that is responsible for regulating medical residency) issued a technical note on May 2020 featuring recommendations regarding the development of residence program activities during the COVID-19 pandemic. In general, this document has guided the medical residency commissions of each institution and the State Medical Residency Commissions (CEREMs) about how to make residency activities more flexible to minimize the harmful effects to physicians during the specialization process. At the same time, it called on medical residents of all specialties to actively engage in health care activities aimed at the pandemic in their cities. In the same vein, the Brazilian Federation of Obstetrics and Gynecology Associations (FEBRASGO), a scientific entity that represents Brazilian gynecologists and obstetricians and is involved in the training of specialists in the area, issued its own recommendations seeking, among other points, to minimize the loss of surgical skills due to the pandemic. The Hospital de Clínicas de Porto Alegre, a university hospital, is one of the largest centers for training medical specialists in southern Brazil. The institution had been preparing for the pandemic since January 2020, establishing contingency plans with staggered restrictions for assistance activities, including specific flows for each stage according to criteria that considered the number of occupied hospital beds and the spread of the disease in the state of Rio Grande do Sul and the city of Porto Alegre, where this study was conducted. The first officially registered case of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection at the institution occurred on March 11, 2020, and the first death, 2 weeks later. Since residence programs generally begin in early March, the new residents arrived at the height of all this activity. Although the official reception was still festive, including hundreds of photos shared on social media, the hospital moved to its first contingency level on March 15, limiting a number of assistance activities and the circulation of people through the hospital, which affected the patients and the medical teams. As the pandemic progressed, restrictions were implemented in residence program teaching activities and, in conformity to local health regulations, bars, restaurants, and other public places were closed to reduce crowds. This exceptional scenario has had repercussions even on specialties that are not traditionally involved in epidemiological problems. Obstetrics and gynecology, for example, is essentially a medical specialty in women's health. The competency matrix, which serves as a guide for the almost 300 residency programs in obstetrics and gynecology in Brazil, was defined through recent CNRM legislation. This detailed document divides the competences to be achieved into several axes of care according to each year of the residency. Although it does not establish a minimum number of expected procedures, there is a clear indication that the resident must master certain techniques, which presupposes having acquired such competence through exhaustive training. As in any other surgical specialty, it requires, in addition to knowledge and attitudes, the development of a series of motor skills, which residents obtain through continuous exposure to a significant and varied number of situations in which they can exercise their clinical reasoning in an appropriate, independent, and safe way for their patient and themselves. In this complex trajectory towards specialized medical practice, the interrelationship of peers plays a crucial role. Although the formation of a physician's professional identity involves an eminently individual path, it always occurs alongside others who are undergoing the same situation. For newcomers, being accepted into a peer community that has a structured professional identity can pose additional stress during the transition from student to doctor. This is true among the residents and the health team members they relate to professionally. In other words, becoming a doctor is also directly linked to the process of appearing like a doctor to other people. Thus, forming a group of residents has been a hallmark of medical residency since its beginning. In other words, the formation of professional identity is a two-way street: from the inside out (the doctor must feel like a doctor) and from the outside in (other people must see him or her as a doctor). Although the millennials' way of learning has led to changes in teaching and medical residency in recent years, some traits, such as group support, have remained equally important despite generational differences and increasing technological sophistication. Regardless of preparation, which can begin prior to entering medical school, the transition from being a student to a resident is still a huge challenge in any specialty. Despite previous familiarity with some of the activities of medical school, being registered with state agencies and “having the seal”, that is, formal authorization to practice medicine, puts professional activity in a different perspective. Likewise, coexistence and relationships with new colleagues—some completely unknown and others who were former competitors for a place in the program—is an important element in this equation. Interaction between colleagues in outpatient clinics, in the operating room, on duty and in seminars is not only inevitable, but it can also help residents bear the moments of stress and emotional overload that occur during the program. Often, it is this type of connection that allows young doctors to satisfactorily reach the end of their journey toward obtaining the title of specialist after finishing the residency program. Following the rapid expansion of information, communication, and technology resources, there seems to be no doubt that this generation has a different learning style than its predecessors. Nevertheless, despite such autonomous learning and skill development, interpersonal relationships and collaborative work among peers remain among the most important pillars in resident training, although different programs do not always recognize them in the same way. The medical field in Brazil reflects certain particularities observed in the country as a whole. Deciphering Brazilian culture has, in itself, been a constant challenge for generations of anthropologists. However, despite great regional diversity, we can say that Brazilian culture has a more permissive tendency than many other countries, both in romantic relationships and in friendships. Among other characteristics, it is a culture marked by physical proximity; for example, greetings with hugs and kisses are not unusual. This greater permissiveness in interpersonal relations is also reflected in the work environment. In medical residencies, for example, new friends are made quickly through numerous social activities during the program, although mainly outside the formal structure. The welcoming rituals for new residents, reception parties, and other informal gatherings that occur in the first weeks of the program are good examples of this. Such events, especially among residents of the same year, become habitual and strengthen friendships and the feeling of belonging to the larger group. These feelings continue even after the residents go their separate ways as trained professionals. The testimonies of two final-year residents illustrate this dimension well: “Towards the end of the first year, we start working together, with two first-year residents and one second-year resident. But there was always this thing about ‘after our shift, let's go out and get something to eat, after our shift let's go out and...’” “To deal with stress, I talk to my fellow residents. We have a very good group, my colleagues from the same year. We're pretty united. We try to help each other. Just yesterday, two colleagues were going through some situations and we [said] ‘Let's go have some coffee together, let's go have lunch together’, we help each other a lot. And this is in addition to family support” (these and the following quotes have been translated from Portuguese). During a medical residency program, any activity can have an educational role, even informal ones and those involving interpersonal and group relationships. It is also in these environments that information about jobs, courses, publications, and articles of interest are exchanged. As Bonet notes in his study on the training of family physicians in Argentina and Brazil, it is interesting to observe the different environments in which professional socialization takes place. It is in these spaces that, more or less explicitly, the transmission of values and performance standards occurs. Thus, the group of residents becomes the space in which they share their personal stories and reveal their personalities in a more open way, without the risk of being judged by preceptors. In fact, it is when residents share their problems with the group, such as difficulties with patients or conflicts within the staff, that they receive collective support. As a second-year resident reports: “I was having a very stressful month that was turned around by my great team. I had an episode that I guess should be called burnout, in which I found myself yelling at a patient. I understood I was in a bad environment. But my colleagues took me aside and said: ‘What's going on? Let's sit down and talk.’ And then I was able to get myself together, and the month ended very well.” Obviously, the group dynamics depend essentially on how these relationships are established and maintained over time. On one hand, there can be a spirit of camaraderie and cooperation, but on the other, overexposure can be dangerous because it reveals weaknesses, faults, or other behaviors that are considered out of step with the group's principles. These discrepancies, if striking, can even compromise a resident's participation in the group. This is because, despite theoretically enjoying independence, more flexible limits, and their own set of values, groups of residents assume a position of relative submission to the larger structure of the residency program. In other words, there is a permanent tension between the central core of the program and the groups of residents who, being the weak end, are “shaped” by the more consolidated and, therefore, more powerful group. Two residents' reflections about how a colleague abandoned the program the previous year are very enlightening about this type of thinking: “The eight of us who entered [the program] are progressing, for better or worse, by leaps and bounds, but, when necessary, we get together and hammer things out. I think the group we have now is very good, we help each other. Still, there are, of course, differences, little intrigues, but I think that, in general, it is a supportive group. There was a resident who requested a transfer to another state, and, recently, one of our oldest residents gave up, taking a test for another specialty. And there was another one who gave up at the beginning of the year and is doing another clinical specialty. These people did not feel at home in our program.” “Two colleagues ended up quitting the program. I think it was because of a number of things. I think the first one gave up because she didn't really fit in; she gave up in July. I think it's because she didn't really like what she was doing. But the second one, I think, was because of built-up pressure. She had a sick relative and the pressure her colleagues–her teammates–put on her was also great.” Even in this environment, interesting changes have occurred more recently regarding group formation. The growing and widespread use of technology has proven inexorable in all dimensions of our lives. Interpersonal and intergroup communication should be highlighted, especially during the pandemic, which has required physical distancing as a way to reduce disease transmission. Communication through platforms such as Facebook, Instagram, and applications—such as WhatsApp and Telegram—have become very popular, especially among young people. With these tools, unlimited instant text and voice messages, images, and videos can be exchanged virtually for free. Thus, other forms of interpersonal communication had to be quickly created to ensure a sense of interactivity and belonging, which is essential for medical residency. “They [only] had 2 weeks of residency! Many people can't make it to meetings. For example, what brought my group together was this: scheduling a happy hour to talk trash about the professors together, call each outer out, to get together, we did that a whole lot. So, now, we can very easily resolve our problems, trade shifts, and admonish each other. [The new residents] often have to go before the full group to ask to trade shifts.” In this context, there was a rush toward online communication tools and virtual meetings. Although most of these applications have been used routinely for several years, the lockdown has intensified their use. What was formerly resolved in person began happening online—even the most common questions. As one resident said: “ WhatsApp saved us! ” Nevertheless, she pointed out that “ it's not the same thing, but you can still feel supported. ” The changes in these relationships seem to indicate adjustments in both form and content. They were creative adaptations implemented very quickly during an uncertain situation and an urgent need for interaction. Thus, a quick solution for a communication problem has now assumed the characteristics of professional identity formation. Assessing the real impact of these changes will require, in addition to a longer follow-up time, a more comprehensive and careful perspective. |
Histopathology and SARS-CoV-2 Cellular Localization in Eye Tissues of COVID-19 Autopsies | 44b81d33-136a-4ba2-a5b8-757f5549a46d | 10032059 | Forensic Medicine[mh] | Postmortem Human Eye Procurement and Processing Postmortem human eyes (E1 to E25) from 25 patients with confirmed SARS-CoV-2 infection were obtained at autopsy performed at the NIH Clinical Center following consent of the legal next of kin as previously described. The study adhered to the tenets of the Declaration of Helsinki. The cases had a range of illness durations before death and were categorized as early ( n = 11), mid ( n = 7), or late ( n = 7) by illness day (D) at the time of death being D14 or earlier, D15 to D30, or D31 or later, respectively ( and ). Previously, one eye from E1 to E25 was freshly dissected, and tissues were preserved in RNA later for subsequent quantification of SARS-CoV-2 RNA by ddPCR. For the study described here, the contralateral eye from E1 to E21 was fixed in 10% neutral buffered formalin for a minimum of 72 hours with the globe intact, opened for macroscopic examination, and then embedded in paraffin for histopathologic evaluation. For cellular localization of SARS-CoV-2 RNA, one eye from E22 to E25 was dissected to isolate cornea, retina, choroid/sclera, and optic nerve. Each tissue type was divided into two pieces—one for ISH and one for SARS-CoV-2 ddPCR. The piece prepared for ISH was fixed for 24 hours in neutral buffered formalin because prolonged formalin fixation interfered with RNA preservation. This fixed piece was then transferred to 70% ethanol for a minimum of 2 days before impregnation with paraffin. The piece prepared for ddPCR was placed in RNA later media for nucleic acid preservation and subsequent SARS-CoV-2 RNA quantification. SARS-CoV-2 RNA quantification was used to guide selection of tissues for analysis by ISH. Histopathology Each eye was cut horizontally along the pupillary-optic nerve head (P-O) axis. For other lesions outside the P-O section, a segment through the lesion was also obtained for histologic analysis. After macroscopic examination of the open eye under a stereo dissecting microscope, all P-O sections and segments were then processed for routine paraffin embedding and sectioning. Consecutive sections of 4-μm thickness were sliced from each paraffin block and stained with hematoxylin and eosin or periodic acid–Schiff reagent. Immunohistochemistry CD61, a marker of platelet aggregation and thrombosis, and cytomegalovirus (CMV) expressions were evaluated on formalin-fixed, paraffin-embedded (FFPE) slides using immunohistochemistry. Briefly, the presence of CD61 was detected by immunohistochemistry performed on FFPE sections with an anti-CD61 mouse monoclonal antibody (clone 2f2; Roche Diagnostics, Mannheim, Germany) performed on a Roche Discovery Ultra instrument with 3,3′-diaminobenzidine detection per clinical laboratory protocol. The presence of CMV was also detected by a similar method with an anti-CMV antibody (clone CCH2+DDG9; Agilent Dako, Santa Clara, CA). SARS-CoV-2 Spike RNA ISH ISH was performed using manual RNAscope 2.5 HD Reagent Kit-BROWN (catalog number 322350; Advanced Cell Diagnostics, Newark, CA) according to the user manual (322360-USM) and as previously described. Briefly, FFPE slides were deparaffinized with xylene followed by 100% ethanol and then treated with hydrogen peroxide at room temperature for 10 minutes to block endogenous peroxidase activity. After antigen retrieval at 99°C for 15 minutes, the slides were incubated with protease at 40°C for 20 minutes. For each ocular tissue, RNAscope Probe-V-nCoV2019-S (catalog number 848561; Advanced Cell Diagnostics), RNAscope Positive Control Probe-Hs-PPIB (catalog number 313901; Advanced Cell Diagnostics), and RNAscope Negative Control Probe-DapB (catalog number 310043; Advanced Cell Diagnostics) were applied to three sequential slides and incubated at 40°C for 2 hours. After rinsing, the ISH signal was amplified using six amplifiers and incubated with a diaminobenzidine substrate for 10 minutes at room temperature. Slides were then counterstained with 50% hematoxylin, air-dried, and mounted.
Postmortem human eyes (E1 to E25) from 25 patients with confirmed SARS-CoV-2 infection were obtained at autopsy performed at the NIH Clinical Center following consent of the legal next of kin as previously described. The study adhered to the tenets of the Declaration of Helsinki. The cases had a range of illness durations before death and were categorized as early ( n = 11), mid ( n = 7), or late ( n = 7) by illness day (D) at the time of death being D14 or earlier, D15 to D30, or D31 or later, respectively ( and ). Previously, one eye from E1 to E25 was freshly dissected, and tissues were preserved in RNA later for subsequent quantification of SARS-CoV-2 RNA by ddPCR. For the study described here, the contralateral eye from E1 to E21 was fixed in 10% neutral buffered formalin for a minimum of 72 hours with the globe intact, opened for macroscopic examination, and then embedded in paraffin for histopathologic evaluation. For cellular localization of SARS-CoV-2 RNA, one eye from E22 to E25 was dissected to isolate cornea, retina, choroid/sclera, and optic nerve. Each tissue type was divided into two pieces—one for ISH and one for SARS-CoV-2 ddPCR. The piece prepared for ISH was fixed for 24 hours in neutral buffered formalin because prolonged formalin fixation interfered with RNA preservation. This fixed piece was then transferred to 70% ethanol for a minimum of 2 days before impregnation with paraffin. The piece prepared for ddPCR was placed in RNA later media for nucleic acid preservation and subsequent SARS-CoV-2 RNA quantification. SARS-CoV-2 RNA quantification was used to guide selection of tissues for analysis by ISH.
Each eye was cut horizontally along the pupillary-optic nerve head (P-O) axis. For other lesions outside the P-O section, a segment through the lesion was also obtained for histologic analysis. After macroscopic examination of the open eye under a stereo dissecting microscope, all P-O sections and segments were then processed for routine paraffin embedding and sectioning. Consecutive sections of 4-μm thickness were sliced from each paraffin block and stained with hematoxylin and eosin or periodic acid–Schiff reagent.
CD61, a marker of platelet aggregation and thrombosis, and cytomegalovirus (CMV) expressions were evaluated on formalin-fixed, paraffin-embedded (FFPE) slides using immunohistochemistry. Briefly, the presence of CD61 was detected by immunohistochemistry performed on FFPE sections with an anti-CD61 mouse monoclonal antibody (clone 2f2; Roche Diagnostics, Mannheim, Germany) performed on a Roche Discovery Ultra instrument with 3,3′-diaminobenzidine detection per clinical laboratory protocol. The presence of CMV was also detected by a similar method with an anti-CMV antibody (clone CCH2+DDG9; Agilent Dako, Santa Clara, CA).
ISH was performed using manual RNAscope 2.5 HD Reagent Kit-BROWN (catalog number 322350; Advanced Cell Diagnostics, Newark, CA) according to the user manual (322360-USM) and as previously described. Briefly, FFPE slides were deparaffinized with xylene followed by 100% ethanol and then treated with hydrogen peroxide at room temperature for 10 minutes to block endogenous peroxidase activity. After antigen retrieval at 99°C for 15 minutes, the slides were incubated with protease at 40°C for 20 minutes. For each ocular tissue, RNAscope Probe-V-nCoV2019-S (catalog number 848561; Advanced Cell Diagnostics), RNAscope Positive Control Probe-Hs-PPIB (catalog number 313901; Advanced Cell Diagnostics), and RNAscope Negative Control Probe-DapB (catalog number 310043; Advanced Cell Diagnostics) were applied to three sequential slides and incubated at 40°C for 2 hours. After rinsing, the ISH signal was amplified using six amplifiers and incubated with a diaminobenzidine substrate for 10 minutes at room temperature. Slides were then counterstained with 50% hematoxylin, air-dried, and mounted.
Autopsy Cohort Overview Postmortem ocular specimens were collected from 25 cases (E1 to E25) with confirmed COVID-19 between April 2020 and December 2021. E24 was the only case that received a COVID-19 vaccine. Eight of the 25 cases (32%) were female, and the mean age at the time of death was 58 years (range, 6 to 91 years). Twenty-four of the 25 (96%) cases had at least one comorbidity, and 15 (60%) cases had three or more comorbidities, with hypertension [15 (60%)] and obesity [10 (40%)] being the most common. The median interval from symptom onset to death was 18 days (range, 4 to 204 days), and the median interval between death and autopsy was 23 hours (range, 10 to 106 hours). Systemic steroids and systemic anticoagulation were commonly administered to treat COVID-19. Case-by-case demographic and clinical data can be found in . Macroscopic Retinal Lesions Pathologic findings (E1 to E21) are described in alongside previously reported SARS-CoV-2 RNA ddPCR results from contralateral ocular tissues of the same case. Six of the 21 eyes examined showed potential COVID-19–associated macroscopic lesions in the retina. Specifically, one eye (E18) had vascular sheathing similar to vasculitis in multiple branch retinal vessels ( A). Two eyes (E2 and E11) showed focal whitish cotton-wool spots in the posterior pole ( B). Two eyes had focal retinal hemorrhages in the posterior pole (E8 and E17) and midperipheral retina (E8) ( C). One eye (E19) showed several patches of subretinal whitish infiltrates ( D). Microscopic examination showed the whitish infiltrates being severe retinal edema and disorganization, accumulation of serous fluids in the retinal outer plexiform layer, as well as debris and strands in the vitreous ( D, inset). In terms of lesions not associated with COVID-19, four eyes (E3, E7, E13, and E17) were pseudophakic with residual cortex material in the peripheral capsule; E3 also showed two old, large hypopigmented chorioretinal scars with focal hyperpigmentation in the peripheral retina. A large, whitish, fragile soft mass (myelin expulsion due to artifact) attached to the optic nerve head was noted in case E14. All the other 11 eyes (E1, E4-6, E9, E10, E12, E15, E16, E20, and E21) were grossly unremarkable. Microscopic Retinal and Choroidal Lesions Retinal lesions potentially associated with SARS-CoV-2 infection are summarized in and , E–L. They include small neovascularization ( E), retinal sclerotic vessels ( F), retinal vascular occlusion ( G) with fibrin/thrombi in retinal vessels ( H), and vitreous hemorrhages. Eleven (E1, E4, E5, E7, E11, E13-15, E17, E18, and E21) of the 21 eyes (52.4%) contained cytoid bodies (the histopathologic constituent of clinical cotton-wool spots) in the nerve fiber layer of the neuroretina ( , F, H and I) and/or on the optic nerve head ( J). Focal loss of photoreceptor cells ( , E, I–L) was noted in seven eyes (33.3%) (E6, E8, E11, E16, E17, E18, and E19). Fourteen eyes (66.7%) (E1, E2, E4, E6-8, E12 to 15, and E18 to 21) showed retinal/choroidal vascular dilation, congestion, and tortuosity ( , F–H, J, and K). In addition, serous retinal edema in the inner and outer nuclear layers was noted in six eyes (28.6%, E3, E8, E12, E13, E17, and E19) ( , D and K), whereas optic nerve head edema was seen in five eyes (23.8%) (E2, E4, E9, E13, and E21) ( J). Fluid accumulation in the outer plexiform layer was also noted in two eyes (9.5%) (E8 and E11 in the macula) ( K). Only one eye (E18) had viral inclusions, which stained negative for CMV, in a few retinal ganglion cells ( L, inset). Except for one eye (E3) showing minimal to mild lymphocyte infiltration on the optic nerve and one eye with minimal choroidal monocytic inflammation (E16), no inflammatory cells were noted in the other ocular tissues or the other eyes. Importantly, cytoid bodies, retinal edema, or vascular congestions/hemorrhage were observed either alone or in combination in 20 of the 21 eyes (95.2%) and were associated with ddPCR positivity in the contralateral eye in 14 cases ( ). Positive CD61 staining of platelet thrombi ( ), sometimes associated with fibrin and mononuclear cells, was observed in the retinal vascular wall in five eyes (E7, E13, E14, E17, and E18), all with positive ddPCR results in the contralateral eye and the presence of cytoid bodies in the nerve fiber layer of the retina. One eye (E18) also showed retinal vascular occlusion ( G) and neovascularization ( E). No CMV positivity was detected in any cases. Cellular Localization of SARS-CoV-2 Spike Gene RNA by ISH ISH for SARS-CoV-2 spike gene RNA was performed on ocular tissues from E22 to E25 ( and ). SARS-CoV-2 RNA was detected mostly in the outer nuclear layer (photoreceptor cells) of the retina in case E22 ( A). Some SARS-CoV-2 RNA was also detected in the inner nuclear layer and in the ganglion cells of the E22 retina. Examination of E23 demonstrated the presence of SARS-CoV-2 within retinal inner and outer nuclear cells and a few ganglion cells ( B). No signal was consistently observed using a standard negative control probe for the bacterial gene DapB to demonstrate specificity of the assay. As examples, images of retinal tissues (from E22 and E23) treated with the negative control probe are shown in , C and D, respectively. A focus of SARS-CoV-2 RNA was detected in the scleral fibroblasts of E24 ( A). Additionally, analysis of E25 demonstrated the presence of SARS-CoV-2 RNA within oligodendrocytes of the optic nerve ( B) as well as the retinal cells ( C) and corneal epithelium ( D).
Postmortem ocular specimens were collected from 25 cases (E1 to E25) with confirmed COVID-19 between April 2020 and December 2021. E24 was the only case that received a COVID-19 vaccine. Eight of the 25 cases (32%) were female, and the mean age at the time of death was 58 years (range, 6 to 91 years). Twenty-four of the 25 (96%) cases had at least one comorbidity, and 15 (60%) cases had three or more comorbidities, with hypertension [15 (60%)] and obesity [10 (40%)] being the most common. The median interval from symptom onset to death was 18 days (range, 4 to 204 days), and the median interval between death and autopsy was 23 hours (range, 10 to 106 hours). Systemic steroids and systemic anticoagulation were commonly administered to treat COVID-19. Case-by-case demographic and clinical data can be found in .
Pathologic findings (E1 to E21) are described in alongside previously reported SARS-CoV-2 RNA ddPCR results from contralateral ocular tissues of the same case. Six of the 21 eyes examined showed potential COVID-19–associated macroscopic lesions in the retina. Specifically, one eye (E18) had vascular sheathing similar to vasculitis in multiple branch retinal vessels ( A). Two eyes (E2 and E11) showed focal whitish cotton-wool spots in the posterior pole ( B). Two eyes had focal retinal hemorrhages in the posterior pole (E8 and E17) and midperipheral retina (E8) ( C). One eye (E19) showed several patches of subretinal whitish infiltrates ( D). Microscopic examination showed the whitish infiltrates being severe retinal edema and disorganization, accumulation of serous fluids in the retinal outer plexiform layer, as well as debris and strands in the vitreous ( D, inset). In terms of lesions not associated with COVID-19, four eyes (E3, E7, E13, and E17) were pseudophakic with residual cortex material in the peripheral capsule; E3 also showed two old, large hypopigmented chorioretinal scars with focal hyperpigmentation in the peripheral retina. A large, whitish, fragile soft mass (myelin expulsion due to artifact) attached to the optic nerve head was noted in case E14. All the other 11 eyes (E1, E4-6, E9, E10, E12, E15, E16, E20, and E21) were grossly unremarkable.
Retinal lesions potentially associated with SARS-CoV-2 infection are summarized in and , E–L. They include small neovascularization ( E), retinal sclerotic vessels ( F), retinal vascular occlusion ( G) with fibrin/thrombi in retinal vessels ( H), and vitreous hemorrhages. Eleven (E1, E4, E5, E7, E11, E13-15, E17, E18, and E21) of the 21 eyes (52.4%) contained cytoid bodies (the histopathologic constituent of clinical cotton-wool spots) in the nerve fiber layer of the neuroretina ( , F, H and I) and/or on the optic nerve head ( J). Focal loss of photoreceptor cells ( , E, I–L) was noted in seven eyes (33.3%) (E6, E8, E11, E16, E17, E18, and E19). Fourteen eyes (66.7%) (E1, E2, E4, E6-8, E12 to 15, and E18 to 21) showed retinal/choroidal vascular dilation, congestion, and tortuosity ( , F–H, J, and K). In addition, serous retinal edema in the inner and outer nuclear layers was noted in six eyes (28.6%, E3, E8, E12, E13, E17, and E19) ( , D and K), whereas optic nerve head edema was seen in five eyes (23.8%) (E2, E4, E9, E13, and E21) ( J). Fluid accumulation in the outer plexiform layer was also noted in two eyes (9.5%) (E8 and E11 in the macula) ( K). Only one eye (E18) had viral inclusions, which stained negative for CMV, in a few retinal ganglion cells ( L, inset). Except for one eye (E3) showing minimal to mild lymphocyte infiltration on the optic nerve and one eye with minimal choroidal monocytic inflammation (E16), no inflammatory cells were noted in the other ocular tissues or the other eyes. Importantly, cytoid bodies, retinal edema, or vascular congestions/hemorrhage were observed either alone or in combination in 20 of the 21 eyes (95.2%) and were associated with ddPCR positivity in the contralateral eye in 14 cases ( ). Positive CD61 staining of platelet thrombi ( ), sometimes associated with fibrin and mononuclear cells, was observed in the retinal vascular wall in five eyes (E7, E13, E14, E17, and E18), all with positive ddPCR results in the contralateral eye and the presence of cytoid bodies in the nerve fiber layer of the retina. One eye (E18) also showed retinal vascular occlusion ( G) and neovascularization ( E). No CMV positivity was detected in any cases.
ISH for SARS-CoV-2 spike gene RNA was performed on ocular tissues from E22 to E25 ( and ). SARS-CoV-2 RNA was detected mostly in the outer nuclear layer (photoreceptor cells) of the retina in case E22 ( A). Some SARS-CoV-2 RNA was also detected in the inner nuclear layer and in the ganglion cells of the E22 retina. Examination of E23 demonstrated the presence of SARS-CoV-2 within retinal inner and outer nuclear cells and a few ganglion cells ( B). No signal was consistently observed using a standard negative control probe for the bacterial gene DapB to demonstrate specificity of the assay. As examples, images of retinal tissues (from E22 and E23) treated with the negative control probe are shown in , C and D, respectively. A focus of SARS-CoV-2 RNA was detected in the scleral fibroblasts of E24 ( A). Additionally, analysis of E25 demonstrated the presence of SARS-CoV-2 RNA within oligodendrocytes of the optic nerve ( B) as well as the retinal cells ( C) and corneal epithelium ( D).
This study illustrates the histopathologic characterization and molecular localization of SARS-CoV-2 within specific ocular tissues and cells in a cohort of 25 fatal COVID-19 cases. Abnormalities suggestive of indirect viral effect, including cytoid bodies, vascular changes, retinal edema, and minimal inflammation, were observed in all 21 eyes that underwent gross and histopathologic evaluation, irrespective of illness duration before death. Fourteen of 21 eyes (67%) were from cases whose contralateral eyes were positive for SARS-CoV-2 by ddPCR. To the best of our knowledge, this is the first report of localization of SARS-CoV-2 RNA in the retina demonstrated by ISH, including the inner and outer nuclear layers and ganglion cells. These findings are consistent with those of Casagrande et al who previously reported detection of SARS-CoV-2 RNA in retinal biopsies from three of 14 eyes by RT-PCR but differ from Bayyoud et al, who did not detect SARS-CoV-2 RNA in 10 cadaveric eyes from five deceased COVID-19 donors. In this study, three of four eyes evaluated by ISH showed localization of SARS-CoV-2 to the retina. Importantly, these four eyes were dissected at autopsy and rapidly fixed to preserve RNA integrity, which contrasts with the standard approach for immersing the globe in toto , risking overfixation. Widespread SARS-CoV-2 dissemination in the cohort suggests hematogenous seeding of the eye with disruption of the blood-retinal barrier. However, localization of SARS-CoV-2 to oligodendrocytes of the optic nerve, focal neuronal loss in the optic nerve of three cases reported here, and the 59% optic nerve SARS-CoV-2 positivity rate that we previously reported raise the possibility of neuronal spread of SARS-CoV-2. There are conflicting reports of expression of angiotensin-converting enzyme 2 (ACE2) and transmembrane protease serine 2 (TMPRSS2) in human ocular tissues perhaps due to different methods for tissue preparation and protein detection. Nonetheless, some reports provide evidence of ACE2 and TMPRSS2 expression in multiple ocular tissues, including optic nerve; retinal neuronal, vascular, and perivascular cells; and visual processing centers of the brain, supporting the possibility of both neuronal and hematogenous infection routes. , Given the expression of ACE2 and TMPRSS2 reported in cornea and conjunctival tissue, concern exists that the ocular surface can serve as a route for SARS-CoV-2 entry. Casagrande et al reported 55% SARS-CoV-2 RT-PCR positivity of corneal buttons in 11 deceased patients with COVID-19. Sawant et al reported 13% SARS-CoV-2 RT-PCR positivity of ocular specimens collected from 33 donors eliminated from corneal donation due to suspected or confirmed SARS-CoV-2 infection, with the highest rate of detection in vitreous and posterior corneal tissue. The study described here is the first to demonstrate localization of SARS-CoV-2 to the ocular surface using ISH, supporting the RT-PCR-based findings from previous studies ( , A and D). Non-CMV viral inclusion bodies, suggestive of active viral replication, were observed in only one eye (E18, with longer disease duration) within a few retinal ganglion cells ( L). However, the contralateral eye from this patient was SARS-CoV-2 positive in the aqueous not retina by ddPCR. It is not uncommon to present discrepancies between the two eyes in ocular diseases, including infection. CMV retinitis tends to be unilateral in presentation. In support of this finding, SARS-CoV-2 was observed in the retinal ganglion cell layer by ISH ( B). To the best of our knowledge, viral inclusions have not previously been reported in other histologic studies on postmortem ocular tissues following COVID-19 infection, although presumed SARS-CoV-2 viral particles have been reported in the retina of three patients with COVID-19 using transmission electron microscopy. The presence of viral inclusion bodies within the retina (ganglion cells) is classically associated with herpetic viral retinitis. Notably, in this case, CMV test results were negative, and no additional inflammatory infiltrate and retinal necrosis consistent with necrotizing retinitis were observed. The lack of frank inflammatory cellular infiltration in the 21 eyes that underwent histologic evaluation is notable, given that 67% of contralateral eyes from these cases were ddPCR positive ( ). The only exceptions were that E16 showed mild monocytic choroidal inflammation and E3 showed minimal inflammation in the optic nerve, with the contralateral eye being SARS-CoV-2 ddPCR positive in both cases. The most common ocular histopathologic findings in the cohort were nerve fiber layer cytoid bodies; retinal vascular abnormalities, including dilation, congestion, tortuosity, and/or occlusion; and optic disc edema. These findings are consistent with microvascular insults in both early and late COVID-19 cases. However, all but one patient with cytoid bodies and retinal vascular dilation had concurrent cardiovascular risk factors, which might at least in part account for these findings. Clinical reports and autopsy studies indicate that multisystem microvascular injury and vasculopathy contribute to COVID-19 pathogenesis, including in the central nervous system, where acute hypoxic injury, microinfarcts, microhemorrhages, and hemorrhages have been observed. , These observations and the lack of inflammatory cells in the eyes of patients with COVID-19 suggest that ocular microscopic changes are likely due to indirect rather than direct viral cytopathic effects. Ocular microangiopathic changes, including the finding of platelet thrombi ( ) in capillaries identified by CD61 expression in five cases, are supportive of thromboembolic events reported with SARS-CoV-2 infection. Specifically, one eye had CD61 positivity (E18) with histopathologic evidence of retinal vascular occlusion. CD61 has been implicated in having a central role in platelet activation and aggregation, and platelet microthrombi were described in early SARS-CoV-2 autopsy series. This could lead to microvascular ischemia, resulting in cytoid bodies in the retinal nerve fiber layers and optic nerve head. Reinhold et al reported in their cohort of 10 autopsy eyes congestion of the choriocapillaris and accumulation of multiple small thrombi within the choriocapillaris in eight eyes from four patients. Retinal microvascular changes have also been observed in studies of patients with COVID-19, including microhemorrhages, retinal vascular tortuosity, and cotton-wool spots, as well as quantitative alternations in the retinal vasculature. In addition to retinal microvascular changes, reports of retinal vein occlusion associated with COVID-19 have been described, including in younger patients without known risk factors. The hypercoagulable state associated with COVID-19 has been implicated in the pathogenesis of these retinal vein occlusions. , , , , Lastly, cases of optic neuritis have been reported following COVID-19 infection. In the study described herein, SARS-CoV-2 RNA was detected in the optic nerve oligodendrocytes in one eye ( B). Although the exact mechanisms remain unclear, antigenic mimicry and breakdown of the blood brain barrier have been postulated. , , Whether direct viral invasion can cause or potentiate these mechanisms is not known. As referenced above, this study has limitations. Given that ocular history and clinical examinations from the donors are not available, it is not known whether any of these observed changes preceded SARS-CoV-2 infection. The diverse comorbidities of the donor cohort can contribute to the observed ocular pathology; consequently, it is difficult to draw clear conclusions about whether the pathology was solely and/or mainly caused by COVID-19 or not. In particular, many of the donors in the study cohort had cardiovascular comorbidities, which could contribute risk factors for more severe COVID-19 and ocular pathology. Despite these limitations, the study shows important new insights into SARS-CoV-2 ocular pathogenesis. Specifically, this is the first report to definitively localize SARS-CoV-2 to the retinal inner and outer nuclear cells, retinal ganglion cells, and ocular surface by ISH, validating previous studies that have exclusively used PCR-based methods. These observations highlight the need to better elucidate mechanisms of SARS-CoV-2 infection and persistence in the eye, associated direct and indirect pathogenesis, and long-term ocular sequelae among COVID-19 survivors.
|
Occupational risks to pregnant obstetrics and gynaecology trainees and physicians: Is it time to think about this? | 85fd8e88-1176-40a9-a0af-f9c63992908e | 10032316 | Gynaecology[mh] | The proportion of women in the workforce has been steadily increasing worldwide. Women now constitute approximately 75% of the global health care workforce and almost 90% in nursing and midwifery professions . In India, females form 38% of all health care workers (HCW) and about 16.8% of allopathic doctors. Roughly 1 in every 3 HCW is a female . The present times have witnessed a dramatic gender shift in the speciality of obstetrics and gynaecology. Women, now comprise a significant proportion of practicing obstetrics and gynaecology specialists all over the world . In 2018, more than 80% of resident doctors and nearly 60% of physicians in the speciality were female, far exceeding any other surgical speciality . This is in stark contrast to 2012, when women comprised >50% of Fellows and Junior Fellows in the American College of Obstetricians and Gynecologists . In India, this figure is much more than 90%, as the patients in this world favour female physicians for their gynaecological issues. The male trainees often feel bias because of patients preferring female physicians . The majority of resident doctors and a significant proportion of physicians in Obstetrics and Gynaecology are in the reproductive age group. They are or will become pregnant at some point in their training program or career. Pregnant HCWs are faced with numerous challenges, as they need to balance their health and the health of their unborn child along with that of their patients. Proper performance of their duties may at times constitute a risk to their own health. Although pregnant women are not more susceptible to most diseases than their non-pregnant counterparts, the consequences of even a mild infection can be far-reaching. Rubella and chickenpox, although most cases are self-resolving, can lead to abortions and congenital abnormalities in the offspring. It is quite common for a health care worker to feel torn between duties towards her patients and co-workers and her responsibilities towards her family and her unborn foetus. A pregnant trainee or even a consultant physician in obstetrics and gynaecology faces unique occupational challenges and hazards. Apart from the physically taxing nature of work in the labour room, where each normal delivery needs continuous monitoring and vigilance for at least 6–8 hours, a number of other occupational risks are unique to the speciality, which probably at the time of pregnancy become matters of concern. To the best of my knowledge, there is no comprehensive review that identifies occupational risks to pregnant obstetricians and gynaecologists. This review focuses on all work-related exposure risks, such as risks of infectious diseases, radiation, stress, violence against doctors, and even peer support, or lack of support, that can have deleterious effects on the health of pregnant physicians and the health of their unborn foetuses. The recent literature related to pregnant health care workers and occupational risks was searched, from various governmental agencies including the World Health Organisation (WHO), Centers for Disease Control and Prevention (CDC), Scientific Advisory Group for Emergencies (SAGE), Occupational Safety and Health Administration (OSHA), and English peer-reviewed journals from databases such as PubMed, Scopus, Google Scholar, EMBASE, and others. The literature regarding workplace regulations for pregnant or lactating health care staff was also reviewed. The search terms used were: ‘pregnant health care worker’ AND ‘occupational risks’; OR ‘radiation exposure’; OR ‘violence against doctors’; OR ‘infectious diseases’ OR ‘physician burn out or stress; Or ‘anaesthetic gases’. The articles referring to obstetrics and gynaecology trainees and physicians were studied in detail to work out all the occupational risks faced by them during pregnancy. Radiation exposure Radiation exposure in early pregnancy is very well known to be associated with teratogenic effects. Although the risk is usually overestimated, the consequences can be significant in cumulative doses. Threshold radiation effects (deterministic effects) occur over a dose threshold and result in cellular injury. Stochastic effects of radiation are incremental, appearing in a dose-response function without a threshold, and are thought to be the primary mechanism of increased risk of cancers . Various agencies have given thresholds that should be taken care of once pregnancy is confirmed, to minimize the effects on the foetus, mainly during organogenesis. The International Commission on Radiological Protection recommends that after a worker declares her pregnancy, the occupational radiation dose should not exceed one mSv during the remainder of the pregnancy. The National Council on Radiation Protection and Measurements, in the United States, recommends a radiation dose limit of 0.5 mSv per month once pregnancy is confirmed to ensure a low exposure during susceptible periods of gestation . The US Environmental Protection Agency recommends a limit of 5 mSv for the entire gestational period . Although only very few procedures in obstetrics and gynaecology use ionizing radiation, hysterosalpingography (HSG) is a standard procedure used to evaluate the patency of fallopian tubes in women presenting with subfertility. The HSG procedure requires the radiologist or preferably a gynaecologist (who is not trained to handle ionizing radiation) to hold the cannula and inject the contrast medium into the patient's cervix while she is being irradiated. The supporting personnel also remain close to the patient. Though a lead apron is worn, frequent or multiple procedures can lead to significant exposure to ionizing radiation, which can be worrisome, especially in the first trimester due to teratogenic effects . A few studies have demonstrated that the dose to the extremities may also be significant enough to warrant monitoring, especially when the procedure is done frequently. Sentinel lymph node mapping, used to identify the affected or sentinel node in gynaecological malignancies, also uses radioactive tracers such as Technetium-99 to identify the affected nodes. Although preliminary studies have indicated that exposure usually falls within the safe limits, many factors, such as time from injection to surgery and distance between the patient's injection site and the surgeon's abdomen, play a significant role . Physicians and trainees predominantly dealing with gynaecological malignancy surgeries, should consider the cumulative dose received. Due to the fear of discrimination by their peers or senior consultants, some of the trainees or resident doctors prefer not to disclose their pregnant status till late. If unaware of the risks during such procedures, they may expose themselves to ionizing radiation, causing inadvertent self-harm. There is a need for a cordial work environment in which the resident doctors can declare their pregnant status, free from the fear of discrimination, and can hold posts in areas that do not involve radiation exposure. Infectious diseases Women during pregnancy are more susceptible to certain viral infections, predominantly due to impaired pathogen clearance and hormonal and immunological alterations. The risk for health care workers, including doctors, is increased manifold mainly due to the prolonged contact with infected patients and that too in closed areas such as birthing suites, high dependency units, and intensive care units. If a pregnant health care worker acquires viral infections such as rubella, cytomegalovirus, herpes simplex, and varicella, particularly at the time of organogenesis, it can be devastating for the foetus. The teratogenic effects of these viruses are well known. Pregnant doctors should not be involved in the care of patients with these infectious diseases. The current pandemic of COVID-19 has made this risk of infectious diseases even more apparent. CDC recognizes health care workers, including doctors, nurses, dentists, paramedics, emergency medical technicians, laboratory personnel collecting and handling samples from infected persons, and morgue workers performing autopsies as the group at highest risk of acquiring Coronavirus . Among health care professionals, certain professionals such as anaesthesiologists, otorhinolaryngologists, dentists, ophthalmologists are at exceptionally high risk because their work demands proximity with patients’ respiratory tracts. Obstetricians also come under very high risk, mainly due to prolonged exposure, especially during labour. CDC and WHO recommend using N95 masks for health care workers, especially during the care of patients with diseases that involve droplet transmissions, such as tuberculosis, severe acute respiratory syndrome (SARS), and COVID-19 . Even after complying with proper protection and preventive measures and using personal protective equipment to protect themselves, health care workers have been affected by the disease. During the initial months of the pandemic, a few efforts were taken to limit the amount of exposure, such as delaying elective surgical procedures and undertaking only emergency ones. However, these practice guidelines cannot be extended to specialities like obstetrics. Every case presenting in labour or needing labour induction due to feto-maternal indications can be treated as an emergency, as any delay can be life-threatening. Obstetrics is probably the only speciality in medicine where the number of cases and surgeries did not decrease, despite the fear of infection. The risks extend beyond the pandemic period and apply to other infectious diseases such as influenza and tuberculosis. The dilemma with personal protective equipment in pregnancy Pregnancy is associated with profound changes in normal respiratory physiology. Dyspnoea is usually a common symptom in late pregnancy. Both mechanical factors (due to the enlarged gravid uterus) and hormonal factors play a role in this. Oxygen consumption increases from the first trimester, increasing by around 30% per term due to maternal metabolic processes and foetal demands . Increased oestrogen causes hyperaemia, oedema, hypersecretion, and friability of the mucosa of the respiratory tract . There is an increase in the number and sensitivity of hypothalamic and medullary progesterone receptors in pregnancy, leading to a rise in the sensitivity of peripheral chemoreceptors to hypoxic conditions . Progesterone also leads to a decreased threshold and increased respiratory centre sensitivity to carbon dioxide. These physiological changes increase the load on the respiratory system in pregnancy. Keeping in mind the airborne transmission of COVID-19 , the Scientific Advisory Group for Emergencies (SAGE) even recommends that HCW caring for patients with suspected or confirmed COVID-19 may need higher grade protective masks, such as FFP3 masks equivalent to N99, to protect them from contracting the virus through the air. Although a few studies suggest that these masks are not associated with adverse effects in pregnancy, these studies are primarily restricted by limited time of exposure, i.e., a maximum of one hour . But in today's scenario, the duration of mask-wearing by pregnant women, especially health care workers, is at least 6–8 hours at a stretch. Several side effects have been reported in health care workers using these face masks for a prolonged duration. These include headache, dryness in the eyes and nose, acne, epistaxis, skin breakdown, and even impaired cognition . Pregnant women, especially in the late second or third trimester, may not be able to maintain their required minute ventilation while breathing through N95 respirators. The workload on breathing increases significantly, leading to decreased oxygen uptake and increased carbon dioxide concentration . Hypoxia and hypercarbia, mainly due to re-breathing caused by retained carbon dioxide in the mask's dead space, occur on prolonged mask usage These changes are evident even at rest and may be exacerbated on mild to moderate exertion. Long-term exposure of the foetus to this increased carbon dioxide level has not been studied. However, some studies suggest that it affects foetal cerebral oxygenation, which may be by regulating the cerebral blood flow and shifting the oxyhaemoglobin dissociation curve . Pregnant women with respiratory ailments such as bronchial asthma or other chronic lung diseases could be at much higher risk. The use of medical and surgical masks and other external airflow resistive load devices has been found to impact some hemodynamic parameters such as diastolic blood pressure and mean blood pressure significantly, in pregnant women and non-pregnant women alike. Although the effect noted was mild, even an increase of 10 mm Hg in a patient with preeclampsia or chronic hypertension could be harmful to the mother and the unborn foetus. Sharps injuries and bloodborne infections All surgeons have a very high risk of needle-stick injury, and obstetrics and gynaecology as a speciality are no different. Resident doctors are at an exceptionally high risk as they are not trained in personal protection measures, and most of them are learning to hold and manipulate the instruments for the first time. A survey of around 700 resident doctors found that almost 99% of them had experienced a sharps injury . The probability of acquiring infection from large-bore needle-stick injury has been reported to be as high as 40% in workers not vaccinated against hepatitis B virus, 1.8% for hepatitis C virus, and 0.3% for human immunodeficiency virus (HIV) . Until the time they are properly trained in handling and manipulating surgical instruments and needles, pregnant resident doctors should not be involved in the surgery of patients with HIV, hepatitis B, and hepatitis C. Further adequate vaccination and good antibody titers against hepatitis B should be a rule for trainee doctors joining any surgical speciality. They should also be trained to handle blood and body fluid spills and be adequately informed regarding post-exposure prophylaxis in case of accidental needle-stick injury. Physician burnout and stress Pregnancy during residency and speciality training in medicine and surgery is challenging. The residency period, especially in clinical specialties like obstetrics and gynaecology, is marked by long duty hours, rotating night shifts, and prolonged standing. Working long hours during the first trimester of pregnancy is associated with threatened abortion and preterm birth . A recent survey on 347 general surgeons, who had at least one pregnancy during residency, reported unmitigated work schedules during pregnancy. There is a negative stigma associated with pregnancy during training. They were also dissatisfied with maternity leave options and inadequate lactation and childcare support. They also desired a better mentorship on work-life integration . Inadequate support by the co-doctors is expected because they are themselves engrossed in their heavy duties. Several studies in the past have stressed that most residents felt inconvenienced by the presence of pregnant or lactating colleagues, as they were forced to cover their responsibilities during their absence (24). Resident doctors during pregnancy and lactation face unique challenges such as arranging for child care during their extended period of absence, maintaining lactation during intense night duties, and frequent breaks for pumping breast milk to ensure proper milk output. Inadequate policies related to pregnancy and parenting may sometimes even adversely affect their career preferences, sometimes even promoting them to quit their career as medical professional . Peer support Fulfilling lactation and child care goals is another challenge for health care workers across all specialities. Maintaining an adequate breast milk supply requires either frequent feeding or frequent pumping, both of which need frequent short breaks in the working schedule. More than half of the doctors and supporting staff opt to quit breastfeeding at an earlier stage than they wished. The nature of the work of health care professionals is such that taking even a short break without proper replacement can even cost lives. There is presently no provision adjustment in the nature of duties of pregnant and lactating health care workers. A written policy regarding avoidance of long duty hours and prolonged standing, and provision of intermittent periods of rest, should be made and brought into practice in health care settings. Provision of lactation rooms with facilities for pumping and storing breast milk should be mandatory. Lactating employees should be provided with frequent short breaks to pump or breastfeed. Although some hospitals do offer an in-house creche and child care facility, taking the baby to hospitals is again a dilemma, especially at the time of the spread of a pandemic which is highly infectious. Exposure to anaesthetic gases and surgical smoke and other chemicals Nitrous oxide and halogenated agents constitute the predominant inhalational agents used for anaesthesia in operation theatres. When inhalational agents are used for induction predominantly for day care procedures or minor surgeries in gynaecology, some waste gases are inadvertently released into the operating room and inhaled by surgeons and their supporting staff. These gases have been associated with adverse pregnancy outcomes such as spontaneous abortions and congenital anomalies in the foetus when inhaled by pregnant women, especially during earlier gestation . Therefore, adequate scavenging systems should be a must in all operation theatres to minimize exposure . Surgical smoke refers to waste gases emitted in the operation theatres due to the burning of tissues with energy sources such as electrocautery. The content of surgical smoke includes water gases containing chemicals such as benzene, 1,2-dichloroethane, and toluene, which are associated with miscarriages, congenital birth defects, foetal growth restriction , and preterm labour . Many studies have found a very high concentration of fine and ultrafine particulate matter when smoke was released during laparoscopic procedures. Although these particles and chemicals have not been studied in much detail, the effects that these particles and chemicals have on the unborn foetus could be significantly grave. Cytoreductive surgery and hyperthermic intraperitoneal chemotherapy (HIPEC) are increasingly used as a treatment modality for ovarian malignancies and peritoneal carcinomatosis. The chemotherapy agents used include mitomycin c and platinum-based compounds such as cisplatin and carboplatin. Pregnant doctors can be exposed through both inhalation and skin contact. These agents are associated with multiple harmful effects in pregnancy, including miscarriage and congenital malformations . Current recommendations are that pregnant women or those planning to become pregnant should keep themselves away from chemotherapy agents and in operation theatres where HIPEC is being done . Violence against doctors It is a paradox that a profession as noble as health care, with the mission to care for people when their need for care is at a maximum, that is, when they are unwell or terminally ill, is at significant risk of workplace violence. The World Health Organisation defines workplace violence as 'incidents where staff is abused, threatened or assaulted in the circumstances related to their work, including commuting to and from work, involving an explicit or implicit challenge to their safety, well-being or health' . It has been observed that around one-fourth of all violent accidents at work occur in the health care sector, and more than half of all health care workers have experienced violence in some form at their workplace . Women, predominantly of the reproductive age group, represent nearly 80% of the health care workforce . The effects of direct physical violence are well known in foetal injury and death, abruptio placentae, and premature rupture of membranes. The indirect effects of verbal, physical, and even sexual abuse include psychological stress and anxiety, which is well known to cause adverse pregnancy outcomes . What can be done? Pregnant health care professionals including those specialising in obstetrics and gynaecology are themselves often not prepared to identify risk factors that can adversely affect their health at the work place. Occupational risk assessment models which incorporate all possible risk factors , should be implemented in all hospitals. Flexible working policies for their pregnant employees, including avoidance of night shifts and long shifts, especially during the trimesters that involve the highest risk to the foetus have been rightly introduced by some universities such as Indiana University's emergency and internal medicine programs. Employment conditions that are pregnancy- and breastfeeding-friendly are the need of the hour. Radiation exposure in early pregnancy is very well known to be associated with teratogenic effects. Although the risk is usually overestimated, the consequences can be significant in cumulative doses. Threshold radiation effects (deterministic effects) occur over a dose threshold and result in cellular injury. Stochastic effects of radiation are incremental, appearing in a dose-response function without a threshold, and are thought to be the primary mechanism of increased risk of cancers . Various agencies have given thresholds that should be taken care of once pregnancy is confirmed, to minimize the effects on the foetus, mainly during organogenesis. The International Commission on Radiological Protection recommends that after a worker declares her pregnancy, the occupational radiation dose should not exceed one mSv during the remainder of the pregnancy. The National Council on Radiation Protection and Measurements, in the United States, recommends a radiation dose limit of 0.5 mSv per month once pregnancy is confirmed to ensure a low exposure during susceptible periods of gestation . The US Environmental Protection Agency recommends a limit of 5 mSv for the entire gestational period . Although only very few procedures in obstetrics and gynaecology use ionizing radiation, hysterosalpingography (HSG) is a standard procedure used to evaluate the patency of fallopian tubes in women presenting with subfertility. The HSG procedure requires the radiologist or preferably a gynaecologist (who is not trained to handle ionizing radiation) to hold the cannula and inject the contrast medium into the patient's cervix while she is being irradiated. The supporting personnel also remain close to the patient. Though a lead apron is worn, frequent or multiple procedures can lead to significant exposure to ionizing radiation, which can be worrisome, especially in the first trimester due to teratogenic effects . A few studies have demonstrated that the dose to the extremities may also be significant enough to warrant monitoring, especially when the procedure is done frequently. Sentinel lymph node mapping, used to identify the affected or sentinel node in gynaecological malignancies, also uses radioactive tracers such as Technetium-99 to identify the affected nodes. Although preliminary studies have indicated that exposure usually falls within the safe limits, many factors, such as time from injection to surgery and distance between the patient's injection site and the surgeon's abdomen, play a significant role . Physicians and trainees predominantly dealing with gynaecological malignancy surgeries, should consider the cumulative dose received. Due to the fear of discrimination by their peers or senior consultants, some of the trainees or resident doctors prefer not to disclose their pregnant status till late. If unaware of the risks during such procedures, they may expose themselves to ionizing radiation, causing inadvertent self-harm. There is a need for a cordial work environment in which the resident doctors can declare their pregnant status, free from the fear of discrimination, and can hold posts in areas that do not involve radiation exposure. Women during pregnancy are more susceptible to certain viral infections, predominantly due to impaired pathogen clearance and hormonal and immunological alterations. The risk for health care workers, including doctors, is increased manifold mainly due to the prolonged contact with infected patients and that too in closed areas such as birthing suites, high dependency units, and intensive care units. If a pregnant health care worker acquires viral infections such as rubella, cytomegalovirus, herpes simplex, and varicella, particularly at the time of organogenesis, it can be devastating for the foetus. The teratogenic effects of these viruses are well known. Pregnant doctors should not be involved in the care of patients with these infectious diseases. The current pandemic of COVID-19 has made this risk of infectious diseases even more apparent. CDC recognizes health care workers, including doctors, nurses, dentists, paramedics, emergency medical technicians, laboratory personnel collecting and handling samples from infected persons, and morgue workers performing autopsies as the group at highest risk of acquiring Coronavirus . Among health care professionals, certain professionals such as anaesthesiologists, otorhinolaryngologists, dentists, ophthalmologists are at exceptionally high risk because their work demands proximity with patients’ respiratory tracts. Obstetricians also come under very high risk, mainly due to prolonged exposure, especially during labour. CDC and WHO recommend using N95 masks for health care workers, especially during the care of patients with diseases that involve droplet transmissions, such as tuberculosis, severe acute respiratory syndrome (SARS), and COVID-19 . Even after complying with proper protection and preventive measures and using personal protective equipment to protect themselves, health care workers have been affected by the disease. During the initial months of the pandemic, a few efforts were taken to limit the amount of exposure, such as delaying elective surgical procedures and undertaking only emergency ones. However, these practice guidelines cannot be extended to specialities like obstetrics. Every case presenting in labour or needing labour induction due to feto-maternal indications can be treated as an emergency, as any delay can be life-threatening. Obstetrics is probably the only speciality in medicine where the number of cases and surgeries did not decrease, despite the fear of infection. The risks extend beyond the pandemic period and apply to other infectious diseases such as influenza and tuberculosis. Pregnancy is associated with profound changes in normal respiratory physiology. Dyspnoea is usually a common symptom in late pregnancy. Both mechanical factors (due to the enlarged gravid uterus) and hormonal factors play a role in this. Oxygen consumption increases from the first trimester, increasing by around 30% per term due to maternal metabolic processes and foetal demands . Increased oestrogen causes hyperaemia, oedema, hypersecretion, and friability of the mucosa of the respiratory tract . There is an increase in the number and sensitivity of hypothalamic and medullary progesterone receptors in pregnancy, leading to a rise in the sensitivity of peripheral chemoreceptors to hypoxic conditions . Progesterone also leads to a decreased threshold and increased respiratory centre sensitivity to carbon dioxide. These physiological changes increase the load on the respiratory system in pregnancy. Keeping in mind the airborne transmission of COVID-19 , the Scientific Advisory Group for Emergencies (SAGE) even recommends that HCW caring for patients with suspected or confirmed COVID-19 may need higher grade protective masks, such as FFP3 masks equivalent to N99, to protect them from contracting the virus through the air. Although a few studies suggest that these masks are not associated with adverse effects in pregnancy, these studies are primarily restricted by limited time of exposure, i.e., a maximum of one hour . But in today's scenario, the duration of mask-wearing by pregnant women, especially health care workers, is at least 6–8 hours at a stretch. Several side effects have been reported in health care workers using these face masks for a prolonged duration. These include headache, dryness in the eyes and nose, acne, epistaxis, skin breakdown, and even impaired cognition . Pregnant women, especially in the late second or third trimester, may not be able to maintain their required minute ventilation while breathing through N95 respirators. The workload on breathing increases significantly, leading to decreased oxygen uptake and increased carbon dioxide concentration . Hypoxia and hypercarbia, mainly due to re-breathing caused by retained carbon dioxide in the mask's dead space, occur on prolonged mask usage These changes are evident even at rest and may be exacerbated on mild to moderate exertion. Long-term exposure of the foetus to this increased carbon dioxide level has not been studied. However, some studies suggest that it affects foetal cerebral oxygenation, which may be by regulating the cerebral blood flow and shifting the oxyhaemoglobin dissociation curve . Pregnant women with respiratory ailments such as bronchial asthma or other chronic lung diseases could be at much higher risk. The use of medical and surgical masks and other external airflow resistive load devices has been found to impact some hemodynamic parameters such as diastolic blood pressure and mean blood pressure significantly, in pregnant women and non-pregnant women alike. Although the effect noted was mild, even an increase of 10 mm Hg in a patient with preeclampsia or chronic hypertension could be harmful to the mother and the unborn foetus. All surgeons have a very high risk of needle-stick injury, and obstetrics and gynaecology as a speciality are no different. Resident doctors are at an exceptionally high risk as they are not trained in personal protection measures, and most of them are learning to hold and manipulate the instruments for the first time. A survey of around 700 resident doctors found that almost 99% of them had experienced a sharps injury . The probability of acquiring infection from large-bore needle-stick injury has been reported to be as high as 40% in workers not vaccinated against hepatitis B virus, 1.8% for hepatitis C virus, and 0.3% for human immunodeficiency virus (HIV) . Until the time they are properly trained in handling and manipulating surgical instruments and needles, pregnant resident doctors should not be involved in the surgery of patients with HIV, hepatitis B, and hepatitis C. Further adequate vaccination and good antibody titers against hepatitis B should be a rule for trainee doctors joining any surgical speciality. They should also be trained to handle blood and body fluid spills and be adequately informed regarding post-exposure prophylaxis in case of accidental needle-stick injury. Pregnancy during residency and speciality training in medicine and surgery is challenging. The residency period, especially in clinical specialties like obstetrics and gynaecology, is marked by long duty hours, rotating night shifts, and prolonged standing. Working long hours during the first trimester of pregnancy is associated with threatened abortion and preterm birth . A recent survey on 347 general surgeons, who had at least one pregnancy during residency, reported unmitigated work schedules during pregnancy. There is a negative stigma associated with pregnancy during training. They were also dissatisfied with maternity leave options and inadequate lactation and childcare support. They also desired a better mentorship on work-life integration . Inadequate support by the co-doctors is expected because they are themselves engrossed in their heavy duties. Several studies in the past have stressed that most residents felt inconvenienced by the presence of pregnant or lactating colleagues, as they were forced to cover their responsibilities during their absence (24). Resident doctors during pregnancy and lactation face unique challenges such as arranging for child care during their extended period of absence, maintaining lactation during intense night duties, and frequent breaks for pumping breast milk to ensure proper milk output. Inadequate policies related to pregnancy and parenting may sometimes even adversely affect their career preferences, sometimes even promoting them to quit their career as medical professional . Fulfilling lactation and child care goals is another challenge for health care workers across all specialities. Maintaining an adequate breast milk supply requires either frequent feeding or frequent pumping, both of which need frequent short breaks in the working schedule. More than half of the doctors and supporting staff opt to quit breastfeeding at an earlier stage than they wished. The nature of the work of health care professionals is such that taking even a short break without proper replacement can even cost lives. There is presently no provision adjustment in the nature of duties of pregnant and lactating health care workers. A written policy regarding avoidance of long duty hours and prolonged standing, and provision of intermittent periods of rest, should be made and brought into practice in health care settings. Provision of lactation rooms with facilities for pumping and storing breast milk should be mandatory. Lactating employees should be provided with frequent short breaks to pump or breastfeed. Although some hospitals do offer an in-house creche and child care facility, taking the baby to hospitals is again a dilemma, especially at the time of the spread of a pandemic which is highly infectious. Nitrous oxide and halogenated agents constitute the predominant inhalational agents used for anaesthesia in operation theatres. When inhalational agents are used for induction predominantly for day care procedures or minor surgeries in gynaecology, some waste gases are inadvertently released into the operating room and inhaled by surgeons and their supporting staff. These gases have been associated with adverse pregnancy outcomes such as spontaneous abortions and congenital anomalies in the foetus when inhaled by pregnant women, especially during earlier gestation . Therefore, adequate scavenging systems should be a must in all operation theatres to minimize exposure . Surgical smoke refers to waste gases emitted in the operation theatres due to the burning of tissues with energy sources such as electrocautery. The content of surgical smoke includes water gases containing chemicals such as benzene, 1,2-dichloroethane, and toluene, which are associated with miscarriages, congenital birth defects, foetal growth restriction , and preterm labour . Many studies have found a very high concentration of fine and ultrafine particulate matter when smoke was released during laparoscopic procedures. Although these particles and chemicals have not been studied in much detail, the effects that these particles and chemicals have on the unborn foetus could be significantly grave. Cytoreductive surgery and hyperthermic intraperitoneal chemotherapy (HIPEC) are increasingly used as a treatment modality for ovarian malignancies and peritoneal carcinomatosis. The chemotherapy agents used include mitomycin c and platinum-based compounds such as cisplatin and carboplatin. Pregnant doctors can be exposed through both inhalation and skin contact. These agents are associated with multiple harmful effects in pregnancy, including miscarriage and congenital malformations . Current recommendations are that pregnant women or those planning to become pregnant should keep themselves away from chemotherapy agents and in operation theatres where HIPEC is being done . It is a paradox that a profession as noble as health care, with the mission to care for people when their need for care is at a maximum, that is, when they are unwell or terminally ill, is at significant risk of workplace violence. The World Health Organisation defines workplace violence as 'incidents where staff is abused, threatened or assaulted in the circumstances related to their work, including commuting to and from work, involving an explicit or implicit challenge to their safety, well-being or health' . It has been observed that around one-fourth of all violent accidents at work occur in the health care sector, and more than half of all health care workers have experienced violence in some form at their workplace . Women, predominantly of the reproductive age group, represent nearly 80% of the health care workforce . The effects of direct physical violence are well known in foetal injury and death, abruptio placentae, and premature rupture of membranes. The indirect effects of verbal, physical, and even sexual abuse include psychological stress and anxiety, which is well known to cause adverse pregnancy outcomes . Pregnant health care professionals including those specialising in obstetrics and gynaecology are themselves often not prepared to identify risk factors that can adversely affect their health at the work place. Occupational risk assessment models which incorporate all possible risk factors , should be implemented in all hospitals. Flexible working policies for their pregnant employees, including avoidance of night shifts and long shifts, especially during the trimesters that involve the highest risk to the foetus have been rightly introduced by some universities such as Indiana University's emergency and internal medicine programs. Employment conditions that are pregnancy- and breastfeeding-friendly are the need of the hour. The major employment issues faced by pregnant health care workers include pregnancy-related discrimination, accommodations in the distribution of work or duties, keeping in mind the health of mother-foetus duo, job-protected leave, and wage replacement while on maternity leave. Employment conditions that create more optimal work environments for pregnant employees are the need of the hour. Women, predominantly of the reproductive age group, constitute a significant proportion of the health care work force. Pregnant obstetrics and gynaecology trainees and physicians face numerous occupational risks, including those of infectious diseases, radiation exposure, stress and burnout, violence against doctors, and even lack of peer support. Employment conditions that create more optimal work environments for pregnant employees are the need of the hour. |
Human scent signature on cartridge case survives gun being fired: A preliminary study on a potential of scent residues as an identification tool | 5023c69a-9f4a-40d9-83fa-7cb2e3c5f0c7 | 10032514 | Forensic Medicine[mh] | Nowadays, standard forensic procedures regarding firearms and crime scenes involve finding and securing fingerprints, gunshot residues, and scent traces. Specifically, human scent traces, if secured from a gun handle or gun slide on the crime scene, are generally compared by specially trained dogs . The ability of trained dogs to distinguish between human individuals based on their ’scent signatures’ is well known and described [ – ]. Human scent is a complex mixture of thousands of chemical compounds, mainly hydrocarbons, heterocompounds such as ketones, sterols, sterones, fatty acids, and esters of fatty acids [ – ]. Although the chemical character of the scent compounds could suggest a lesser resistance of the scent traces when exposed to heating, fire, and explosions, there are studies indicating that the scent signatures survive such conditions . This phenomenon can be explained by the multiplicity of scent signature. The study of Doležal et al. confirmed that specially trained dogs can identify individuals by different chemical fractions of their scents. In addition, the least volatile fraction of chemical compounds proved to be the most relevant for olfactory identification. However, tests performed by specially trained dogs are very often challenged in court as there is no concrete numerical proof of the results. The aim of this study was to acquire evidence whenever the human scent can survive extreme conditions during the gunshot, to analyze the scent residues by two-dimensional chromatography coupled with mass spectrometry and perform comparison analyzes while mimicking the procedure performed by specially trained dogs. The results indeed supported the assumption that it is possible to secure human scent traces from post-blast items (fired cartridges) and compare such residue samples against samples collected under the laboratory condition.
Ethics statements This study was approved by the Institutional Review Board of the University of Chemistry and Technology, Prague (approval number EK/8/2020) and complies with the Declaration of Helsinki for Medical Research involving Human Subjects. Volunteers gave their written consent to collect, analyze, and store their scent samples ( )–for this subbranch of research of human scent, volunteers were not asked to fill questionary data, as those data were outside the scope of the experiment. Volunteers In total, twenty volunteers (ten males and ten females aged from 25 to 40 years) provided their scent samples through the experiments. Repetitive scent collection took place approximately once a week at the same place at the same time. Sampling procedure and sample collection First, the volunteers washed their hands with non-perfumed soap (Cormen, CZ), rinsed them with a warm tap water until there were no soap bubbles, and let the hands dry in the air to complete dryness. Subsequently, the volunteers rubbed their palms to activate the scent glands. After 5 minutes, the volunteers received cartridges (S&B 7.65 Browning). The sampling itself continued for 10 minutes. Volunteers were holding and rubbing the projectiles in both hands. Once the collection procedure was over, the cartridges were transferred to vials and sealed. Comparative samples on glass beads were collected with the same procedure and were referred as standard scent samples in the following text. In an experimental shooting room, the cartridges were loaded by a designated person (a researcher who was not involved in the volunteer sample pool) into the pistol and fired (the shooter was another researcher who was not involved in the volunteer sample pool). The fired cartridges were left to cool down on the sterile surface (waterproof surgical drapes) and then moved into the vials (see ). Both persons present at the shooting site were wearing surgical gloves and respirators to prevent sample contamination. The time between both sampling procedures (glass beads sampling and projectile sampling) was as short as possible (less than 12 hours). The extraction procedure of the collected scent samples was the same for both the scent carriers, for the glass bead samples and for the fired cartridge cases samples. First, 2 mL of ethanol (> = 99.8%, Penta CZ) were added to each vial with the collected scent sorbent. At this initial stage of extraction, the sample vials were placed in an ultrasonic bath (Fisher Scientific, UK) for 10 minutes and then on a shaker (1000 rpm, IKAVIBRAX VXR basic, Germany) for an additional 10 minutes. The solution was pipetted into 2 mL champagne vials and dried to complete dryness under reduced pressure (concentrator Genevac EZ-2, USA). Right before the analysis, the sample was dissolved again in 70 μL of ethanol. For contaminant identification purposes, blank samples of glass beads and fired cartridge cases were taken for each sampling procedure.
This study was approved by the Institutional Review Board of the University of Chemistry and Technology, Prague (approval number EK/8/2020) and complies with the Declaration of Helsinki for Medical Research involving Human Subjects. Volunteers gave their written consent to collect, analyze, and store their scent samples ( )–for this subbranch of research of human scent, volunteers were not asked to fill questionary data, as those data were outside the scope of the experiment.
In total, twenty volunteers (ten males and ten females aged from 25 to 40 years) provided their scent samples through the experiments. Repetitive scent collection took place approximately once a week at the same place at the same time.
First, the volunteers washed their hands with non-perfumed soap (Cormen, CZ), rinsed them with a warm tap water until there were no soap bubbles, and let the hands dry in the air to complete dryness. Subsequently, the volunteers rubbed their palms to activate the scent glands. After 5 minutes, the volunteers received cartridges (S&B 7.65 Browning). The sampling itself continued for 10 minutes. Volunteers were holding and rubbing the projectiles in both hands. Once the collection procedure was over, the cartridges were transferred to vials and sealed. Comparative samples on glass beads were collected with the same procedure and were referred as standard scent samples in the following text. In an experimental shooting room, the cartridges were loaded by a designated person (a researcher who was not involved in the volunteer sample pool) into the pistol and fired (the shooter was another researcher who was not involved in the volunteer sample pool). The fired cartridges were left to cool down on the sterile surface (waterproof surgical drapes) and then moved into the vials (see ). Both persons present at the shooting site were wearing surgical gloves and respirators to prevent sample contamination. The time between both sampling procedures (glass beads sampling and projectile sampling) was as short as possible (less than 12 hours). The extraction procedure of the collected scent samples was the same for both the scent carriers, for the glass bead samples and for the fired cartridge cases samples. First, 2 mL of ethanol (> = 99.8%, Penta CZ) were added to each vial with the collected scent sorbent. At this initial stage of extraction, the sample vials were placed in an ultrasonic bath (Fisher Scientific, UK) for 10 minutes and then on a shaker (1000 rpm, IKAVIBRAX VXR basic, Germany) for an additional 10 minutes. The solution was pipetted into 2 mL champagne vials and dried to complete dryness under reduced pressure (concentrator Genevac EZ-2, USA). Right before the analysis, the sample was dissolved again in 70 μL of ethanol. For contaminant identification purposes, blank samples of glass beads and fired cartridge cases were taken for each sampling procedure.
All analyzes were performed by the GC×GC system with a LECO Pegasus 4D-C detector (LECO Corp, USA), a 7890B gas chromatograph (Agilent, USA), and a MPS robotic MPS multipurpose sampler autosampler from Gerstel (Germany). The first column was a 30 m long Rtx-200 column (0.25 mm; 0.25 μm—Restek, USA) with an additional 2 m as the precolumn. The second column was a 1 m Thermo-5HT (0.25 mm; 0.25 μm—Thermo Scientific, USA). Helium (purity ≥ 99.9995 from Linde CZ) was used as the carrier gas with the flow set at 1.5 mL / min. The temperature program was the following: 40°C—2 minutes hold—5°C/min ramp—320°C—10 minutes hold. The modulator offset against the secondary oven was set at 15° C and the secondary oven offset against the primary oven was set at 5° C. There were three modulation periods during the run: 6s (start→1700 th s), 8s (1700 th to 2588 th s), and 10s (2588 th → end). The injection chamber was heated to 280° C. The injection volume of the sample was 1μL (splitless). The temperature for the transfer line was 280° C and for the ion source was 250° C. The electron ionization energy was 70eV. 29–800 m/z fragments were scanned at rate of 200 spectra per second. This method was previously evaluated as the best possible option for measuring human scent samples . Data processing The LECO® software (version 4.72.0.0, LECO Corp., USA) was used for the basic evaluation of the measured chromatograms. The retention indexes (RI) were calculated for each chromatogram to treat the time shift in the first dimension. The RIs could not be based on the elution of n -alkanes because of their lack or overlap with the contaminant in the collected samples. All RI calculations were based on siloxanes. The main advantage of the siloxane approach was their presence as contaminants, even in the blank samples, which was crucial for the automatic sorting of the detected peaks. After data alignment via RIs, all detected peaks were inspected, and the subset of 75 selected chemical compounds presented in standard and fired sample datasets was created and further processed. Each sample, its peaks, and their areas, respectively, were transformed to a matrix of peak area ratio ( 75×75 dimension–if the chemical compound was absent in all inspected samples, the dimensionality was reduced) and the matrix was then rewritten as a single row representation. Peak area ratios where the divisor was equal to zero were set to zero. The peak area ratios with the ’infinite’ value were also approximated by zero. Eliminating the peak area ratios with zero values was the next step in the data processing procedure. The selection algorithm for the ’significant’ peak area ratios for model-based analyses was the following: For each volunteer, only the peak area ratios with incidence rate ( T 1 ) of 75% (for example, peak area ratios had a nonzero value in eight out of ten samples) were chosen. It means that the zero_count parameter in the used script was set to 25% (incidence thresholds T x are equal to 100% − zero_count parameter). Then, the N number of peak area ratios with the least standard intra-subject deviation (sorted in ascending order) through all samples were included in further calculations. The N parameter was estimated and optimised based on the results of the PCA. The N parameter can range from 1 to all the peak area ratios. The N range tested in this study ranged from 25 up to 500 with a step of 25). Finally, the ’significant’ peak area ratios of all volunteers were merged and, again, the incidence rate threshold ( T 2 ) was established. The T 2 was also estimated and optimised in range from 25% up to 75% with step of 25%. When only the most common peak area ratios were involved in all samples, the T 2 threshold was higher. If more ’ volunteer-specific’ peak area ratios were targeted in the data set, the T 2 threshold was set to lower values. The models are named after the used parameters: Model_ Nparamater_T 2 parameter . Data comparison and evaluation The normality of the experimental data set was tested with the multivariate normality test, specifically Mardia’s test . Information about the (non)normality of the distribution was important for the correct estimation of the parameter N value and an evaluation of the results from the similarity comparison. There were two main strategies for setting the N parameter. First one was to set rigid standard deviation threshold value, but this approach was not suitable for non-normal distributed data. Therefore, an approach of setting the rigid count of the peak area ratios with the least standard deviation was prioritized. Two different statistical approaches were used for the following data analyzes: Principal Component Analysis and similarity comparison method. Principal Component Analysis (PCA) was used solely as a visualization of possible trends in data variance and as a tool for parameter optimization–models with higher percentage of variance explained for 7 principal components were considered as more reliable. Different sets of zero-value filters (incidence rate threshold) and number of characteristic ratios ( N– see section Data processing) were tested. Subsequently, the best PCA models (models that describe the most variance in dataset and with separated volunteer clusters) served as the basis for similarity analyses in the next stages of our experiment. In addition, K- Nearest Neighbor (KNN) , Radius Nearest Neighbor (RNN), and Random Forest (RF) classification algorithms were applied on samples used in the model creation steps. All three algorithms were optimized via Grid Search algorithm with 5-fold cross validation with accuracy as the scoring argument, and the data had a train/test split ratio of 75% / 25%. The KNN was optimized on grid: K in range from 1 to 19 with weights set to uniform or distance weighted . The RNN had same weights grid setup and the radius parameter was tested in range from 1 to 99 with step of 10. The RF classification model was optimized on the following parameter grid: number of estimators – 1 to 200 wit step of 10; max_ features – 1 to 11 with step of 2; bootstrapping set to active or inactive, and if the bootstrapping was active, the oob scoring was tested on active and inactive level as well. For complete analysis of the data, first three principal components of PCA models were feed to the machine learning models (KNN, RNN and RF) and compared with the performance of models applied on original data. F 1 was used as the performance metric. Regarding similarity comparison, four similarity algorithms were used: Cosine similarity (CS) , Pearson’s correlation (PC) , Spearman’s correlation (SC) , and Kendall’s Tau correlation (KT) . Mardia’s test was performed as a built-in function of the Unscrambler® software (version 10.5). The PCA, KNN, RNN, RF, CS, PC, SC, and KT scripts were written in Python using Matplotlib , Numpy , Pandas , Scikit-learn and Scipy modules. All written scripts are published in the opened repository (see ). depicts a workflow of data processing and data comparison steps from the creation of similarity models to the final application of best performing models on standard scent samples versus shot cartridge samples comparison.
The LECO® software (version 4.72.0.0, LECO Corp., USA) was used for the basic evaluation of the measured chromatograms. The retention indexes (RI) were calculated for each chromatogram to treat the time shift in the first dimension. The RIs could not be based on the elution of n -alkanes because of their lack or overlap with the contaminant in the collected samples. All RI calculations were based on siloxanes. The main advantage of the siloxane approach was their presence as contaminants, even in the blank samples, which was crucial for the automatic sorting of the detected peaks. After data alignment via RIs, all detected peaks were inspected, and the subset of 75 selected chemical compounds presented in standard and fired sample datasets was created and further processed. Each sample, its peaks, and their areas, respectively, were transformed to a matrix of peak area ratio ( 75×75 dimension–if the chemical compound was absent in all inspected samples, the dimensionality was reduced) and the matrix was then rewritten as a single row representation. Peak area ratios where the divisor was equal to zero were set to zero. The peak area ratios with the ’infinite’ value were also approximated by zero. Eliminating the peak area ratios with zero values was the next step in the data processing procedure. The selection algorithm for the ’significant’ peak area ratios for model-based analyses was the following: For each volunteer, only the peak area ratios with incidence rate ( T 1 ) of 75% (for example, peak area ratios had a nonzero value in eight out of ten samples) were chosen. It means that the zero_count parameter in the used script was set to 25% (incidence thresholds T x are equal to 100% − zero_count parameter). Then, the N number of peak area ratios with the least standard intra-subject deviation (sorted in ascending order) through all samples were included in further calculations. The N parameter was estimated and optimised based on the results of the PCA. The N parameter can range from 1 to all the peak area ratios. The N range tested in this study ranged from 25 up to 500 with a step of 25). Finally, the ’significant’ peak area ratios of all volunteers were merged and, again, the incidence rate threshold ( T 2 ) was established. The T 2 was also estimated and optimised in range from 25% up to 75% with step of 25%. When only the most common peak area ratios were involved in all samples, the T 2 threshold was higher. If more ’ volunteer-specific’ peak area ratios were targeted in the data set, the T 2 threshold was set to lower values. The models are named after the used parameters: Model_ Nparamater_T 2 parameter .
The normality of the experimental data set was tested with the multivariate normality test, specifically Mardia’s test . Information about the (non)normality of the distribution was important for the correct estimation of the parameter N value and an evaluation of the results from the similarity comparison. There were two main strategies for setting the N parameter. First one was to set rigid standard deviation threshold value, but this approach was not suitable for non-normal distributed data. Therefore, an approach of setting the rigid count of the peak area ratios with the least standard deviation was prioritized. Two different statistical approaches were used for the following data analyzes: Principal Component Analysis and similarity comparison method. Principal Component Analysis (PCA) was used solely as a visualization of possible trends in data variance and as a tool for parameter optimization–models with higher percentage of variance explained for 7 principal components were considered as more reliable. Different sets of zero-value filters (incidence rate threshold) and number of characteristic ratios ( N– see section Data processing) were tested. Subsequently, the best PCA models (models that describe the most variance in dataset and with separated volunteer clusters) served as the basis for similarity analyses in the next stages of our experiment. In addition, K- Nearest Neighbor (KNN) , Radius Nearest Neighbor (RNN), and Random Forest (RF) classification algorithms were applied on samples used in the model creation steps. All three algorithms were optimized via Grid Search algorithm with 5-fold cross validation with accuracy as the scoring argument, and the data had a train/test split ratio of 75% / 25%. The KNN was optimized on grid: K in range from 1 to 19 with weights set to uniform or distance weighted . The RNN had same weights grid setup and the radius parameter was tested in range from 1 to 99 with step of 10. The RF classification model was optimized on the following parameter grid: number of estimators – 1 to 200 wit step of 10; max_ features – 1 to 11 with step of 2; bootstrapping set to active or inactive, and if the bootstrapping was active, the oob scoring was tested on active and inactive level as well. For complete analysis of the data, first three principal components of PCA models were feed to the machine learning models (KNN, RNN and RF) and compared with the performance of models applied on original data. F 1 was used as the performance metric. Regarding similarity comparison, four similarity algorithms were used: Cosine similarity (CS) , Pearson’s correlation (PC) , Spearman’s correlation (SC) , and Kendall’s Tau correlation (KT) . Mardia’s test was performed as a built-in function of the Unscrambler® software (version 10.5). The PCA, KNN, RNN, RF, CS, PC, SC, and KT scripts were written in Python using Matplotlib , Numpy , Pandas , Scikit-learn and Scipy modules. All written scripts are published in the opened repository (see ). depicts a workflow of data processing and data comparison steps from the creation of similarity models to the final application of best performing models on standard scent samples versus shot cartridge samples comparison.
Projectiles and fired cartridges as sampling materials First, a test was performed whenever the surface of the projectile is usable as a scent sampling material. The scent samples collected on the projectiles were qualitatively similar to those collected on the glass beads. However, the chromatograms of non-fired projectiles were, in general, less intensive than those of the glass beads’ sample. This means that the number of detected peaks and areas of detected peaks were lower than in the samples collected on glass beads. This could be the result of the surface characteristics (relief) of both sampling materials, as both glass and metal are, from the human scent point of view, inert to the collected samples. After it was proven that metal cases of projectiles can be used as a sampling material for sample collection, blank chromatogram and cartridge scent sample chromatogram (the metal case of the projectile left on the floor after shooting) were measured. Chromatograms of standard scent sample collected on glass beads, blank projectile, blank fired cartridge, and fired cartridge with sampled scent chromatograms are demonstrated in the ( ). The chromatograms of the blank projectiles and blank fired cartridges were qualitatively analogous to blank samples of glass beads. Furthermore, the blank samples of the fired cartridges contained a ‘cloud’ of aliphatic, both branched and non-branched alkanes, alkenes, and cyclic compounds ( , areas A and E)–this contamination likely comes from lubricating oils used for the maintenance of the weapons. Scent traces on fired cartridges The overall chemical composition of the scent trace endured conditions during the weapon’s firing. However, the elution area of fatty acid esters and less volatile compounds ( , area C) contained a smaller number of peaks, and the peaks were generally less intensive (as compared to standard scent samples). The elution pattern of the most frequent chemical compounds (such as ethyl esters, amines, and steroid derivates) and their summary are shown in and . Considering the nature of the compounds found (mostly C 16 and higher), it was clear that the volatility of the chemical compound would be a key factor in the study. The higher ethyl esters were previously reported in human scent samples . Squalene was also found in the cartridge samples, but its´ elution area overlapped with the “aliphatic cloud”. Thus, it was excluded from the further calculations. Nitrogen compounds as derivates of urea or nitrogen heterocycles were also reported in samples of human scent . Steroid chemical compounds have also been reported . In addition, steroids were marked as a ’scent precursors’ and products of the sebaceous glands along with squalene, fatty acids and their esters . Furthermore, compounds of the same chemical species were detected by gas chromatography in the samples of a fingerprints . Note that only a mass spectra comparison was done for the identification of the detected chemical compounds. Thus, the findings are discussed as a chemical group, not as the specific chemical compounds. Model-based analysis of fired cartridge samples First, it was necessary to determine whether the remaining scent compounds on the fired cartridges included enough information for the distinction between the volunteers. The PCA model was built on ratios with the least standard deviation (SD) through the samples of each volunteer. The best model ( ) was based on N = 75 ratios. However, to reduce the peak noise and keep the relevant values, we considered ratios with more than 50% nonzero values ( T 2 = 50%) through all samples (Model_75_50) only. The experimental data set exhibited a non-normal distribution (normal skewness, but kurtosis differed significantly). Originally, five volunteers provided five samples for this experiment, but one volunteers´ set exhibited severe contamination and therefore it excluded from the dataset. Additionally, two samples from two volunteers (Vol1 and Vol4) had to be excluded due to the severe contamination as well. KNN and RF trained models had F 1 score equal to zero. The RNN model failed the classification and the F 1 score was equal to 0.5. Similarity analysis of fired cartridge samples. One random sample from each volunteer was chosen as the ’unknown’ sample, and the rest of the samples (’standards’) were compared to that ’unknown’ sample. Every classification step was successful with the four approaches (CS, PC, SC, and KT) (see ). Model-based analysis of standard samples Five volunteers produced 11×5 samples in total, 10 standard scent samples, and one projectile sample. However, three standard samples (2 × Vol 2 and 1 × Vol 5) were excluded from further analyses due to low intensities (due to sampling or handling error, the measured chromatograms were almost blank). The initial PCA model was constructed on the ratios that were considered in Model 75_50 (Section Model-based analysis of fired cartridge samples). PCA analysis of standard samples based on this model failed to reveal differences between volunteers. Another model was built, and multiple least SD ratio setups were tested. The best discrimination power was achieved with N = 50 ratios (for each volunteer) and with the T 2 set to 25% across all inspected samples as demonstrated in . The experimental data set exhibited a non-normal distribution (skewness and kurtosis differed significantly from the normal distribution). The best performing KNN model reached F 1 score equal to 0.85. RNN model best F 1 score was equal to 0.60. The RF performed with the F 1 score equal to 1. The confusion matrix of the RF model is shown in the . The results and the model parameters are supplied in the ( ). For the completion of the data inspection, all models were retrained on the reduced data (first three principal components). RF model retained F 1 score equal to 1. F 1 score improved for both NN models: KNN scored 0.92 and RNN scored 0.85 in terms of F 1 score. The similarity analyses were carried out on the non-reduced (original) data, thus, the performance of the models when applied on original data was more significant for the proof-of-concept tests. Similarity analysis of fired cartridge samples versus standard scent samples The Spearman’s correlation coefficient was the best performing correlation from all tested approaches (for the full representation of the results see and ). For Volunteers 1, 3, and 5 all their standard scent samples had higher correlation coefficient compared to the fired cartridge sample than the standard samples of the other volunteers. For Volunteer 2, one standard sample a had lower correlation coefficient compared to the fired cartridge sample than the standard samples of another Volunteer. The same case occurred for Volunteer 4. Cosine similarity was achieved at a level similar to that of SC. It was correct in all cases for Volunteers 1, 4, and 5, and made one error for Volunteers 2 and 3. Kendall’s tau reached the top score for the same volunteers as SC did, and similarly, made one error in the case of Volunteer 2. However, KT made two errors when comparing Volunteer 4’s samples. Pearson’s correlation reached the highest score only in two cases, Volunteer 1 and 5. PC made one error in comparing the samples of Volunteer 3 and 4, and two errors for Volunteer 2. The reason why SC and KT comparisons achieved better results is probably that, unlike PC, these correlations are more suitable for data with a non-normal distribution. Fired cartridge sample versus standard scent sample comparison: Model application For the final comparison, a group of ten volunteers different from the group from previous experiments was asked to provide two samples–one collected on the glass beads (’standard sample’), and one collected on the projectile (’fired cartridge sample’). The samples were treated as stated in section Sampling procedure and sample collection. The structure of an experiment was designed to simulate the same procedure as that performed by specially trained dogs . Each shot cartridge sample was compared to ten standard samples–a comparison row . For every comparison row , the algorithm was supposed to detect and mark the standard sample that was most similar to the fired cartridge sample. The 50_75 model (Section Model-based analysis of standard samples) method was applied to a comparison process. However, the classification power of the model failed and was successful in less than 50% of the classification attempts. The main reason might be the fact that the model was built on standard scent sample rather than on scent residue samples. This probably led to a selection of peak area rations significant for distinguishing between standard scent samples, not shot cartridge ones. Model 75_50 (Section Model-based analysis of fired cartridge samples) performed with the same results (successful in less than 50% of the classification attempts) for Pearson’s correlation, cosine similarity, and Kendall’s tau. However, the Spearman’s correlation reached a 100% success rate (see ) because it was built on the significant peak area ratios detected on the shot cartridges. However, the standard scent sample of Vol 4 had a higher similarity with the fired cartridge sample of Vol 6. The relevant ratios are listed in . To inspect if the top two correlations are significantly different, both results were transformed to the z-scores and the test for the significance of the difference between them was calculated with p-value set to 0.05 (see ).
First, a test was performed whenever the surface of the projectile is usable as a scent sampling material. The scent samples collected on the projectiles were qualitatively similar to those collected on the glass beads. However, the chromatograms of non-fired projectiles were, in general, less intensive than those of the glass beads’ sample. This means that the number of detected peaks and areas of detected peaks were lower than in the samples collected on glass beads. This could be the result of the surface characteristics (relief) of both sampling materials, as both glass and metal are, from the human scent point of view, inert to the collected samples. After it was proven that metal cases of projectiles can be used as a sampling material for sample collection, blank chromatogram and cartridge scent sample chromatogram (the metal case of the projectile left on the floor after shooting) were measured. Chromatograms of standard scent sample collected on glass beads, blank projectile, blank fired cartridge, and fired cartridge with sampled scent chromatograms are demonstrated in the ( ). The chromatograms of the blank projectiles and blank fired cartridges were qualitatively analogous to blank samples of glass beads. Furthermore, the blank samples of the fired cartridges contained a ‘cloud’ of aliphatic, both branched and non-branched alkanes, alkenes, and cyclic compounds ( , areas A and E)–this contamination likely comes from lubricating oils used for the maintenance of the weapons.
The overall chemical composition of the scent trace endured conditions during the weapon’s firing. However, the elution area of fatty acid esters and less volatile compounds ( , area C) contained a smaller number of peaks, and the peaks were generally less intensive (as compared to standard scent samples). The elution pattern of the most frequent chemical compounds (such as ethyl esters, amines, and steroid derivates) and their summary are shown in and . Considering the nature of the compounds found (mostly C 16 and higher), it was clear that the volatility of the chemical compound would be a key factor in the study. The higher ethyl esters were previously reported in human scent samples . Squalene was also found in the cartridge samples, but its´ elution area overlapped with the “aliphatic cloud”. Thus, it was excluded from the further calculations. Nitrogen compounds as derivates of urea or nitrogen heterocycles were also reported in samples of human scent . Steroid chemical compounds have also been reported . In addition, steroids were marked as a ’scent precursors’ and products of the sebaceous glands along with squalene, fatty acids and their esters . Furthermore, compounds of the same chemical species were detected by gas chromatography in the samples of a fingerprints . Note that only a mass spectra comparison was done for the identification of the detected chemical compounds. Thus, the findings are discussed as a chemical group, not as the specific chemical compounds.
First, it was necessary to determine whether the remaining scent compounds on the fired cartridges included enough information for the distinction between the volunteers. The PCA model was built on ratios with the least standard deviation (SD) through the samples of each volunteer. The best model ( ) was based on N = 75 ratios. However, to reduce the peak noise and keep the relevant values, we considered ratios with more than 50% nonzero values ( T 2 = 50%) through all samples (Model_75_50) only. The experimental data set exhibited a non-normal distribution (normal skewness, but kurtosis differed significantly). Originally, five volunteers provided five samples for this experiment, but one volunteers´ set exhibited severe contamination and therefore it excluded from the dataset. Additionally, two samples from two volunteers (Vol1 and Vol4) had to be excluded due to the severe contamination as well. KNN and RF trained models had F 1 score equal to zero. The RNN model failed the classification and the F 1 score was equal to 0.5. Similarity analysis of fired cartridge samples. One random sample from each volunteer was chosen as the ’unknown’ sample, and the rest of the samples (’standards’) were compared to that ’unknown’ sample. Every classification step was successful with the four approaches (CS, PC, SC, and KT) (see ).
Five volunteers produced 11×5 samples in total, 10 standard scent samples, and one projectile sample. However, three standard samples (2 × Vol 2 and 1 × Vol 5) were excluded from further analyses due to low intensities (due to sampling or handling error, the measured chromatograms were almost blank). The initial PCA model was constructed on the ratios that were considered in Model 75_50 (Section Model-based analysis of fired cartridge samples). PCA analysis of standard samples based on this model failed to reveal differences between volunteers. Another model was built, and multiple least SD ratio setups were tested. The best discrimination power was achieved with N = 50 ratios (for each volunteer) and with the T 2 set to 25% across all inspected samples as demonstrated in . The experimental data set exhibited a non-normal distribution (skewness and kurtosis differed significantly from the normal distribution). The best performing KNN model reached F 1 score equal to 0.85. RNN model best F 1 score was equal to 0.60. The RF performed with the F 1 score equal to 1. The confusion matrix of the RF model is shown in the . The results and the model parameters are supplied in the ( ). For the completion of the data inspection, all models were retrained on the reduced data (first three principal components). RF model retained F 1 score equal to 1. F 1 score improved for both NN models: KNN scored 0.92 and RNN scored 0.85 in terms of F 1 score. The similarity analyses were carried out on the non-reduced (original) data, thus, the performance of the models when applied on original data was more significant for the proof-of-concept tests.
The Spearman’s correlation coefficient was the best performing correlation from all tested approaches (for the full representation of the results see and ). For Volunteers 1, 3, and 5 all their standard scent samples had higher correlation coefficient compared to the fired cartridge sample than the standard samples of the other volunteers. For Volunteer 2, one standard sample a had lower correlation coefficient compared to the fired cartridge sample than the standard samples of another Volunteer. The same case occurred for Volunteer 4. Cosine similarity was achieved at a level similar to that of SC. It was correct in all cases for Volunteers 1, 4, and 5, and made one error for Volunteers 2 and 3. Kendall’s tau reached the top score for the same volunteers as SC did, and similarly, made one error in the case of Volunteer 2. However, KT made two errors when comparing Volunteer 4’s samples. Pearson’s correlation reached the highest score only in two cases, Volunteer 1 and 5. PC made one error in comparing the samples of Volunteer 3 and 4, and two errors for Volunteer 2. The reason why SC and KT comparisons achieved better results is probably that, unlike PC, these correlations are more suitable for data with a non-normal distribution.
For the final comparison, a group of ten volunteers different from the group from previous experiments was asked to provide two samples–one collected on the glass beads (’standard sample’), and one collected on the projectile (’fired cartridge sample’). The samples were treated as stated in section Sampling procedure and sample collection. The structure of an experiment was designed to simulate the same procedure as that performed by specially trained dogs . Each shot cartridge sample was compared to ten standard samples–a comparison row . For every comparison row , the algorithm was supposed to detect and mark the standard sample that was most similar to the fired cartridge sample. The 50_75 model (Section Model-based analysis of standard samples) method was applied to a comparison process. However, the classification power of the model failed and was successful in less than 50% of the classification attempts. The main reason might be the fact that the model was built on standard scent sample rather than on scent residue samples. This probably led to a selection of peak area rations significant for distinguishing between standard scent samples, not shot cartridge ones. Model 75_50 (Section Model-based analysis of fired cartridge samples) performed with the same results (successful in less than 50% of the classification attempts) for Pearson’s correlation, cosine similarity, and Kendall’s tau. However, the Spearman’s correlation reached a 100% success rate (see ) because it was built on the significant peak area ratios detected on the shot cartridges. However, the standard scent sample of Vol 4 had a higher similarity with the fired cartridge sample of Vol 6. The relevant ratios are listed in . To inspect if the top two correlations are significantly different, both results were transformed to the z-scores and the test for the significance of the difference between them was calculated with p-value set to 0.05 (see ).
The scent samples collected from the fired cartridges indicated an analogous chemical composition to the scent samples undamaged by shooting collected on the glass beads. The results of the qualitative analysis of scent residues found on the cartridges suggested the sebaceous gland products and their derivatives as the compounds of the main interest for the rest of the conducted experiments. PCA models were developed based on the peak area ratios of the sampled volunteers (Section Model-based analysis of fired cartridge samples). The significant peak area ratios for the comparison identification were selected (Section Data processing) and used for the identification experiments. To prove the concept of presented workflow, the basic classification algorithms such as K-nearest neighbor and Random Forest were trained on the input data. The most successful model was RF with F 1 score equal to 1. As the algorithms were successful in a term of classification, the next step was the similarity comparison. The most suitable similarity comparison was the Spearman´s correlation (Section Similarity analysis of fired cartridge samples versus standard scent samples). The final comparison mimicking the sample comparison performed by the specially trained dogs (Section Fired cartridge sample versus standard scent sample comparison) was 100% successful. If the experiment considered standard scent samples as unknown samples, the method would make one error, misclassifying Volunteer 4 as Volunteer 6. The difference between two top correlations was statistically significant in four out of ten cases. The objective of this study was to point out the forensic potential of scent evidence found on the fired cartridges for the scent identification. It is important to point out that all steps of the study were conducted under laboratory conditions that did not involve real-world contaminations such as soil contamination, scents of different origin, or any other environmental contaminations; every step of this experiment was designed to reduce the risk of contamination as much as possible, and thus answer the question if there is a possibility to discover patterns in the residues of human scent after being exposed to gunshot. The biggest limitation of this kind of study so far is the available instrumentation. Even with an advanced approach such as GC×GC, the number of chemical compounds below the level of detection might be in the tens or hundreds. With improved sensitivity, the additional higher esters of the fatty acids could be quantified and involved in similarity experiments, which could lead to more convincing results.
S1 File Detailed results of the KNN and RF approaches with links to developed scripts and mathematical formulas used for the calculations. (PDF) Click here for additional data file. S2 File Official wording of the written consent collected from participating volunteers. (PDF) Click here for additional data file. S3 File Original data used as input for the study along with data results (see file ). (RAR) Click here for additional data file. S1 Fig Visual comparison of chromatograms. A = area of saturated and unsaturated hydrocarbons (the cluster most likely originated from gun lubricants); B = Squalen; C = Esters of higher fatty acids; D = Cholesterol; E = Ethyl esters of carboxylic acids; F = Heterocycles, amides, and more polar compounds in general. (TIF) Click here for additional data file. S1 Table Summary of the most abundant chemical compounds found in samples (26) collected from fired cartridges. (XLSX) Click here for additional data file. S2 Table List of relevant ratios for Spearman’s correlation (The model 75_50; Section 3.3.2). (XLSX) Click here for additional data file. S3 Table Overall results of similarity analyses of fired cartridge samples versus standard scent samples. (XLSX) Click here for additional data file. S1 Graphical abstract (TIF) Click here for additional data file.
|
Reporting of measures of accuracy in systematic reviews of diagnostic literature | 2500e913-c41d-420b-95b9-940dd12a8c04 | 100326 | Pathology[mh] | The manner in which accuracy of clinical tests is mathematically summarised in the biomedical literature has important implications for clinicians. Appropriate accuracy measures would be expected to sensibly convey the meaning of the study results with scientifically robust statistics without exaggerating or underestimating the clinical significance of the findings. Lack of use of appropriate measures may lead authors of primary accuracy studies to draw biased conclusions. In systematic reviews of test accuracy literature, there are many ways of synthesising results from several studies, not all of which are considered to be scientifically robust. For example, measures such as sensitivity and specificity commonly used in primary studies are not considered suitable for pooling separately in meta-analysis. Variations in reporting of summary accuracy and use of inappropriate summary statistics may increase the risk of misinterpretation of clinical value of tests. A recent study evaluated a small sample of meta-analytical reviews of screening tests to demonstrate the variety of approaches used to quantitatively summarise accuracy results. This study confined itself to a limited Medline search. It exclusively examined meta-analytical studies so reviews not using quantitative synthesis were excluded. It did not look at accuracy measures used to report results of primary studies separately from those used for meta-analyses. In order to address these issues, we undertook a comprehensive search to survey systematic reviews (with and without meta-analysis) of test accuracy literature to assess the measures used for reporting results of included primary studies as well as their quantitative synthesis.
We manually searched for relevant reviews in the Database of Abstracts of Reviews of Effectiveness (DARE). In order to limit the impact of human error inherent in manual searching, we complemented it with electronic searching. DARE was searched electronically with word variants of relevant terms (diagnostic, screening, test, likelihood ratio, sensitivity, specificity, positive and negative predictive value) combined using OR. From 1994 to 2000 DARE has identified 1897 reviews of different types by regular electronic searching of several bibliographic databases, hand searching of key major medical journals, and by scanning grey literature (search strategy and selection criteria can be found at ). The structured abstracts of these reviews were screened independently by the authors to identity systematic reviews of test accuracy. The full texts were obtained of those abstracts judged to be potentially relevant. Reviews addressing test development and diagnostic effectiveness or cost effectiveness were excluded. Any disagreements about review selection were resolved by consensus. Information from each of the selected reviews was extracted for the measures of test accuracy used to report the results of the primary studies included in the review. If a meta-analysis was conducted, information was also extracted for the summary accuracy measures. The various accuracy measures are shown in Table . We sought the following in the primary studies: sensitivity or specificity, predictive values, likelihood ratios and diagnostic odds ratio. For meta-analysis, we sought the summary measures pooling the above results and summary receiver operating characteristics (ROC) plot or values. All extracted data were double-checked. We divided the reviews into two groups arbitrarily according to time of publication; one group covering the period 1994–97 (50 reviews) and another covering 1998–2000 (40 reviews). This allowed us to assess whether there were any significant differences in measures being used to report test accuracy results among reviews published earlier and those published later. As the approaches to summarising results are not mutually exclusive, we evaluated and reported the most commonly used measures and their most common combinations. We used chi-squared statistical test for comparison of differences between proportions.
Of the abstracts available in DARE, 150 were considered to be potentially relevant. Excluding reviews that addressed test development and diagnostic effectiveness or cost, 90 reviews of test accuracy were left for inclusion in our survey. There were 45 reviews of dichotomous test results, 42 reviews of continuous results dichotomised by the original authors, and 3 reviews that contained both result types. Meta-analysis was used in 60/90 (67 %) reviews, 50 in 1994–97 and 40 in 1998–2000. (See : BMC_IncludedRefList_04032002 for a complete listing of the 90 reviews included in our study). As shown in Table , sensitivity or specificity was used for reporting the results of primary studies in 65/90 (72%) reviews, predictive values in 26/90 (28%), and likelihood ratios in 20/90 (22%). For meta-analysis, independently pooled sensitivity or specificity was used in 35/60 (58%) reviews, pooled predictive values in 11/60 (18%), pooled likelihood ratios in 13/60 (22%), and pooled diagnostic odds ratio in 5/60 (8%). Summary ROC was used in 44/60 (73%) of the meta-analyses. There were no significant differences between reviews published earlier and those published later as shown in Table .
Our study showed that sensitivity and specificity remain in frequent use, both for primary studies and for meta-analyses over the time period surveyed. Sensitivity and specificity are considered inappropriate for meta-analyses, as they do not behave independently when they are pooled from various primary studies to generate separate averages. In our survey, separate pooling of sensitivities or specificity was used frequently in meta-analyses where summary ROC would have been more appropriate. [ - ]. Our findings about reporting of summary accuracy measures in meta-analyses are different to those reported previously. We found a higher rate of use of summary ROC, though use of independent summaries of sensitivity, specificity and predictive values were similar. These differences may be due to differences in searching strategies (databases and time frames) and selection criteria. Our search was more recent and comprehensive, using DARE, which has covered seven different databases (Medline, CINAHL, BIOSIS, Allied and Alternative Medicine, ERIC, Current Contents clinical medicine and PsycLIT), and hand-searched 68 peer-reviewed journals and publications from 33 health technology assessment centres around the world since February 1994. Moreover, as we did not restrict our selection to meta-analytical reviews only, we were able to examine reviews summarising accuracy results of primary studies without quantitative synthesis, which constituted 33% (30/90) of our sample. Therefore, compared to the previous publication on this topic, our survey provided a broader and more up-to-date overview of the state of reporting of accuracy measure in test accuracy reviews.
The use of inappropriate accuracy measures has the potential to bias judgement about the value of tests. Of the various approaches to reporting accuracy of dichotomous test results, likelihood ratios are considered to be more clinically powerful than sensitivities or specificities. Crucially, it has been empirically shown that authors of primary studies may overstate the value of tests in the absence of likelihood ratios. There is also evidence that readers themselves may misinterpret test accuracy measures following publication. It is conceivable that the problem of inconsistent usage of test accuracy measures in published reviews, as found in our survey, may contribute to misinterpretation by clinical readership. The reason for variation in reported accuracy measures may, in part, be attributed to a lack of consensus regarding the best ways to summarise test results. It is worth noting that despite authoritative publications about appropriate summary accuracy measures in the past,[ , , ] (we have only quoted a few references) inconsistent and inappropriate use of summary measures has remained prevalent in the period 1994–2000. Our paper highlights the need for consensus to support change in this field of research.
None declared
The pre-publication history for this paper can be accessed here:
Additional file Reference list of the 90 reviews included in the survey BMC_IncludedRefList_04032002
|
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 |
null | f98c0286-86f0-485d-b948-dc78b53f72bc | 10033568 | Microbiology[mh] | Legionella are Gram-negative bacteria that can be found in natural (Atkinson et al., ; Zhan et al., ) and artificial aquatic environments (Prussin et al., ; Volker et al., ). The genus includes 64 species and more than 70 different serogroups (LPSN, ). Legionella are responsible for legionellosis, an infectious disease acquired by inhalation or aspiration of aerosols from contaminated water (De Giglio et al., ). The presence of amoebas, biofilms, and stagnant water and a temperature between 37 and 42 °C promote the growth of this microorganism (Van der Koiji et al., ). However, it is able to adapt to a wide range of water parameters, including temperatures of 5.7–63 °C, pH values of 5.5–8.1, and oxygen concentrations of 0.3–8.2 mg/L (Fliermans, ; Schwake et al., ; De Giglio et al., ). Recently, the new European Drinking Water Directive (Directive EU ) has inserted Legionella as mandatory among microbiological parameters to detect in water intended for human consumption for the risk assessment of domestic distribution systems in health and community facilities. In Europe, notification rates of Legionnaires’ disease vary from fewer than 0.5 cases to 5.7 cases per 100,000 of the population (ECDC, ). Although over 80% of human cases are caused by L. pneumophila ( Lpn ) serogroup 1 ( Lpn 1), the real number of Lpn non-serogroup 1 and Legionella non- pneumophila ( L-np ) cases (e.g., L. micdadei , L. bozemanii , L. longbeachae , L. dumofii , L. feeleii , and L. anisa ) is poorly documented (Beauté et al., ; Craun et al., ; ESGLI, ; Girolamini et al., ; Head et al., ; Vaccaro et al., ; Yang et al., ). In particular, L. anisa is the most commonly detected L-np in aquatic environments, and rarely causes infections in humans (Doebelling et al., ; Akermi et al., ). L. bozemanii and L. micdadei which are not frequently found in the environment are the most common causes of culture-verified L-np infections (Doebelling et al., ; Svarrer & Uldum., ; Neiderud et al., ; Miller et al., ). The risk of Legionella transmission with severe outcomes has been found to be affected by the complexity of hospital water systems and the vulnerability of hospitalized patients (De Giglio et al., ). Two thousand seven hundred and twenty-six cases of legionellosis were reported in Italy in 2021 with an incidence rate of 46 cases per 1 million of which 3.7% was of nosocomial origin (Istituto Superiore di Sanità, ). When it was possible to use culture-based methods, Lpn was identified in 100% of cases by Istituto Superiore di Sanità ( ). However, data describing the distribution of L-np in water systems of healthcare facilities are scarce (Arrigo et al., ; Girolamini et al., , , ; Mazzotta et al., ; Napoli et al., ). To our knowledge, there have been no reports of Italian legionellosis cases of nosocomial origin associated with L-np . However, it is likely that cases due to L-np are underestimated because the diagnostic methods commonly used in routine microbiological investigations are not sufficient for identification of species other than Lpn . Although culture-based method is considered the “gold standard” for microbiological surveillance of Legionella , it has some limitations in the identification of L-np (Scaturro et al., ). In recent years, many laboratories use a rapid identification method through the analysis of ribosomal protein pattern based on array-assisted laser desorption ionization mass spectrometry (MALDI-TOF MS) (Dilger et al., ; Pascale et al., ). Additionally, various genotyping techniques such as total genomic DNA analysis and specific ( mip gene) and internal gene sequencing are available for identification of Legionella (Svarrer and Uldum ). However, the use of different molecular techniques (e.g., pulsed-field gel electrophoresis (PFGE), RAPD, Rep-PCR, and SAU-PCR) is required to assess whether clinical patient isolates correspond to environmental isolates (Linee Guida per la prevenzione e il Controllo della Legionellosi, ; Haroon et al., ; Corich et al., ; Yamamoto et al., ). In this study, L-np strains isolated during environmental surveillance conducted for 13 months in four pavilions of a large hospital in Apulia region (southern Italy) were identified using traditional methods (i.e., culture and latex agglutination) and spectrometry (MALDI-TOF MS) in conjunction with mip -gene sequencing. The aim of the study was to genetically correlate strains isolated from the water networks of the different hospital pavilions using traditional molecular methods as PFGE compared with alternative techniques (RAPD, Rep-PCR, and SAU-PCR) not usually applied to L-np strains.
Our Environmental and Food Hygiene Laboratory (ACCREDIA n.1683), which is recognized as a Regional Reference Center in Apulia (Southern Italy), has conducted clinical and environmental surveillance of legionellosis in nosocomial and community facilities since 2001. In January 2020, systematic monitoring of the water network of a large Apulian public hospital was started as part of the implementation program of a Water Safety Plan. The surveillance plan covered 33 separate buildings and provided for the quarterly microbiological control of 50% of the water supply points uniformly distributed in each pavilion. Sampling and culture-based investigation From 1 March 2021 to 30 March 2022, 346 water samples (1 L) were collected from taps of water system of four pavilions: pI (81), pII (109), pIII (128), and pIV (28). In detail, 277 samples were collected from the hot water systems and 69 from the cold water networks when it was not possible to obtain hot water samples (e.g., due to exhaustion of hot water in the boilers). The temperature was recorded for each water sample. According to Italian Guideline (Linee guida per la Prevenzione e il Controllo della Legionellosi, ), the collected samples were stored in sterile dark glass bottles containing sodium thiosulfate pentahydrate (0.01%, w/v) to neutralize chloride present in water. Samples were then transported, by an isothermal container, to the laboratory at room temperature (19.2 °C; range 18.5–24.2 °C). According to ISO 11731:2017 (ISO ), each sample was first filtered through isopore polycarbonate membranes (47 mm in diameter with a pore size of 0.22 µm) (Millipore Corporation, Bedford, MA, USA) then resuspended in 10 mL of the non-filtered water sample. After vortexing, an aliquot of each sample (200 µL) was seeded onto selective culture medium GVPC plates added with glycine, vancomycin, polymyxin, and cycloheximide B (Liofilchem Srl, Teramo, Italy) followed by incubation at 36 ± 2 °C for 7–10 days in a humid environment (to prevent desiccation of the plates) (ISO ). Presumptive colonies (at least three different colonies for each plate) of Legionella were then inoculated in buffered charcoal yeast extract (BCYE) agar plates (BioMérieux, Marcy l'Etoile, France) with and without l -cysteine. The identification of the colonies of Legionella spp. grown only on BCYE agar plates added with cysteine was performed using a latex agglutination test with polyvalent (Biolife Italiana Srl, Milan, Italy) and monovalent antisera (Biogenetics Srl, Tokyo, Japan) (Linee guida per la Prevenzione e il Controllo della Legionellosi, ). In particular, regarding Legionella species group, the polyvalent antisera provided by the manufacturer (Biolife Italiana Srl, Milan, Italy) identify only few species implicated in clinical cases: L. anisa , L. micdadei , L. bozemanii 1 and 2 , L. gormanii , L. longbeachae 1 and 2 , L. dumoffii , and L. jordanis . Moreover, the monovalent sera commercially available allow to identify all Legionella species mentioned above except L. anisa , L. longbeachae 1 and 2 , and L. jordanis (Biogenetics Srl, Tokyo, Japan). All the strains identified as Legionella spp. were viewed under a long-wave ultraviolet light (360 ± 20 nm) emitted by a wood lamp to evaluate the emission of autofluorescence due to the presence of an intracellular pigment. The results of Legionella contamination in water samples were expressed as colony-forming units per liter (CFU/L), and the detection limit was 50 CFU/L. Identification by MALDI-TOF MS analysis and mip-gene sequencing Legionella species strains were identified by MALDI-TOF using the Vitek MS system (BioMérieux, Italy) at level serogroup according to instruction of manufacturer. Also the mip gene sequence analysis was performed as reported by Federici et al. ( ). Briefly, thermal shock was used to extract genomic DNA from colonies of Legionella grown on BCYE medium, followed by PCR amplification of the mip gene. PCR reactions were set up on a total volume of 20 µL containing FIREPol 5 × Master Mix (Solis BioDyne, Tartu, Estonia), 0.4 µM of Legmip-f and Legmip-r primers (Ratcliff et al., ), and 4 µl of template DNA. PCR amplification protocol consisted of 34 cycles at 95 °C for 30 s of denaturation, 58 °C for 1 min of annealing, and 72 °C for 1 min of extension. At the end, amplicons were purified with the EuroSAP PCR Enzymatic Clean-up kit (Euroclone SpA, Pero, MI, IT) and sequenced by the Sanger method (Eurofins Genomics, Ebersberg; Germany). Sequencing results were analyzed and processed with Geneious Prime 2022.1.1 (Biomatters Inc., San Diego, CA, USA) running the BLAST (Basic Local Alignment Search Tool) tool to find similar mip genes belonging to Legionella reference strains in the GenBank database. A phylogenetic tree was constructed by the unweighted pair group method with arithmetic mean (UPGMA) method using Jukes-Cantor as a genetic distance model after aligning the obtained mip gene sequences with those of the culture collections strains downloaded from GenBank. Sequence alignment was performed using the multiple sequence comparison by log-expectation (MUSCLE) program with default settings (Edgar, ). Genetic correlation of Legionella non-pneumophila strains Only 14 (C, F, G1, G2, G3, G4, H, I, L, M, P, R, S, and T) strains isolated from the water networks of the four different pavilions (pI, pII, pIII, and pIV) were possible to analyze establishing a possible genetic correlation. Strains L, M, P, R, S, and T were isolated from pavilion pI; C from pavilion pII; H and I from pavilion pIII; and F, G1, G2, G3, and G4 from pavilion pIV. Pulsed-field gel electrophoresis PFGE of the 14 strains was performed using a modification of the CDC PulseNet standardized PFGE protocol (Ribot et al., ; Sabrià et al., ). Briefly, isolates were grown on BCYE agar plates at 37 °C for 48 h, after which they were suspended in cell suspension buffer (10 mM Tris, 1 M NaCl, pH 8.0). The cell suspensions were then adjusted to an optical density of 0.5–0.6 at a wavelength of 610 nm and mixed with equal volumes of 1.0% pulsed-field certified agarose (BioRad, Milan, Italy) and 10 µL of proteinase K (20 mg/mL). Next, the agarose plugs were transferred into cell lysis buffer (0.5 M EDTA, pH 8.0, 1% sarcosine, and 200 µg/ml of proteinase K) and incubated at 50 °C overnight. DNA embedded in the plugs was digested with 40 U of AscI or SfiI (New England Biolabs, Schwahlbach, Germany) at 37 °C for 4 h. Plugs of Salmonella braenderup strain H9812 were digested with 40 U XbaI (TaKaRa Bio, Dalian, China) and used in each gel as universal size standards. Fragments of DNA were separated in a CHEF DR III System (BioRad Laboratories, Richmond, CA, USA) with a constant voltage of 6 V cm −1 , an included angle of 120°, and increasing pulse times (5.6–50 s) at 14 °C for 21 h. All PFGE profiles were digitalized and the phylogenetic relationship was assessed through the fingerprinting software GelJ (Heras et al., ). The phylogenetic relationship was represented through a phylogenetic tree obtained using the Dice coefficient with clustering by the unweighted pair-group method with arithmetic mean (UPGMA). A 2% tolerance in band position differences was applied ( Martínez-Puchol et al., ). RAPD, Rep-PCR, and SAU-PCR analyses DNA extraction for PCR-based genetic fingerprint of the strains was performed using the GenElute™ Bacterial Genomic DNA Kit (Sigma-Aldrich, Milan, Italy) following the manufacturer’s instruction. The initial step was modified as follows: isolates were collected from Legionella agar plates (Sigma-Aldrich, Milan, Italy) using a 10-µL sterile loop, then resuspended thoroughly in 180 μL of lysis solution and the protocol was regularly followed. DNA was then quantified and standardized at 50 ng/µL (Nanodrop One system, Thermo Scientific, Marietta, OH, USA). PCRs were performed using the amplification condition using a C1000 Touch Thermal Cycler (BioRad, Milan, Italy). The following primers were used for the reactions: M13 (5′-GAG GGT GGC GGT TCT-30′) (Huey & Hall, ), (GTG) 5 (5′-GTGGTGGTGGTGGTG-3′) (Gevers et al., ), and SAG1 (5′-CCGCCGCGATCAG-3′) for RAPD, Rep-PCR, and SAU-PCR, respectively. Conditions for PCR reaction, gel electrophoresis of the amplicons, and cluster analysis were performed according to Iacumin et al. ( ). The 14 strains of L-np species were subjected to fingerprint analysis at least three times to confirm the reproducibility of the obtained profiles. Statistical analysis The data were presented as numbers or percentages for the categorical variables. To determine if the distribution of load (cfu /L) of Legionella spp. was normal, a Shapiro–Wilk normality test was performed. Chi-square or Fisher’s exact test was performed to compare two or more mutually exclusive proportions or percentages in groups. The Kruskal–Wallis rank sum test and the Wilcoxon rank sum test were used to determine if water temperature influenced the presence of Legionella spp. and L-np species. To accomplish this, the water samples were grouped into negative and positive for L-np and L. pneumophila in the hot and cold water networks. R version 3.6.3 (The R Project for Statistical Computing, Vienna, Austria) was used for analysis. A p-value < 0.05 was considered to indicate statistical significance.
From 1 March 2021 to 30 March 2022, 346 water samples (1 L) were collected from taps of water system of four pavilions: pI (81), pII (109), pIII (128), and pIV (28). In detail, 277 samples were collected from the hot water systems and 69 from the cold water networks when it was not possible to obtain hot water samples (e.g., due to exhaustion of hot water in the boilers). The temperature was recorded for each water sample. According to Italian Guideline (Linee guida per la Prevenzione e il Controllo della Legionellosi, ), the collected samples were stored in sterile dark glass bottles containing sodium thiosulfate pentahydrate (0.01%, w/v) to neutralize chloride present in water. Samples were then transported, by an isothermal container, to the laboratory at room temperature (19.2 °C; range 18.5–24.2 °C). According to ISO 11731:2017 (ISO ), each sample was first filtered through isopore polycarbonate membranes (47 mm in diameter with a pore size of 0.22 µm) (Millipore Corporation, Bedford, MA, USA) then resuspended in 10 mL of the non-filtered water sample. After vortexing, an aliquot of each sample (200 µL) was seeded onto selective culture medium GVPC plates added with glycine, vancomycin, polymyxin, and cycloheximide B (Liofilchem Srl, Teramo, Italy) followed by incubation at 36 ± 2 °C for 7–10 days in a humid environment (to prevent desiccation of the plates) (ISO ). Presumptive colonies (at least three different colonies for each plate) of Legionella were then inoculated in buffered charcoal yeast extract (BCYE) agar plates (BioMérieux, Marcy l'Etoile, France) with and without l -cysteine. The identification of the colonies of Legionella spp. grown only on BCYE agar plates added with cysteine was performed using a latex agglutination test with polyvalent (Biolife Italiana Srl, Milan, Italy) and monovalent antisera (Biogenetics Srl, Tokyo, Japan) (Linee guida per la Prevenzione e il Controllo della Legionellosi, ). In particular, regarding Legionella species group, the polyvalent antisera provided by the manufacturer (Biolife Italiana Srl, Milan, Italy) identify only few species implicated in clinical cases: L. anisa , L. micdadei , L. bozemanii 1 and 2 , L. gormanii , L. longbeachae 1 and 2 , L. dumoffii , and L. jordanis . Moreover, the monovalent sera commercially available allow to identify all Legionella species mentioned above except L. anisa , L. longbeachae 1 and 2 , and L. jordanis (Biogenetics Srl, Tokyo, Japan). All the strains identified as Legionella spp. were viewed under a long-wave ultraviolet light (360 ± 20 nm) emitted by a wood lamp to evaluate the emission of autofluorescence due to the presence of an intracellular pigment. The results of Legionella contamination in water samples were expressed as colony-forming units per liter (CFU/L), and the detection limit was 50 CFU/L.
Legionella species strains were identified by MALDI-TOF using the Vitek MS system (BioMérieux, Italy) at level serogroup according to instruction of manufacturer. Also the mip gene sequence analysis was performed as reported by Federici et al. ( ). Briefly, thermal shock was used to extract genomic DNA from colonies of Legionella grown on BCYE medium, followed by PCR amplification of the mip gene. PCR reactions were set up on a total volume of 20 µL containing FIREPol 5 × Master Mix (Solis BioDyne, Tartu, Estonia), 0.4 µM of Legmip-f and Legmip-r primers (Ratcliff et al., ), and 4 µl of template DNA. PCR amplification protocol consisted of 34 cycles at 95 °C for 30 s of denaturation, 58 °C for 1 min of annealing, and 72 °C for 1 min of extension. At the end, amplicons were purified with the EuroSAP PCR Enzymatic Clean-up kit (Euroclone SpA, Pero, MI, IT) and sequenced by the Sanger method (Eurofins Genomics, Ebersberg; Germany). Sequencing results were analyzed and processed with Geneious Prime 2022.1.1 (Biomatters Inc., San Diego, CA, USA) running the BLAST (Basic Local Alignment Search Tool) tool to find similar mip genes belonging to Legionella reference strains in the GenBank database. A phylogenetic tree was constructed by the unweighted pair group method with arithmetic mean (UPGMA) method using Jukes-Cantor as a genetic distance model after aligning the obtained mip gene sequences with those of the culture collections strains downloaded from GenBank. Sequence alignment was performed using the multiple sequence comparison by log-expectation (MUSCLE) program with default settings (Edgar, ).
Only 14 (C, F, G1, G2, G3, G4, H, I, L, M, P, R, S, and T) strains isolated from the water networks of the four different pavilions (pI, pII, pIII, and pIV) were possible to analyze establishing a possible genetic correlation. Strains L, M, P, R, S, and T were isolated from pavilion pI; C from pavilion pII; H and I from pavilion pIII; and F, G1, G2, G3, and G4 from pavilion pIV.
PFGE of the 14 strains was performed using a modification of the CDC PulseNet standardized PFGE protocol (Ribot et al., ; Sabrià et al., ). Briefly, isolates were grown on BCYE agar plates at 37 °C for 48 h, after which they were suspended in cell suspension buffer (10 mM Tris, 1 M NaCl, pH 8.0). The cell suspensions were then adjusted to an optical density of 0.5–0.6 at a wavelength of 610 nm and mixed with equal volumes of 1.0% pulsed-field certified agarose (BioRad, Milan, Italy) and 10 µL of proteinase K (20 mg/mL). Next, the agarose plugs were transferred into cell lysis buffer (0.5 M EDTA, pH 8.0, 1% sarcosine, and 200 µg/ml of proteinase K) and incubated at 50 °C overnight. DNA embedded in the plugs was digested with 40 U of AscI or SfiI (New England Biolabs, Schwahlbach, Germany) at 37 °C for 4 h. Plugs of Salmonella braenderup strain H9812 were digested with 40 U XbaI (TaKaRa Bio, Dalian, China) and used in each gel as universal size standards. Fragments of DNA were separated in a CHEF DR III System (BioRad Laboratories, Richmond, CA, USA) with a constant voltage of 6 V cm −1 , an included angle of 120°, and increasing pulse times (5.6–50 s) at 14 °C for 21 h. All PFGE profiles were digitalized and the phylogenetic relationship was assessed through the fingerprinting software GelJ (Heras et al., ). The phylogenetic relationship was represented through a phylogenetic tree obtained using the Dice coefficient with clustering by the unweighted pair-group method with arithmetic mean (UPGMA). A 2% tolerance in band position differences was applied ( Martínez-Puchol et al., ).
DNA extraction for PCR-based genetic fingerprint of the strains was performed using the GenElute™ Bacterial Genomic DNA Kit (Sigma-Aldrich, Milan, Italy) following the manufacturer’s instruction. The initial step was modified as follows: isolates were collected from Legionella agar plates (Sigma-Aldrich, Milan, Italy) using a 10-µL sterile loop, then resuspended thoroughly in 180 μL of lysis solution and the protocol was regularly followed. DNA was then quantified and standardized at 50 ng/µL (Nanodrop One system, Thermo Scientific, Marietta, OH, USA). PCRs were performed using the amplification condition using a C1000 Touch Thermal Cycler (BioRad, Milan, Italy). The following primers were used for the reactions: M13 (5′-GAG GGT GGC GGT TCT-30′) (Huey & Hall, ), (GTG) 5 (5′-GTGGTGGTGGTGGTG-3′) (Gevers et al., ), and SAG1 (5′-CCGCCGCGATCAG-3′) for RAPD, Rep-PCR, and SAU-PCR, respectively. Conditions for PCR reaction, gel electrophoresis of the amplicons, and cluster analysis were performed according to Iacumin et al. ( ). The 14 strains of L-np species were subjected to fingerprint analysis at least three times to confirm the reproducibility of the obtained profiles.
The data were presented as numbers or percentages for the categorical variables. To determine if the distribution of load (cfu /L) of Legionella spp. was normal, a Shapiro–Wilk normality test was performed. Chi-square or Fisher’s exact test was performed to compare two or more mutually exclusive proportions or percentages in groups. The Kruskal–Wallis rank sum test and the Wilcoxon rank sum test were used to determine if water temperature influenced the presence of Legionella spp. and L-np species. To accomplish this, the water samples were grouped into negative and positive for L-np and L. pneumophila in the hot and cold water networks. R version 3.6.3 (The R Project for Statistical Computing, Vienna, Austria) was used for analysis. A p-value < 0.05 was considered to indicate statistical significance.
Identification by agglutination test, MALDI-TOF MS technique, and mip -gene sequencing Overall, 86/346 (24.9%) water samples were positive for Legionella spp. Analysis of positive samples using the polyvalent antiserum revealed that 45/86 (52.3%) samples were positive solely for Lpn 1 (median value = 1,200 cfu/L; range 50–83,000 cfu/L), 18/86 (20.9%) for Lpn 2–15 (median value = 1,150 cfu/L; range 50–22,500 cfu/L), 6/86 (7.1%) for Lpn 1 + Lpn 2–15 (median value = 980 cfu/L; range 450–4,400 cfu/L), 15/86 (17.4%) solely for L-np (median value = 200 cfu/L; range 50–13,000 cfu/L), and 2/86 (2.3%) for L-np + Lpn 1 (median value = 905 cfu/L; range 410–1,400 cfu/L). All colonies of L-np showed blue-white autofluorescence under UV light at 365 nm and were identified by monovalent antiserum as Legionella bozemanii . Subsequent MALDI-TOF analysis and mip -gene sequencing indicated that they were all L. anisa (Fig. ). To assess the relationship between L. anisa isolates and the closest Legionella spp., a UPGMA-phylogenetic tree was built (Fig. ) (Pascale et al., ). All isolates grouped into a single clade with the L. anisa ATCC strain, indicating 100% homology both among isolates and with the reference sequence. However, the remaining reference sequences were located in other clades, confirming the phylogenetic distance between L. anisa isolates and L. bozemanii (95.19% sequence identity). Table summarizes the distribution of isolates in the four pavilions. In particular, it shows two water samples revealed the presence of L. anisa mixed with Lpn 1 in pII ( Lpn 1 200 cfu/L + L. anisa 210 cfu/L) and pIV ( Lpn 1 200 cfu/L + L. anisa 1,200 cfu/L). Temperature analysis The cold water network was found to have more Legionella contamination than the hot water network (39.1% vs 21.3%; χ 2 = 9.40, p-value = 0.002). The same result was observed for L. anisa (44.5% cold water vs 5.1% hot water; Fisher’s exact test, p-value < 0.001). Conversely, more Lpn were present in the hot water system than the cold water system (94.9% vs 48.1%, χ 2 = 22.68, p -value < 0.0001). No significant differences in the distribution of Lpn + L. anisa were observed between the hot and cold water systems (Fisher’s exact test, p-value = 0.09) (Table ). Table Distribution of Legionella in hot and cold water samples collected from water networks in the four investigated pavilions (March 2021–March 2022). As shown in Table , Legionella distribution differed significantly among temperature ranges. Specifically, L. anisa was more often isolated from hot water samples with lower median temperatures than from other samples (Kruskal-Wallis χ 2 =12.545, p-value =0.001), i.e., positive for Lpn (pairwise comparisons using Wilcoxon rank sum test, p-value =0.05) and negative samples (pairwise comparisons using Wilcoxon rank sum test, p-value =0.04). Correlation among Legionella species strains PFGE typing Figure shows the PFGE patterns after restriction with AscI . PFGE patterns differing by 1 or more bands were classified into single clusters. Eleven out of fourteen analyzed strains showed an indistinguishable PFGE profile and were clustered within the same group (termed A). Single strains assigned to groups B and C had > 90% similarity to group A and differed from the latter for 2 and 3 bands, respectively. Group D (also composed of a single strain) had 86% similarity to group A from which it differed for 3 bands. Analysis of PFGE profiles generated with Sfi I showed data comparable to that observed with Asc I. In particular, strains clustered in group A for Asc I showed an indistinguishable PFGE profile with Sfi I too. Likewise, single strains that composed groups B, C, and D with Asc I differed among them and were assigned to single groups. Clonal relatedness was determined by criteria for interpreting PFGE profiles published by Tenover and colleagues (Tenover et al., ). Strains of group A were found to be clonal for both restriction patterns), while strains of groups B and D (according to Tenover’s criteria and percentage of similarity) might be closely related to group prevailing (Tenover et al., ). RAPD, Rep-PCR, and SAU-PCR typing Three other genetic fingerprint techniques based on different genetic targets and with different degrees of stringency were applied to confirm the PFGE results and evaluate strain diversity. RAPD, Rep-PCR, and SAU-PCR effectively produced fingerprints for each individual isolate (Fig. ). Moreover, these techniques showed greater discriminative power than PFGE with restriction enzyme AscI, allowing confirmation that the isolates were individual strains and not clones. However, the use of all three techniques was critical in establishing that the strains were different, but genetically related. Calculation of similarities in the profiles of bands based on Pearson product-moment correlation coefficients allowed production of dendrograms using UPGMA clustering algorithms (Fig. ). The RAPD technique allowed discrimination of the strains according to the place of origin. In fact, cutting the cluster at a similarity of 90% resulted in the majority of the strains being grouped into two main clusters, B and C. Cluster B contained all of the strains isolated from pavilion pIV while cluster C included five strains isolated from pavilion pI and two from pavilion pIII. It should be noted that although strain H belonged to group C, it differed greatly from the others. This finding was confirmed by the SAU dendrogram. Using SAU-PCR technique, a higher degree of discrimination among strains emerged. Using a similarity cutoff of 70% resulted in the formation of two large clusters, A and B. Cluster A contained all of the strains isolated from pavilion pIV and a single outlier from pavilion pI (strain M). Although present within cluster A, strain M differed from the others strains within the cluster (82% similarity). This finding agreed with the PFGE results and clearly demonstrated that this strain was one of three that differed from the other strains. Cluster B contained all remaining strains. However, when a cutoff of 88% similarity was used, cluster B broke up into two clusters; namely, cluster C, which contained all strains from pavilion pI, and cluster D, which contained all of the strains isolated from pavilion pIII in one group and a strain isolated from pavilion pII. The Rep-PCR technique confirmed this genetic variability between strains, albeit with different clusters. Specifically, this method denoted a differentiation based on the site of isolation, although there were some strains that were more similar to other isolates at distant points.
mip -gene sequencing Overall, 86/346 (24.9%) water samples were positive for Legionella spp. Analysis of positive samples using the polyvalent antiserum revealed that 45/86 (52.3%) samples were positive solely for Lpn 1 (median value = 1,200 cfu/L; range 50–83,000 cfu/L), 18/86 (20.9%) for Lpn 2–15 (median value = 1,150 cfu/L; range 50–22,500 cfu/L), 6/86 (7.1%) for Lpn 1 + Lpn 2–15 (median value = 980 cfu/L; range 450–4,400 cfu/L), 15/86 (17.4%) solely for L-np (median value = 200 cfu/L; range 50–13,000 cfu/L), and 2/86 (2.3%) for L-np + Lpn 1 (median value = 905 cfu/L; range 410–1,400 cfu/L). All colonies of L-np showed blue-white autofluorescence under UV light at 365 nm and were identified by monovalent antiserum as Legionella bozemanii . Subsequent MALDI-TOF analysis and mip -gene sequencing indicated that they were all L. anisa (Fig. ). To assess the relationship between L. anisa isolates and the closest Legionella spp., a UPGMA-phylogenetic tree was built (Fig. ) (Pascale et al., ). All isolates grouped into a single clade with the L. anisa ATCC strain, indicating 100% homology both among isolates and with the reference sequence. However, the remaining reference sequences were located in other clades, confirming the phylogenetic distance between L. anisa isolates and L. bozemanii (95.19% sequence identity). Table summarizes the distribution of isolates in the four pavilions. In particular, it shows two water samples revealed the presence of L. anisa mixed with Lpn 1 in pII ( Lpn 1 200 cfu/L + L. anisa 210 cfu/L) and pIV ( Lpn 1 200 cfu/L + L. anisa 1,200 cfu/L).
The cold water network was found to have more Legionella contamination than the hot water network (39.1% vs 21.3%; χ 2 = 9.40, p-value = 0.002). The same result was observed for L. anisa (44.5% cold water vs 5.1% hot water; Fisher’s exact test, p-value < 0.001). Conversely, more Lpn were present in the hot water system than the cold water system (94.9% vs 48.1%, χ 2 = 22.68, p -value < 0.0001). No significant differences in the distribution of Lpn + L. anisa were observed between the hot and cold water systems (Fisher’s exact test, p-value = 0.09) (Table ). Table Distribution of Legionella in hot and cold water samples collected from water networks in the four investigated pavilions (March 2021–March 2022). As shown in Table , Legionella distribution differed significantly among temperature ranges. Specifically, L. anisa was more often isolated from hot water samples with lower median temperatures than from other samples (Kruskal-Wallis χ 2 =12.545, p-value =0.001), i.e., positive for Lpn (pairwise comparisons using Wilcoxon rank sum test, p-value =0.05) and negative samples (pairwise comparisons using Wilcoxon rank sum test, p-value =0.04).
PFGE typing Figure shows the PFGE patterns after restriction with AscI . PFGE patterns differing by 1 or more bands were classified into single clusters. Eleven out of fourteen analyzed strains showed an indistinguishable PFGE profile and were clustered within the same group (termed A). Single strains assigned to groups B and C had > 90% similarity to group A and differed from the latter for 2 and 3 bands, respectively. Group D (also composed of a single strain) had 86% similarity to group A from which it differed for 3 bands. Analysis of PFGE profiles generated with Sfi I showed data comparable to that observed with Asc I. In particular, strains clustered in group A for Asc I showed an indistinguishable PFGE profile with Sfi I too. Likewise, single strains that composed groups B, C, and D with Asc I differed among them and were assigned to single groups. Clonal relatedness was determined by criteria for interpreting PFGE profiles published by Tenover and colleagues (Tenover et al., ). Strains of group A were found to be clonal for both restriction patterns), while strains of groups B and D (according to Tenover’s criteria and percentage of similarity) might be closely related to group prevailing (Tenover et al., ).
Figure shows the PFGE patterns after restriction with AscI . PFGE patterns differing by 1 or more bands were classified into single clusters. Eleven out of fourteen analyzed strains showed an indistinguishable PFGE profile and were clustered within the same group (termed A). Single strains assigned to groups B and C had > 90% similarity to group A and differed from the latter for 2 and 3 bands, respectively. Group D (also composed of a single strain) had 86% similarity to group A from which it differed for 3 bands. Analysis of PFGE profiles generated with Sfi I showed data comparable to that observed with Asc I. In particular, strains clustered in group A for Asc I showed an indistinguishable PFGE profile with Sfi I too. Likewise, single strains that composed groups B, C, and D with Asc I differed among them and were assigned to single groups. Clonal relatedness was determined by criteria for interpreting PFGE profiles published by Tenover and colleagues (Tenover et al., ). Strains of group A were found to be clonal for both restriction patterns), while strains of groups B and D (according to Tenover’s criteria and percentage of similarity) might be closely related to group prevailing (Tenover et al., ).
Three other genetic fingerprint techniques based on different genetic targets and with different degrees of stringency were applied to confirm the PFGE results and evaluate strain diversity. RAPD, Rep-PCR, and SAU-PCR effectively produced fingerprints for each individual isolate (Fig. ). Moreover, these techniques showed greater discriminative power than PFGE with restriction enzyme AscI, allowing confirmation that the isolates were individual strains and not clones. However, the use of all three techniques was critical in establishing that the strains were different, but genetically related. Calculation of similarities in the profiles of bands based on Pearson product-moment correlation coefficients allowed production of dendrograms using UPGMA clustering algorithms (Fig. ). The RAPD technique allowed discrimination of the strains according to the place of origin. In fact, cutting the cluster at a similarity of 90% resulted in the majority of the strains being grouped into two main clusters, B and C. Cluster B contained all of the strains isolated from pavilion pIV while cluster C included five strains isolated from pavilion pI and two from pavilion pIII. It should be noted that although strain H belonged to group C, it differed greatly from the others. This finding was confirmed by the SAU dendrogram. Using SAU-PCR technique, a higher degree of discrimination among strains emerged. Using a similarity cutoff of 70% resulted in the formation of two large clusters, A and B. Cluster A contained all of the strains isolated from pavilion pIV and a single outlier from pavilion pI (strain M). Although present within cluster A, strain M differed from the others strains within the cluster (82% similarity). This finding agreed with the PFGE results and clearly demonstrated that this strain was one of three that differed from the other strains. Cluster B contained all remaining strains. However, when a cutoff of 88% similarity was used, cluster B broke up into two clusters; namely, cluster C, which contained all strains from pavilion pI, and cluster D, which contained all of the strains isolated from pavilion pIII in one group and a strain isolated from pavilion pII. The Rep-PCR technique confirmed this genetic variability between strains, albeit with different clusters. Specifically, this method denoted a differentiation based on the site of isolation, although there were some strains that were more similar to other isolates at distant points.
The best approach to prevent waterborne diseases is systematic environmental monitoring and accurate identification of strains circulating in water distribution systems coupled with implementation of a risk assessment plan. Over the years, this approach has improved in hospitals, especially for the control and prevention of legionellosis (Mazzotta et al., ). Culture-based investigations still represent the most used method for the isolation of Legionella strains, followed by serological agglutination tests with poly/monovalent antisera (Linee guida per la Prevenzione e il Controllo della Legionellosi, ; ISO ). However, while serological tests can identify Lpn and its serogroups, they cannot identify all of the L-np species. Particularly problematic characteristics of these tests are that the monovalent antiserum for L. anisa is currently not commercially available and anti- L. bozemanii antiserum shows cross-reaction between L. bozemanii and L. anisa (Tateyama, ). Over 13 months, we repeatedly isolated L-np , which were initially identified as L. bozemanii , by the slide agglutination test. Because these strains had not previously been isolated frequently in our hospital, we confirmed their identification by MALDI-TOF and mip -gene sequencing (Dilger et al., ; Ratcliff et al., ). This approach allowed elucidation of the preliminary misidentification of the strains, which resulted as L. anisa . Both species showed blue-white fluorescence when exposed to a wood lamp. Alternative methods to slide agglutination tests (i.e., molecular methods) are generally only used during epidemiological investigations, not in routine environmental surveillance (Ratcliff et al., ). The misleading identification of L. bozemani in place of L. anisa demonstrates the importance of correct identification of environmental strains, particularly those belonging to L-np species. In addition to ecological relevance, accurate phylogenetic characterization of Legionella contamination is essential to assessing its potential impact on the etiology of nosocomial pneumonia. Indeed, the importance of such information is widely recognized to acquire correct and complete epidemiological data and, consequently, design effective strategies to control legionellosis (Rota et al., ). Although the pathogenicity of L. anisa is considered low, it has been reported that Lpn serogroup (sg) 1 is not the only species or serogroup responsible for clinical cases and other species can cause human pathologies (Cross et al., ; Montagna et al., ). In several countries, including Italy, L. anisa has been often identified as one of the most abundant contaminating species (Federici et al., ; Mazzotta et al., ). Furthermore, L. anisa can mask Lpn contamination in the water supply (Orsini et al., ; van der Mee-Marquet et al., ). In our study, the concentration of L. anisa was higher than that of Lpn sg 1 when mixed cultures were detected. Our findings also confirmed that higher water temperatures affect the load of Lpn , with lower temperatures favoring the presence of L-np (e.g., L. anisa ) (Girolamini et al., ). This phenomenon suggests the need for new diagnostic approaches capable of widespread identification even of L-np . Currently, such identification can be performed by MALDI-TOF analysis and mip -gene sequencing (Dilger et al., ; Ratcliff et al., ) displaying a high concordance level (Pascale et al., ). Epidemiological investigations uncover different aspects of Legionella contamination because the goal of such studies is identification of the source of infection to enable its control (Mazzotta et al., ). PFGE is one of the molecular techniques indicated for comparison of the genomic profiles among Legionella strains of different origin by the National Guidelines (Linee Guida per la Prevenzione e il Controllo della Legionellosi, ). However, our results suggest that the use of other molecular techniques could be a useful tool that more accurately characterizes the identity of circulating Legionella strains, according to the results found in other studies (Haroon et al., ; Yamamoto et al., ; Yan et al., ). The unusual presence of L. anisa strains in the water supply of the four investigated pavilions indicated that understanding the genetic variability of the isolated strains was necessary to determine if the hospital was contaminated with one or more different L. anisa strains. PFGE, which is known as a laborious and time-consuming technique (Dilger et al., ), was initially used. This was followed by RAPD, rep-PCR, and SAU-PCR analysis, which are rapid, simple, reproducible, and inexpensive (Haroon et al., ). Although these techniques are not usually applied to L-np (Haroon et al., ; Corich et al., ; Yamamoto et al., ), they demonstrated a greater discriminative power than PFGE, allowing us to establish that not all of the isolated strains belonged to the same clone. In addition, genetic variability between the strains denoting a differentiation based on the site of origin was highlighted. The reason about genetic differences among the strains from different pavilions is still unclear. Diverse eras of pavilion construction, different entry points of the water network, water temperatures, disinfection systems, or pipe materials could be responsible for the genetic variability (Girolamini et al., ). The presence of pavilion-specific clusters indicates the rooting of strains that have become characteristic for particular isolation sites. This phenomenon may depend on the natural selection of strains adapted to the specific environmental conditions in which they have been isolated (Akermi et al., ). If extended to the entire hospital water network, our study could allow development of a map of Legionella distribution, which would enable investigation of variability in terms of strain diversity between environments. The study of Legionella belonging to L-np could improve knowledge regarding less-documented species in hospitals and other surveillance protocols (Mazzotta et al., ). Moreover, a regional map of Legionella will support control and prevention of Legionella and it will allow to follow the evolution of the strains over time which is affected by different factors such as mutations, resistance, and pathogenicity (Mazzotta et al., ). Further investigations are still necessary to investigate the role of molecular typing in the environmental surveillance of a water network contaminated by L-np .
Understanding the distribution and interaction between different Legionella species is valuable for obtaining and programming the correct strategies to control these bacteria (e.g., temperature values or concentrations of disinfectants), as well as for evaluation of the efficacy of preventive strategies applied. Moreover, accurate information regarding the relationships among isolates over time and during the disinfection treatment is helpful in the comprehension of the dynamics of Legionella contamination, in order to promptly take effective corrective measures.
|
Risk management patterns in radiation oncology—results of a national survey within the framework of the Patient Safety in German Radiation Oncology (PaSaGeRO) project | fa316d21-aaac-449e-a6d6-af3b87da1f81 | 10033570 | Internal Medicine[mh] | Radiation oncology (RO) plays a vital role in cancer management with almost half of all patients undergoing radiation therapy (RT) during their course of disease . Modern RT techniques complement patient-individualized treatment planning and image-guided dose delivery by adaptive radiation therapy from daily imaging and patient repositioning to daily treatment planning. With modern techniques, even small dosimetric deviations can result in significant errors, leading to a devastating impact in terms of side-effects and therapy success [ – ]. In recent years, several organizations have focused on patient safety during RT, providing recommendations for radiation oncology professionals [ , – ]. Several publications reported about occurrence of incidents during RT, e.g., application of a wrong plan, mistaken patient identity, underdosage on tumor tissues, or ignorance of previously applied dosages [ , , ]. The rate of incidents leading to major or long-lasting harm have decreased during the last few decades; nevertheless, newer data suggest that about 1% of incidents contribute to patients’ premature death . For Germany, numbers of incidents which fulfill certain conditions of severity as reported to a public authority reporting system are available, i.e., 49 RT incidents were reported nationwide in 2020 . As only highly dangerous incidents are reported here, the number of incidence actually occurring during daily treatment might be much higher . Risk management now offers the opportunity to anticipate possible failures, systematically learn from incidents, and implement measures to reduce risks which contributes to a safe therapy environment and is recommended by several authorities [ , – ]. Prospective RM is one key component of RM and is used to evaluate processes in patient care in terms of possible pitfalls, e.g., mistake of identities . It is recommended to perform these analyses with an interprofessional team before implementation of new processes or routinely, e.g., annually [ , , ]. Topic of analysis might be only one step of the treatment process (subprocess) or even several processes altogether . Several methods might be of help and especially failure modes and effects analysis (FMEA) with nomination of values to rank risks or use of a risk matrix are the topic of several publications [ – ]. Based on this analysis, measures can be initiated and implemented in daily routine to minimize the occurrence or the severity of errors. Reactive RM or analysis of adverse events is another key component of RM. It consists of reporting and processing of incidents and the subsequent definition of measures . Multiple systems to report incidents are established and range from international systems, e.g., ROSEIS (“radiation oncology reporting and education system” of the European Society for Radiation Oncology [ESTRO]) to in-house solutions for the department . What is learned from incidents and near-events (meaning failures which did not reach the patient) is valuable knowledge and should be processed preferably in an interprofessional team and translated into implementation of safety barriers [ – ]. Open communication about risk analyses and reported incidents support built-up of a sustainable safety culture and is highly recommended . Common methods to establish RM include conducting morbidity and mortality conferences (m&m conferences) or implementing a safety assurance/RM manager and/or committee . Accreditations of institutional quality management (QM) are a common tool to evaluate QM and in RM (e.g., according to DIN EN 15224) in order to provide high-quality care [ , , ]; however, the role of accreditation regarding patient safety has to be evaluated further . To underline the importance of a systematic risk management for German health care processes, it was declared to be mandatory by the patients’ right act for health care in general and by the new law on radiation protection ordinance and consecutive legislation for RT in 2018 which translates the EU legislation on the national level . From 2011, the ACCIRAD project analyzed Europe-wide approaches on implementation of RM tools in RO . The authors evaluated national legislation and recommendations on prospective and reactive risk management. Several countries having implemented different national regularities were identified, whereas no such implementation in national modus operandi for Germany was reported at that time. Multiple guidelines from different international societies and papers about best practice in RM in RO have been published [ , , , , ]. These offer recommendations on different dimensions of RM as described in the above section and promote growing knowledge on that topic in RO society. To complement these insights, we conducted a nationwide survey within the Patient Safety in German Radiation Oncology (PaSaGeRO) project to evaluate the current modus operandi for RM in Germany. PaSaGeRO is a joint project between the German Society for Medical Physics (DGMP) and the German Society for Radiation Oncology (DEGRO), investigating current approaches concerning patient safety and to define key components for a sustainable safety culture in German RO. In the past, interprofessional discussion rounds, e.g., in the context of meetings of the DGMP working group on risk management demonstrated insecurity and discordance among radiation oncology experts concerning implementation of risk management methods, protagonists or tools. Furthermore, these discussions underlined the strong relevance of risk management for radiation oncology processes not only to fulfill legal requirements, but to ensure safety during patients’ paths in therapy. Therefore, aims of this evaluation are to: Present the status quo of the patterns and needs of risk management in Germany. Hereby, we aim to evaluate the fulfillment of recommendations and regulations as described in the introduction. The results offer the possibility to create specific continuous-education courses or recommendations. Particularly addressed can be processes, tools, or topics of risk management which might differ from international recommendations or legal requirements and implementation in clinical routine. Enhance communication about risk management in general and subtopics such as incident reporting in particular. Systematic elaboration of the results offers the possibility for all German radiation oncology experts to compare our own approaches with current approaches in other departments and to advance the in-house risk management processes. Build a base for future research on questions of risk management or patient safety in general in RO, e.g., to investigate differences in approaches for RM in different institutions, methods for facilitated implementation of RM tools, or assessment of existing approaches. This represents the first systematic evaluation of RM approaches in German RO. The results can contribute knowledge to develop new recommendations and teaching content based on national legislation and existing guidelines regarding modern RO strategies.
We developed a questionnaire to address the following aspects: prospective RM, reactive RM, protagonists of RM, self-assessment concerning RM, and general data on the participating RO department. Items were designed to cover the above-mentioned components of risk management as found in national and international recommendations. It was designed using LimeSurvey (LimeSurvey GmbH, Hamburg, Germany) containing 18 questions in multiple-choice style. Self-assessment concerning knowledge on RM was answered using a 5-point Likert scale: very good (1)–good (2)–satisfying (3)–bad (4)–very bad (5). Process steps as mentioned in the questionnaire were aligned to the WHO radiotherapy risk profile . The set of subprocesses was adapted and finalized after several interprofessional discussion rounds with the aim to keep it well arranged and suitable for several institutional-specific workflows. The full questionnaire is available as supplementary material as an English transcript. An invitation link was first sent via e‑mail trough the official DGMP mailing list, second to a mailing list for medical physicists in Germany which is also open for non-DGMP members, and third also to all university hospital RO department in Germany, to reach possibly all ( n = 294 ) German RO departments. The survey was available between June and August 2021. Institutions were asked to answer the questions once per institution by a physicist, physician, or other member of the team. All answers were returned anonymously. All participants provided consent to data publication. Data were analyzed using SPSS statistics 28 (IBM, Armonk, NY, USA). Normal distribution was tested using the Kolmogorov–Smirnov test. Categorial data were compared via χ 2 test and Fisher’s exact test. The Kruskal–Wallis test was used to compare median values of different groups. A p -value of ≤ 0.05 was defined to be significant.
In all, 48 completed forms were eligible for evaluation: 17 institutions were university hospitals (UH; 35%), 11 were general and non-academic hospitals (23%), and 16 (33.3%) were private practices/medical service centers. Four participants did not name their affiliation. At that time, 21 institutions (44%) had an official accreditation for their QM (e.g., according to DIN EN 15224:2015), of which most were UH ( n = 12, 71% of all UH). Other hospitals and practices infrequently reported an accreditation (4/11 [36%] and 5/16 [31%], respectively). Significant differences were seen in the comparison of UH and private practices ( p = 0.038). Most institutions rated their knowledge concerning risk management as “satisfying” (21, 44%) or “bad” (11, 23%). Median value was three (“satisfying”) for all ratings. No difference in median value of UH vs. hospital vs. practice was seen ( p = 0.34). Sixteen departments ranked their knowledge to be bad or very bad (33%; Fig. ). No significant difference between (university) hospitals and practices was seen ( p = 0.54). No difference between departments with or without accreditations were seen ( p = 0.78). Risk rounds or m&m conferences were implemented in 22 departments (46%). Twenty (42%) departments had an explicit manager or committee for RM, while 16 (33%) had a dedicated specialist from the superior hospital risk management. Two institutions reported about external consultants for risk management purposes. Accredited departments rather tended to have risk rounds/m&m conferences ( p = 0.069), so did UH ( p = 0.04), Figs. and . A total of 31 participants (65%) wished for more information on RM through their respective society, 30 participants (63%) would prefer information to be provided via special courses on RM in RO, and 8 (17%) would prefer more information on the topic during university education.
In all, 42 institutions had undertaken at least one risk analysis so far (88%), of which 23 reported multiple or regular analyses (e.g., yearly, 48%). Five departments had not yet conduct prospective risk analyses (10%); one department did not answer this question. Departments with a risk management committee or risk manager reported significantly more often on risk analyses performed at multiple occasions or on a regular basis ( p < 0.01). Type of institution or achievement of accreditation did not impact number of analyses being performed. In 31 (65%) institutions, analyses were performed with an interprofessional team in multiple conferences; in 17 (35%), only one professional group was responsible, of which medical physics was in charge in 13 cases (76%). No difference between different types of departments was seen in this aspect. The most common method for prospective risk analyses was failure modes and effects analysis (FMEA, used by 25 institutions, 52%). Use of risk matrix or fault tree analyses was less common, while 6 (13%) participants used a combination of different methods. Furthermore, 44 institutions reported information on subprocesses they included in the analyses. The number of subprocesses included ranged between 1 and 12 of 13 suggested steps. Most commonly imaging, planning, patient setup, and first treatment were analyzed. One institution reported on analysis of possible errors during chemo-/immunotherapy. Average total number of subprocesses being included into analyses were higher in departments with risk committee or risk manager (8.9 vs. 7.1) and departments with accreditation of QM (8.8 vs 7.1). No difference between type of institution was seen. Results for respective subprocesses are provided in Table . In 71% ( n = 34) of departments, results were commonly presented to the team, either to the middle management/executive team ( n = 34) or to the complete team ( n = 17). In 10 departments (21%), the results were presented in written form, as a conference or hearing in 21 departments (44%), a combination of written form and conferences were used in 3 departments. In 26 departments (54%), measures were defined to improve processes after analyses.
Concerning critical incidents reporting, 23 institutions used the public authority reporting system for serious events in the last 12 months, of which 6 departments did not use other systems for less serious events (13%). A total of 28 used hospital-specific reporting systems such as CIRS (critical incident reporting system, 58%), while 14 participants used multiple systems (29%). None of the participants used international systems (ROSEIS or “safety in radiation oncology” [SAFRON]). In total, 34 institutions had made a report during the last year, six did not report, and eight did not reply to this question. Most departments reported 1–10 incidents in the last year ( n = 29, 60%). Availability of a risk manager or risk committee did not have significant impact on the number of reports neither did risk rounds/m&m conferences or CIRS newsletters. In ten departments (21%), a member of the physicists’ team was responsible for the workup of incidents or near-incidents. In four departments, the medical head of the department was responsible (8%); in 27 departments (56%), incidents were processed by a committee. No difference between different types of departments was seen. Root cause analyses were most common for evaluation ( n = 13, 27%). In most departments, incidents led to definition of measures for safety improvement in the last year ( n = 31, 65%). In 85% ( n = 41) of departments, information about incidents was presented to the team, either to the middle management/executive team ( n = 41) or to the complete team ( n = 27). Most common were conferences or hearings ( n = 30, 63%). Some departments communicated in written form ( n = 4) or used both oral and written form ( n = 6).
To illustrate approaches for RM in different countries, the European ACCIRAD project that started in 2011 systematically assessed nationwide regularities on incident reporting and prospective risk management . The authors reported on national RM requirements in 10 countries, yet in Germany, the implementation of national regularities for RM in RO took until 2018. As only few scientific papers on approaches of RM in Germany exist [ , , ], our results present a considerable support for enhanced communication about the topic, for the definition of guidelines and teaching content as well as a base for future research.
In all, 87% of departments reported having performed about at least one prospective risk analysis. Furthermore, half of the departments conducted more than one risk analysis, which correlates with availability of a risk manager or committee, so does the total number of subprocesses included. In contrast, 10% of departments did not conduct a risk analysis. Availability of a risk committee/manager is a key part of securing organizational structure for patient safety . As expected, conduction of risk analyses was facilitated after nomination of dedicated staff members for these positions . Most departments adhered to common recommendations concerning methods and protagonists, e.g., involving an FMEA and interprofessional teams. FMEA was recommended in multiple publications about RM in RO [ , – ]; however, deviation in methods might bear certain advantages as described recently . Most departments included planning, patient setup, and first radiation treatment in the analyses, while treatment routine was only assessed by two-third of institutions. Analysis of possible medication errors in general or during chemo-/immunotherapy were only reported by one institution in spite of their increasing role for patient safety . Available data identify the planning process as especially prone to severe errors . Nevertheless, patient assessment, information transfer, and routine RT were also identified as very prone to events and near-events and should be included in risk analyses [ , , ]. A possible reason for the difference in considering different process steps for analyses might be the fact that planning or setup/first treatment as subprocess are often conducted interprofessionally. This implies at least a four-eye-principle control of procedures and commonly discussions about possible improvements (peer reviews). During peer review of plans, possible safety concerns are discussed leading to plan modification in up to 66% of cases . In contrast, patient assessment or routine RT are conducted by few staff members in shorter time and are rarely presented in review rounds . Especially in our cohort, 29% of departments reported that medical physicists were responsible for analyses which might enhance the effect of focusing on planning and setup. A similar effect was previously reported from other countries , underlining the fact that several subprocesses are commonly underrepresented in important peer-review-based quality control, even though it is recommended to pay as much attention to, for example, RT routine as to technical plan quality . The limitation to 13 subprocesses to choose from in the questionnaire might not have fully covered all institution-specific workflows or nomenclatures, which impedes interpretation of the results. Nevertheless, it is conform with several recommendations for an institution to illustrate their own workflows and define specific subprocesses in a selectable extend before risk analyses [ , , ].
Use of incident-reporting systems is recommended by multiple societies and mandatory in Germany throughout healthcare institutions . The majority of departments in our study used incident reporting with a focus on hospital/institution specific reporting systems, e.g., CIRS . This is in line with previous work showing that the use of internal reporting systems is most popular among RO staff . Most departments reported less than 10 incidents in the last year, and some did not even report at all. None of the participants used international reporting systems. Hesitation to report an incident or near-event is described by previous publications and underlines the need for change towards a more open failure culture (e.g. “just culture” ). Another reason for low reporting rates could be a lack of education about availability, handling, and most important benefit of reporting systems for patient safety and should be addressed in educational sessions. Implementation of a low-barrier institutional reporting-system for the RO department is recommended and similarly implemented in German nuclear medicine and radiology departments or in other nations’ RO (e.g., PRISMA-RT in the Netherlands ). Since 2019, the German Department for Radiation Protection offers evaluation and recommendations in form of annually reports based on reported major incidents (Bedeutsame Vorkommnisse in der Medizin [BeVoMed]). Increasing benefit from the elaborated recommendations might encourage experts to aim for a generally more open reporting culture in the future. None of the “tools” of RM impacted the reporting rate. Most departments adhered to common recommendations and were able to define safety measures based on the analyses using recommended methods within an interprofessional team which was similarly described for prospective RM.
Most participants rated their knowledge about RM to be average or below. Most participants wished for more information by respective societies—offered via guidelines or special courses. Neither QM accreditation nor type of institution seemed to influence self-assessment. This is in line with results of a survey among American RO residents which describes low expertise regarding RM, although a program to improve skills on RM/QM existed for the RO curriculum . Even though inclusion of QM and RM in medical education is recommended in Germany , current evaluation of RO residency training in Germany does not offer information on specific education on patient safety/RM issues so far . This might be included in further evaluation as this is a dedicated topic of international recommendations . Division-specific education can be achieved by communicating existing RM processes. In our analyses, results of risk analyses—prospective and reactive—were mainly passed to the departments’ management levels. Especially regarding prospective analyses, only a minority of institutions communicated the results with all staff members (35%). As stated before, education about patient-safety issues shall be one key task of the management level (supported by the risk management committee) . University hospitals reported more often about risk managers/committees, m&m conferences, and support by superior RM. However, neither prospective risk analyses nor incident reporting were performed more frequently in these institutions. It is especially noteworthy that self-assessment of knowledge on RM did not show different values for accredited departments. This could be caused by interpersonal differences in QM and RM knowledge, on the one hand, and the fact that the preparation for the accreditation process does not provide secure handling of RM tools to a satisfying extent, on the other hand, even though accreditation was previously described as an important component of patient safety . An advantage of larger institutions might be the increased accessibility of personnel resources, on the one hand, and support by implemented quality and risk management structures, on the other hand. Translation of recommendations, e.g., about certain tools of RM, might be facilitated. Nevertheless, implementation of a profound safety-culture including educational elements is possible in any RO department . Departmental education should be facilitated by the societies focusing on national conditions. Leaned to existing recommendations , we expect updated guidelines from national societies and increasing focus on RM at national congresses and educational events .
Multiple institutions reported on tasks of RM being processed interprofessionally, so it is expected that answers represent impressions of interprofessional work. Nevertheless, especially for those departments where one person or profession is responsible for RM issues, answers might not completely represent balanced insight about the department’s viewpoints and knowledge about RM. For closer investigation of certain aspects which emerged as interesting from our survey, e.g., the role of accreditation, more detailed investigations are needed to obtain profound insight into the role of these points for development of a sustainable RM/patient safety structure. A limitation of our study is that only 48 evaluable survey forms out of 294 RO departments in Germany were returned. We sent invitations for the survey with the aim to reach as many physicists as possible. Nevertheless, the absolute number of institutions is not completely clear. Assuming we reached all departments in Germany, the response rate would be 16%. As several departments of a region might be pooled under a common administration, participants of these could have answered the questionnaire once referring to practices in all subsumed departments. As the survey was designed to be anonymous, this bias cannot be excluded. A reason for a low response rate could be the lack of time during daily routine allowing staff to complete survey forms . In addition, institutions without implemented processes of risk management might hesitate to take part in a survey about this topic. This may contribute to a low response rate, but also implies a bias of results, as our cohort might represent a rather experienced group in RM. Nevertheless, the return rate is comparable to other publications evaluating similar topics [ , , , ].
Our results reveal disparities in approaches for prospective and reactive RM in German RO departments, even though most departments adhered to current recommendations and regulations. Challenges in terms of implementation and realization of RM were identified and we showed that implementation of certain tools or protagonists such as risk managers can be of use in maintaining RM processes. We showed that the need for updated recommendations and education by technical societies exists throughout the entire institutional landscape.
Based on our results and existing literature, the authors encourage all departments to increase efforts for education of their staff concerning risk management, including outcomes of in-house risk analyses. We suggest increasing the frequency of prospective risk analyses and to incorporate “routine processes”. Access to reporting systems for all personnel including use-briefing and definition of criteria for events should be provided. Existing examples for low-threshold reporting systems from RO worldwide or from other specialties could be adapted for German RO proposes on the institutional or cross-regional level. Future research can evaluate a broader spectrum of patient safety-related issues and safety culture in our field. Questions to address based on our results may be how communication and education concerning incidents, reporting, prospective analyses, and measures can be improved and how incident-learning can be transferred to maintaining changes in processes or how safety concepts can be assessed.
Supplementary material: Survey about risk management – English transcript (original document in German)
|
Large scale crowdsourced radiotherapy segmentations across a variety of cancer anatomic sites | 00331ec6-e91d-443d-aa92-dc4ad12b62b9 | 10033824 | Internal Medicine[mh] | Since the advent of contemporary radiation delivery techniques for cancer treatment, clinician generated segmentation (also termed contouring or delineation) of target structures (e.g., primary tumors and metastatic lymph nodes) and organs at risk (e.g., healthy tissues whose irradiation could lead to damage and/or side effects) on medical images has become a necessity in the radiotherapy workflow . These segmentations are typically provided by trained medical professionals, such as radiation oncologists. While segmentations can be performed on any imaging modality that provides sufficient discriminative capabilities to visualize regions of interest (ROIs), the current radiotherapy workflow prioritizes the use of computed tomography (CT) for ROI segmentation due to its ubiquitous nature and use in radiotherapy dose calculations. Subsequently, clinicians spend a large fraction of their time and effort generating ROI segmentations on CT imaging necessary for the radiotherapy workflow. Interobserver and intraobserver variability are well-documented byproducts of the use of manual human-generated segmentations , . While consensus radiotherapy guidelines to ensure ROI segmentation quality have been developed and shown to reduce variability , these guidelines are not necessarily followed by all practicing clinicians. Therefore, segmentation variability remains a significant concern in maintaining radiotherapy plan quality and consistency. Recent computational improvements in machine learning, particularly deep learning, have prompted the increasing development and deployment of accurate ROI auto-segmentation algorithms to reduce radiotherapy segmentation variability – . However, for auto-segmentation algorithms to be clinically useful, their input data (training data) should reflect high-quality “gold-standard” annotations. While research has been performed on the impact of interobserver variability and segmentation quality for auto-segmentation training – , it remains unclear how “gold-standard” segmentations should be defined and generated. One common approach, consensus segmentation generation, seeks to crowdsource multiple segmentations from different annotators to generate a high-quality ground-truth segmentation. While multi-observer public medical imaging segmentation datasets exist – , there remains a lack of datasets with a large number of annotators for radiotherapy applications. The Contouring Collaborative for Consensus in Radiation Oncology (C3RO) challenge was developed to engage radiation oncologists across various expertise levels in cloud-based ROI crowdsourced segmentation . Through this collaboration, a large number of clinicians generated ROI segmentations using CT images from 5 unique radiotherapy cases: breast, sarcoma, head and neck, gynecologic, and gastrointestinal. In this data descriptor, we present the curation and processing of the data from the C3RO challenge. The primary contribution of this dataset is unprecedented large-scale multi-annotator individual and consensus segmentations of various ROIs crucial for radiotherapy planning in an easily accessible and standardized imaging format. These data can be leveraged for exploratory analysis of segmentation quality across a large number of annotators, consensus segmentation experiments, and auto-segmentation model benchmarking. An overview of this data descriptor is shown in Fig. .
Patient population Five separate patients who had undergone radiotherapy were retrospectively collected from our collaborators at various institutions. Each patient had received a pathologically confirmed diagnosis of cancer of one of the following sites: breast (post-mastectomy intraductal carcinoma), sarcoma (malignant peripheral nerve sheath tumor of the left thigh), head and neck (oropharynx with nodal spread, [H&N]), gynecologic (cervical cancer, [GYN]), and gastrointestinal (anal cancer, [GI]). Clinical characteristics of these patients are shown in Table . Of note, these five disease sites were included as part of the C3RO challenge due to being among the most common disease sites treated by radiation oncologists; additional disease sites were planned but were not realized due to diminishing community participation in C3RO. Specific patient cases were selected by C3RO collaborators on the basis of being adequate reflections of routine patients a generalist radiation oncologist may see in a typical workflow (i.e., not overly complex). Further details on the study design for C3RO can be found in Lin & Wahid et al . . Imaging protocols Each patient received a radiotherapy planning CT scan which was exported in Digital Imaging and Communications in Medicine (DICOM) format. CT image acquisition characteristics are shown in Table . All images were acquired on scanners that were routinely used for radiotherapy planning at their corresponding institutions with appropriate calibration and quality assurance by technical personnel. The sarcoma, H&N, and GI cases received intravenous contrast, the GU case received oral contrast, and the breast case did not receive any contrast. Of note, the H&N case had metal streak artifacts secondary to metallic implants in the upper teeth, which obscured anatomy near the mandible. No other cases contained noticeable image artifacts. Notably, the sarcoma case also received a magnetic resonance imaging (MRI) scan, while the H&N and GI cases received full body positron emission tomography (PET) scans. The sarcoma MRI scan was acquired on a GE Signa HDxt device and corresponded to a post-contrast spin echo T1-weighted image with a slice thickness of 3.0 mm and in-plane resolution of 0.35 mm. The H&N PET scan was acquired on a GE Discovery 600 device with a slice thickness of 3.3 mm and in-plane resolution of 2.73 mm. The GI PET scan was acquired on a GE Discovery STE device with a slice thickness of 3.3 mm and in-plane resolution of 5.47 mm. IRB exemption and data storage The retrospective acquisition, storage, and use of these DICOM files have been reviewed by the Memorial Sloan Kettering (MSK) Human Research Protection Program (HRPP) Office on May 26, 2021 and were determined to be exempt research as per 45 CFR 46.104(d)(3),(i)(a), (ii) and (iii), (i)(b),(ii) and (iii), (i)(c), (ii) and (iii) and 45.CFR.46.111(a)(7). A limited IRB review of the protocol X19-040 A(1) was conducted via expedited process in accordance with 45 CFR 46.110(b), and the protocol was approved on May 26, 2021. DICOM files were obtained and stored on MIMcloud (MIM Software Inc., Ohio, USA), which is a HIPAA-compliant cloud-based storage for DICOM image files that has been approved for use at MSK by MSK’s Information Security team. DICOM anonymization For each image, the DICOM header tags containing the patient name, date of birth, and patient identifier number were consistently removed from all DICOM files using DicomBrowser v. 1.5.2 . The removal of acquisition data and time metadata (if available in DICOM header tags) caused compatibility issues with ProKnow so were kept as is. Moreover, if institution name or provider name were available in the DICOM file, they were not removed as they were not considered protected health information. Select cases (breast, GYN, GI) were previously anonymized using the DICOM Import Export tool (Varian Medical Systems, CA, USA). Participant details To register for the challenge, participants completed a baseline questionnaire that included their name, email address, affiliated institution, country, specialization, years in practice, number of disease sites treated, volume of patients treated per month for the designated tumor site, how they learned about this challenge, and reasons for participation. Registrant intake information was collected through the Research Electronic Data Capture (REDCap) system - a widely used web application for managing survey databases ; an example of the intake form can be found at: https://redcap.mskcc.org/surveys/?s=98ARPWCMAT . The research conducted herein was approved by the HRRP at MSK (IRB#: X19-040 A(1); approval date: May 26, 2021). All subjects prospectively consented to participation in the present study, as well as to the collection, use, and disclosure of de-identified aggregate subject information and responses. Participants were categorized as recognized experts or non-experts. Recognized experts were identified by our C3RO team (EFG, CDF, DL) based on participation in the development of national guidelines or other extensive scholarly activities. Recognized experts were board-certified physicians with expertise in the specific disease site. Non-experts were any participants not categorized as an expert for that disease site. All non-experts had some knowledge of human anatomy, with the majority being composed of practicing radiation oncologists but also included resident physicians, radiation therapists, and medical physicists. Worthy of note, a participant could only be considered an expert for one disease site, but could have participated as a non-expert for other disease sites. Out of 1,026 registrants, 221 participated in generating segmentations, which were used for this dataset; due to the low participation rate, participants may represent a biased sample of registrants. Of note, participants could provide segmentations for multiple cases. Additional demographic characteristics of the participants can be found in Lin & Wahid et al . . ProKnow segmentation platform Participants were given access to the C3RO workspace on ProKnow (Elekta AB, Stockholm, Sweden). ProKnow is a commercially available radiotherapy clinical workflow tool that allows for centralization of data in a secure web-based repository; the ProKnow system has been adopted by several large scale medical institutions and is used routinely in clinical and research environments. Anonymized CT DICOM images for each case were imported into the ProKnow system for participants to segment; anonymized MRI and PET images were also imported for select cases as available. Each case was attributed a short text prompt describing the patient presentation along with any additional information as needed. Participants were allowed to utilize common image manipulation (scrolling capabilities, zooming capabilities, window leveling, etc.) and segmentation (fill, erase, etc.) tools for generating their segmentations. No auto-segmentation capabilities were provided to the participants, i.e., all segmentations were manually generated. Notably, for the sarcoma case, an external mask of the patient’s body and a mask of the left femur was provided to participants. Screenshots of the ProKnow web interface platform for the various cases are shown in Fig. . Segmentation details For each case, participants were requested to segment a select number of ROIs corresponding to target structures or OARs. Notably, not all participants generated segmentations for all ROIs. ROIs for each participant were combined into one structure set in the ProKnow system. ROIs were initially named in a consistent, but non-standardized format, so during file conversion ROIs were renamed based on The Report of American Association of Physicists in Medicine Task Group 263 (TG-263) suggested nomenclature ; TG-263 was chosen due its ubiquity in standardized radiotherapy nomenclature. A list of the ROIs and the number of available segmentations stratified by participant expertise level is shown in Supplementary Table . Image processing and file conversion For each case, anonymized CT images and structure sets for each annotator were manually exported from ProKnow in DICOM and DICOM radiotherapy structure (RTS) format, respectively. The Neuroimaging Informatics Technology Initiative (NIfTI) format is increasingly used for reproducible imaging research – due to its compact file size and ease of implementation in computational models . Therefore, in order to increase the interoperability of these data, we converted all our DICOM imaging and segmentation data to NIfTI format. For all file conversion processes, Python v. 3.8.8 was used. An overview of the image processing workflow is shown in Fig. . In brief, using an in-house Python script, DICOM images and structure sets were loaded into numpy array format using the DICOMRTTool v. 0.4.2 library , and then converted to NIfTI format using SimpleITK v. 2.1.1 . For each annotator, each individual structure contained in the structure set was separately converted into a binary mask (0 = background, 1 = ROI), and was then converted into separate NIfTI files. Notably, voxels fully inside and outside the contour are included and not include in the binary mask, respectively, while voxels that overlapped the segmentation (edge voxels) were counted as surface coordinates and included in the binary mask; additional details on array conversion can be found in the DICOMRTTool documentation . Examples of random subsets of five expert segmentations for each ROI from each case are shown in Fig. . Consensus segmentation generation In addition to ground-truth expert and non-expert segmentations for all ROIs, we also generated consensus segmentations using the Simultaneous Truth and Performance Level Estimation (STAPLE) method, a commonly used probabilistic approach for combining multiple segmentations – . Briefly, the STAPLE method uses an iterative expectation-maximization algorithm to compute a probabilistic estimate of the “true” segmentation by deducing an optimal combination of the input segmentations and incorporating a prior model for the spatial distribution of segmentations as well as implementing spatial homogeneity constraints . For our specific implementation of the STAPLE method, we utilized the SimpleITK STAPLE function with a default threshold value of 0.95. For each ROI, all available binary segmentation masks acted as inputs to the STAPLE function for each expertise level, subsequently generating binary STAPLE segmentation masks for each expertise level (i.e., STAPLE expert and STAPLE non-expert ). An overview of the consensus segmentation workflow is shown in Fig. . Examples of STAPLE expert and STAPLE non-expert segmentations for each ROI are shown in Fig. .
Five separate patients who had undergone radiotherapy were retrospectively collected from our collaborators at various institutions. Each patient had received a pathologically confirmed diagnosis of cancer of one of the following sites: breast (post-mastectomy intraductal carcinoma), sarcoma (malignant peripheral nerve sheath tumor of the left thigh), head and neck (oropharynx with nodal spread, [H&N]), gynecologic (cervical cancer, [GYN]), and gastrointestinal (anal cancer, [GI]). Clinical characteristics of these patients are shown in Table . Of note, these five disease sites were included as part of the C3RO challenge due to being among the most common disease sites treated by radiation oncologists; additional disease sites were planned but were not realized due to diminishing community participation in C3RO. Specific patient cases were selected by C3RO collaborators on the basis of being adequate reflections of routine patients a generalist radiation oncologist may see in a typical workflow (i.e., not overly complex). Further details on the study design for C3RO can be found in Lin & Wahid et al . .
Each patient received a radiotherapy planning CT scan which was exported in Digital Imaging and Communications in Medicine (DICOM) format. CT image acquisition characteristics are shown in Table . All images were acquired on scanners that were routinely used for radiotherapy planning at their corresponding institutions with appropriate calibration and quality assurance by technical personnel. The sarcoma, H&N, and GI cases received intravenous contrast, the GU case received oral contrast, and the breast case did not receive any contrast. Of note, the H&N case had metal streak artifacts secondary to metallic implants in the upper teeth, which obscured anatomy near the mandible. No other cases contained noticeable image artifacts. Notably, the sarcoma case also received a magnetic resonance imaging (MRI) scan, while the H&N and GI cases received full body positron emission tomography (PET) scans. The sarcoma MRI scan was acquired on a GE Signa HDxt device and corresponded to a post-contrast spin echo T1-weighted image with a slice thickness of 3.0 mm and in-plane resolution of 0.35 mm. The H&N PET scan was acquired on a GE Discovery 600 device with a slice thickness of 3.3 mm and in-plane resolution of 2.73 mm. The GI PET scan was acquired on a GE Discovery STE device with a slice thickness of 3.3 mm and in-plane resolution of 5.47 mm.
The retrospective acquisition, storage, and use of these DICOM files have been reviewed by the Memorial Sloan Kettering (MSK) Human Research Protection Program (HRPP) Office on May 26, 2021 and were determined to be exempt research as per 45 CFR 46.104(d)(3),(i)(a), (ii) and (iii), (i)(b),(ii) and (iii), (i)(c), (ii) and (iii) and 45.CFR.46.111(a)(7). A limited IRB review of the protocol X19-040 A(1) was conducted via expedited process in accordance with 45 CFR 46.110(b), and the protocol was approved on May 26, 2021. DICOM files were obtained and stored on MIMcloud (MIM Software Inc., Ohio, USA), which is a HIPAA-compliant cloud-based storage for DICOM image files that has been approved for use at MSK by MSK’s Information Security team.
For each image, the DICOM header tags containing the patient name, date of birth, and patient identifier number were consistently removed from all DICOM files using DicomBrowser v. 1.5.2 . The removal of acquisition data and time metadata (if available in DICOM header tags) caused compatibility issues with ProKnow so were kept as is. Moreover, if institution name or provider name were available in the DICOM file, they were not removed as they were not considered protected health information. Select cases (breast, GYN, GI) were previously anonymized using the DICOM Import Export tool (Varian Medical Systems, CA, USA).
To register for the challenge, participants completed a baseline questionnaire that included their name, email address, affiliated institution, country, specialization, years in practice, number of disease sites treated, volume of patients treated per month for the designated tumor site, how they learned about this challenge, and reasons for participation. Registrant intake information was collected through the Research Electronic Data Capture (REDCap) system - a widely used web application for managing survey databases ; an example of the intake form can be found at: https://redcap.mskcc.org/surveys/?s=98ARPWCMAT . The research conducted herein was approved by the HRRP at MSK (IRB#: X19-040 A(1); approval date: May 26, 2021). All subjects prospectively consented to participation in the present study, as well as to the collection, use, and disclosure of de-identified aggregate subject information and responses. Participants were categorized as recognized experts or non-experts. Recognized experts were identified by our C3RO team (EFG, CDF, DL) based on participation in the development of national guidelines or other extensive scholarly activities. Recognized experts were board-certified physicians with expertise in the specific disease site. Non-experts were any participants not categorized as an expert for that disease site. All non-experts had some knowledge of human anatomy, with the majority being composed of practicing radiation oncologists but also included resident physicians, radiation therapists, and medical physicists. Worthy of note, a participant could only be considered an expert for one disease site, but could have participated as a non-expert for other disease sites. Out of 1,026 registrants, 221 participated in generating segmentations, which were used for this dataset; due to the low participation rate, participants may represent a biased sample of registrants. Of note, participants could provide segmentations for multiple cases. Additional demographic characteristics of the participants can be found in Lin & Wahid et al . .
Participants were given access to the C3RO workspace on ProKnow (Elekta AB, Stockholm, Sweden). ProKnow is a commercially available radiotherapy clinical workflow tool that allows for centralization of data in a secure web-based repository; the ProKnow system has been adopted by several large scale medical institutions and is used routinely in clinical and research environments. Anonymized CT DICOM images for each case were imported into the ProKnow system for participants to segment; anonymized MRI and PET images were also imported for select cases as available. Each case was attributed a short text prompt describing the patient presentation along with any additional information as needed. Participants were allowed to utilize common image manipulation (scrolling capabilities, zooming capabilities, window leveling, etc.) and segmentation (fill, erase, etc.) tools for generating their segmentations. No auto-segmentation capabilities were provided to the participants, i.e., all segmentations were manually generated. Notably, for the sarcoma case, an external mask of the patient’s body and a mask of the left femur was provided to participants. Screenshots of the ProKnow web interface platform for the various cases are shown in Fig. .
For each case, participants were requested to segment a select number of ROIs corresponding to target structures or OARs. Notably, not all participants generated segmentations for all ROIs. ROIs for each participant were combined into one structure set in the ProKnow system. ROIs were initially named in a consistent, but non-standardized format, so during file conversion ROIs were renamed based on The Report of American Association of Physicists in Medicine Task Group 263 (TG-263) suggested nomenclature ; TG-263 was chosen due its ubiquity in standardized radiotherapy nomenclature. A list of the ROIs and the number of available segmentations stratified by participant expertise level is shown in Supplementary Table .
For each case, anonymized CT images and structure sets for each annotator were manually exported from ProKnow in DICOM and DICOM radiotherapy structure (RTS) format, respectively. The Neuroimaging Informatics Technology Initiative (NIfTI) format is increasingly used for reproducible imaging research – due to its compact file size and ease of implementation in computational models . Therefore, in order to increase the interoperability of these data, we converted all our DICOM imaging and segmentation data to NIfTI format. For all file conversion processes, Python v. 3.8.8 was used. An overview of the image processing workflow is shown in Fig. . In brief, using an in-house Python script, DICOM images and structure sets were loaded into numpy array format using the DICOMRTTool v. 0.4.2 library , and then converted to NIfTI format using SimpleITK v. 2.1.1 . For each annotator, each individual structure contained in the structure set was separately converted into a binary mask (0 = background, 1 = ROI), and was then converted into separate NIfTI files. Notably, voxels fully inside and outside the contour are included and not include in the binary mask, respectively, while voxels that overlapped the segmentation (edge voxels) were counted as surface coordinates and included in the binary mask; additional details on array conversion can be found in the DICOMRTTool documentation . Examples of random subsets of five expert segmentations for each ROI from each case are shown in Fig. .
In addition to ground-truth expert and non-expert segmentations for all ROIs, we also generated consensus segmentations using the Simultaneous Truth and Performance Level Estimation (STAPLE) method, a commonly used probabilistic approach for combining multiple segmentations – . Briefly, the STAPLE method uses an iterative expectation-maximization algorithm to compute a probabilistic estimate of the “true” segmentation by deducing an optimal combination of the input segmentations and incorporating a prior model for the spatial distribution of segmentations as well as implementing spatial homogeneity constraints . For our specific implementation of the STAPLE method, we utilized the SimpleITK STAPLE function with a default threshold value of 0.95. For each ROI, all available binary segmentation masks acted as inputs to the STAPLE function for each expertise level, subsequently generating binary STAPLE segmentation masks for each expertise level (i.e., STAPLE expert and STAPLE non-expert ). An overview of the consensus segmentation workflow is shown in Fig. . Examples of STAPLE expert and STAPLE non-expert segmentations for each ROI are shown in Fig. .
Medical images and multi-annotator segmentation data This data collection primarily consists of 1985 3D volumetric compressed NIfTI files (.nii.gz file extension) corresponding to CT images and segmentations of ROIs from various disease sites (breast, sarcoma, H&N, GYN, GI). Analogously formatted MRI and PET images are available for select cases (sarcoma, H&N, GI). ROI segmentation NIfTI files are provided in binary mask format (0 = background, 1 = ROI); file names for each ROI are provided in TG-263 notation. All medical images and ROI segmentations were derived from original DICOM and DICOM RTS files (.dcm file extension) respectively, which for completeness are also provided in this data collection. In addition, Python code to recreate the final NIfTI files from DICOM files is also provided in the corresponding GitHub repository (see Code Availability section). Consensus segmentation data Consensus segmentations for experts and non-experts generated using the STAPLE method for each ROI have also been provided in compressed NIfTI file format (.nii.gz file extension). Consensus segmentation NIfTI files are provided in binary mask format (0 = background, 1 = ROI consensus). Python code to recreate the STAPLE NIfTI files from input annotator NIfTI files is also provided in the corresponding GitHub repository (see Code Availability section). Annotator demographics data We also provide a single Microsoft Excel file (.xlsx file extension) containing each annotator’s gender, race/ethnicity, geographic setting, profession, years of experience, practice type, and categorized expertise level (expert, non-expert). Geographic setting was re-coded as “United States” or “International” to further de-identify the data. Each separate sheet corresponds to a separate disease site (sheet 1 = breast, sheet 2 = sarcoma, sheet 3 = H&N, sheet 4 = GU, sheet 5 = GI). Moreover, in order to foster secondary analysis of registrant data, we also include a sheet containing the combined intake data for all registrants of C3RO, including those who did not provide annotations (sheet 6). Folder structure and identifiers Each disease site is represented by a top-level folder, containing a subfolder for images and segmentations. The annotator demographic excel file is located in the same top-level location as the disease site folders. Image folders contain separate subfolders for NIfTI format and DICOM format images. Segmentation folders contain separate subfolders for expert and non-expert segmentations. Each expertise folder contains separate subfolders for each annotator (which contains separate subfolders for DICOM and NIfTI formatted files) and the consensus segmentation (only available in NIfTI format). The data have been specifically structured such that for any object (i.e., an image or segmentation), DICOM and NIfTI subdirectories are available for facile partitioning of data file types. An overview of the organized data records for an example case is shown in Fig. . Segmentation files (DICOM and NIfTI) are organized by anonymized participant ID numbers and can be cross referenced against the excel data table using this identifier. The raw data, records, and supplemental descriptions of the meta-data files are cited under Figshare doi: 10.6084/m9.figshare.21074182 .
This data collection primarily consists of 1985 3D volumetric compressed NIfTI files (.nii.gz file extension) corresponding to CT images and segmentations of ROIs from various disease sites (breast, sarcoma, H&N, GYN, GI). Analogously formatted MRI and PET images are available for select cases (sarcoma, H&N, GI). ROI segmentation NIfTI files are provided in binary mask format (0 = background, 1 = ROI); file names for each ROI are provided in TG-263 notation. All medical images and ROI segmentations were derived from original DICOM and DICOM RTS files (.dcm file extension) respectively, which for completeness are also provided in this data collection. In addition, Python code to recreate the final NIfTI files from DICOM files is also provided in the corresponding GitHub repository (see Code Availability section).
Consensus segmentations for experts and non-experts generated using the STAPLE method for each ROI have also been provided in compressed NIfTI file format (.nii.gz file extension). Consensus segmentation NIfTI files are provided in binary mask format (0 = background, 1 = ROI consensus). Python code to recreate the STAPLE NIfTI files from input annotator NIfTI files is also provided in the corresponding GitHub repository (see Code Availability section).
We also provide a single Microsoft Excel file (.xlsx file extension) containing each annotator’s gender, race/ethnicity, geographic setting, profession, years of experience, practice type, and categorized expertise level (expert, non-expert). Geographic setting was re-coded as “United States” or “International” to further de-identify the data. Each separate sheet corresponds to a separate disease site (sheet 1 = breast, sheet 2 = sarcoma, sheet 3 = H&N, sheet 4 = GU, sheet 5 = GI). Moreover, in order to foster secondary analysis of registrant data, we also include a sheet containing the combined intake data for all registrants of C3RO, including those who did not provide annotations (sheet 6).
Each disease site is represented by a top-level folder, containing a subfolder for images and segmentations. The annotator demographic excel file is located in the same top-level location as the disease site folders. Image folders contain separate subfolders for NIfTI format and DICOM format images. Segmentation folders contain separate subfolders for expert and non-expert segmentations. Each expertise folder contains separate subfolders for each annotator (which contains separate subfolders for DICOM and NIfTI formatted files) and the consensus segmentation (only available in NIfTI format). The data have been specifically structured such that for any object (i.e., an image or segmentation), DICOM and NIfTI subdirectories are available for facile partitioning of data file types. An overview of the organized data records for an example case is shown in Fig. . Segmentation files (DICOM and NIfTI) are organized by anonymized participant ID numbers and can be cross referenced against the excel data table using this identifier. The raw data, records, and supplemental descriptions of the meta-data files are cited under Figshare doi: 10.6084/m9.figshare.21074182 .
Data annotations Segmentation DICOM and NIfTI files were manually verified by study authors (D.L., K.A.W., O.S.) to be annotated with the appropriate corresponding ROI names. Segmentation interobserver variability We calculated the pairwise interobserver variability (IOV) for each ROI for each disease site across experts and non-experts. Specifically, for each metric all pairwise combinations between all available segmentations in a given group (expert or non-expert) were calculated; median and interquartile range values are reported in Supplementary Table . Calculated metrics included the Dice Similarity coefficient (DSC), average surface distance (ASD), and surface DSC (SDSC). SDSC was calculated based on ROI specific thresholds determined by the median pairwise mean surface distance of all expert segmentations for that ROI as suggested in literature . Metrics were calculated using the Surface Distances Python package , and in-house Python code. For specific equations for metric calculations please see corresponding Surface Distances Python package documentation . Resultant values are broadly consistent with previous work in breast , sarcoma , H&N , , , GYN , and GI – IOV studies.
Segmentation DICOM and NIfTI files were manually verified by study authors (D.L., K.A.W., O.S.) to be annotated with the appropriate corresponding ROI names.
We calculated the pairwise interobserver variability (IOV) for each ROI for each disease site across experts and non-experts. Specifically, for each metric all pairwise combinations between all available segmentations in a given group (expert or non-expert) were calculated; median and interquartile range values are reported in Supplementary Table . Calculated metrics included the Dice Similarity coefficient (DSC), average surface distance (ASD), and surface DSC (SDSC). SDSC was calculated based on ROI specific thresholds determined by the median pairwise mean surface distance of all expert segmentations for that ROI as suggested in literature . Metrics were calculated using the Surface Distances Python package , and in-house Python code. For specific equations for metric calculations please see corresponding Surface Distances Python package documentation . Resultant values are broadly consistent with previous work in breast , sarcoma , H&N , , , GYN , and GI – IOV studies.
The image and segmentation data from this data collection are provided in original DICOM format (where applicable) and compressed NIfTI format with the accompanying excel file containing demographic information indexed by annotator identifiers. We invite all interested researchers to download this dataset for use in segmentation, radiotherapy, and crowdsourcing related research. Moreover, we encourage this dataset’s use for clinical decision support tool development. While the individual number of patient cases for this dataset is too small for traditional machine learning development (i.e., deep learning auto-segmentation training), this dataset could act as a benchmark reference for testing existing auto-segmentation algorithms. Importantly, this dataset could also be used as a standardized reference for future interobserver variability studies seeking to investigate further participant expertise criteria, e.g., true novice annotators (no previous segmentation or anatomy knowledge) could attempt to segment ROI structures on CT images, which could then be compared to our expert and non-expert annotators. Finally, in line with the goals of the eContour collaborative , these data could be used to help develop educational tools for radiation oncology clinical training. The segmentations provided in this data descriptor have been utilized in a study by Lin & Wahid et al . . This study demonstrated several results that were consistent with existing literature, including: (1). target ROIs tended to exhibit greater variability than OAR ROIs , (2). H&N ROIs exhibited higher interobserver variability compared to other disease sites , , and (3). non-expert consensus segmentations could approximate gold-standard expert segmentations . Original DICOM format images and structure sets may be viewed and analyzed in radiation treatment planning software or select digital image viewing applications, depending on the end-user’s requirements. Current open-source software for these purposes includes ImageJ , dicompyler , ITK-Snap , and 3D Slicer with the SlicerRT extension . Processed NIfTI format images and segmentations may be viewed and analyzed in any NIfTI viewing application, depending on the end-user’s requirements. Current open-source software for these purposes includes ImageJ , ITK-Snap , and 3D Slicer .
Supplementary Table 1 Supplementary Table 2
|
Predicting outcomes in chronic kidney disease: needs and preferences of patients and nephrologists | 2b967d47-c63e-4983-8e1b-1df39818956f | 10035227 | Internal Medicine[mh] | The course of chronic kidney disease (CKD) and the risk of progression to end-stage kidney disease (ESKD) vary among patients . Guidelines recommend that nephrologists use clinical prediction models (CPMs) to help identify patients at increased risk of CKD progression and adjust their treatment to help limit further kidney function decline . In addition, multiple studies showed that patients are interested in prognostic information, and that they value this information for behavioural change and treatment planning . CPMs can also be used to help establish the optimal timing of starting education on kidney replacement therapy (KRT) when patients do progress to the more advanced stages of CKD. Timely education and decisional support allow for effective decision-making, and may prevent delays in the decision-making process which are associated with increased patient morbidity, mortality and health care costs . Numerous CPMs have been developed for CKD practice over the years. These include models that predict the risk of progression to ESKD or adverse outcomes of different KRT modalities, such as: 1) mortality after dialysis initiation , and 2) rejection after kidney transplantation . Some of these models, such as the Kidney Failure Risk Equation (KFRE), have been extensively validated and offer good predictive performance . Even though well-validated models are readily available and guidelines recommend that nephrologists use CPMs, the actual use of CPMs in CKD practice seems limited . This may be related to the CPMs themselves (e.g., limitations in predictive performance or user friendliness), and/or to the intended users (e.g., doubts about the reliability and generalizability of CPMs) . CPMs are also often developed without the input of end-users (i.e., patients and nephrologists), and as a consequence, lack clinical relevance . In addition, patients and nephrologists often prioritize different (treatment) outcomes and may have different needs and preferences regarding the use and purpose of CPMs in CKD practice. Therefore, the aim of this study was to: 1) evaluate to what extent CPMs are currently used in the Dutch CKD practice, 2) identify patients’ and nephrologists’ needs and preferences regarding predictions in CKD, and 3) explore determinants that may affect the adoption of CPMs in CKD practice. Our results can be used to guide implementation of CPMs and inform future development of CPMs.
Study design A national survey study among CKD patients and nephrologists in The Netherlands was conducted. First, patients’ attitudes towards different CPMs predicting the course of CKD were explored in semi-structured interviews. Next, two online surveys were developed and distributed: one for patients and one for nephrologists. Semi-structured interviews Patients with CKD were interviewed to explore their attitudes towards the use of CPMs in CKD practice. These interviews were held in the context of a larger study on the development of a CKD dashboard . During these interviews, two different predictions were introduced: 1) the prediction from the KFRE: a 2- and 5-year risk of progression to kidney failure for stages 3 to 5 CKD patients (in %), and 2) a prediction about the time until kidney failure (in years). Mock-ups were used to present these predictions in a similar lay-out to have patients focus on the meaning of the predictions rather than on how these were presented (Additional file ). Patients were asked to ‘think-out-loud’ and give their first impressions on the presented predictions. Patients were subsequently asked whether they would want to be provided with these predictions in (including reasons why), and how they would prefer to receive this information. Online surveys Two surveys were developed: one for CKD patients and one for nephrologists. Each survey started with an introductory text and an explanation of the definition of a CPM. This explanation was supplemented with an infographic to facilitate understanding (Additional file ). Both surveys consisted of questions assessing: 1) the current use of CPMs in Dutch CKD practice, 2) preferences for predictions in CKD, 3) preferences for predictions about CKD progression (to ESKD), and 4) barriers and facilitators for the adoption of CPMs in clinical practice. The patient surveys also included questions about educational levels, which was measured according to the International Standard Classification of Education and health literacy, which was measured with the Set of Brief Screening questions (SBSQ) . The SBSQ assesses perceived difficulties with health information based on three 5-point Likert scale statements ranging from 1–5. An average score of ≤ 3 indicates inadequate health literacy and a score of > 3 adequate health literacy. In the patient survey, the Threatening Medical Situations Inventory (TMSI) was used to assess whether patients handle medically threatening information with either monitoring (attending to the problem) or blunting (avoiding the problem) coping behaviour, since this may affect their views on receiving predictions . In the TMSI, patients are asked how they would handle hypothetical situations. They report on a 5-point Likert scale how likely it would be for them to apply three monitoring and three blunting strategies. Total scores for both the monitoring and blunting strategies are subsequently calculated (ranging from 6–30) . In the nephrologist survey, the Measurement Instrument for Determinants of Innovations (MIDI) was used to identify enablers for the adoption of CPMs in clinical practice . For three domains (the innovation, the user, and the organisation), nephrologists had to pick the two most important determinants that may facilitate the adoption of CPMs in clinical practice. Additional file shows the validated survey instruments used and the study-specific survey questions. Pretesting the surveys Both surveys were tested and amended for face validity by a: 1) communication scientist (CvU), 2) professor of medical decision-making (AS), 3) nephrologist (WB), and 4) cognitive psychologist specialised in communication research (AP). The patient survey was written at the B1 level of the common European framework of reference for languages (CEFRL) to ensure comprehensibility . It was also tested for face validity by five CKD patients recruited by the Dutch Kidney Patients Association. Participants, recruitment and informed consent Patients with CKD were recruited for the interviews by their nephrologists in two Dutch hospitals (St. Antonius hospital and Maasstad hospital) in February 2021. All participants gave informed consent. For the surveys, CKD patients and nephrologists were recruited from November 2021 until March 2022. Patients were approached via e-mail through the online platform of the Dutch Kidney Patients Association. The nephrologists were approached via e-mail through the online platform of the Dutch Federation for Nephrology. Both surveys were anonymous; no personal identifying information was registered. The patients and nephrologists who agreed to participate were asked to consent with the use of their answers for research and publication purposes when they started the survey. According to the Dutch medical research involving human subjects act, ethical approval was not required for the surveys because participants were not subjected to (medical) procedures or behavioural alternations and the survey was anonymous and limited in its burden (i.e., topics and length). Data analysis All interviews were recorded and transcribed verbatim. The transcripts were coded inductively to identify different themes in the data. One researcher (DH) conducted the primary analysis, which were checked by a second coder (NE). All survey data were analysed with IBM SPSS Statistics (version 28). Descriptive statistics were used to describe the demographic characteristics of the participants. Continuous data are expressed as a mean with standard deviation (SD) or as the median with interquartile range (IQR) when appropriate. Categorical data are presented as valid percent (i.e., percentages when missing data are excluded from the calculations), except for data deriving from multiple answer questions; here absolute frequencies were used. One-way ANOVA or Kruskal–Wallis tests were used (depending on the distribution of the data) to determine whether patients’ mean monitor and blunting scores on the TMSI were associated with patients’ preferences for wanting to know predictions.
A national survey study among CKD patients and nephrologists in The Netherlands was conducted. First, patients’ attitudes towards different CPMs predicting the course of CKD were explored in semi-structured interviews. Next, two online surveys were developed and distributed: one for patients and one for nephrologists.
Patients with CKD were interviewed to explore their attitudes towards the use of CPMs in CKD practice. These interviews were held in the context of a larger study on the development of a CKD dashboard . During these interviews, two different predictions were introduced: 1) the prediction from the KFRE: a 2- and 5-year risk of progression to kidney failure for stages 3 to 5 CKD patients (in %), and 2) a prediction about the time until kidney failure (in years). Mock-ups were used to present these predictions in a similar lay-out to have patients focus on the meaning of the predictions rather than on how these were presented (Additional file ). Patients were asked to ‘think-out-loud’ and give their first impressions on the presented predictions. Patients were subsequently asked whether they would want to be provided with these predictions in (including reasons why), and how they would prefer to receive this information.
Two surveys were developed: one for CKD patients and one for nephrologists. Each survey started with an introductory text and an explanation of the definition of a CPM. This explanation was supplemented with an infographic to facilitate understanding (Additional file ). Both surveys consisted of questions assessing: 1) the current use of CPMs in Dutch CKD practice, 2) preferences for predictions in CKD, 3) preferences for predictions about CKD progression (to ESKD), and 4) barriers and facilitators for the adoption of CPMs in clinical practice. The patient surveys also included questions about educational levels, which was measured according to the International Standard Classification of Education and health literacy, which was measured with the Set of Brief Screening questions (SBSQ) . The SBSQ assesses perceived difficulties with health information based on three 5-point Likert scale statements ranging from 1–5. An average score of ≤ 3 indicates inadequate health literacy and a score of > 3 adequate health literacy. In the patient survey, the Threatening Medical Situations Inventory (TMSI) was used to assess whether patients handle medically threatening information with either monitoring (attending to the problem) or blunting (avoiding the problem) coping behaviour, since this may affect their views on receiving predictions . In the TMSI, patients are asked how they would handle hypothetical situations. They report on a 5-point Likert scale how likely it would be for them to apply three monitoring and three blunting strategies. Total scores for both the monitoring and blunting strategies are subsequently calculated (ranging from 6–30) . In the nephrologist survey, the Measurement Instrument for Determinants of Innovations (MIDI) was used to identify enablers for the adoption of CPMs in clinical practice . For three domains (the innovation, the user, and the organisation), nephrologists had to pick the two most important determinants that may facilitate the adoption of CPMs in clinical practice. Additional file shows the validated survey instruments used and the study-specific survey questions.
Both surveys were tested and amended for face validity by a: 1) communication scientist (CvU), 2) professor of medical decision-making (AS), 3) nephrologist (WB), and 4) cognitive psychologist specialised in communication research (AP). The patient survey was written at the B1 level of the common European framework of reference for languages (CEFRL) to ensure comprehensibility . It was also tested for face validity by five CKD patients recruited by the Dutch Kidney Patients Association.
Patients with CKD were recruited for the interviews by their nephrologists in two Dutch hospitals (St. Antonius hospital and Maasstad hospital) in February 2021. All participants gave informed consent. For the surveys, CKD patients and nephrologists were recruited from November 2021 until March 2022. Patients were approached via e-mail through the online platform of the Dutch Kidney Patients Association. The nephrologists were approached via e-mail through the online platform of the Dutch Federation for Nephrology. Both surveys were anonymous; no personal identifying information was registered. The patients and nephrologists who agreed to participate were asked to consent with the use of their answers for research and publication purposes when they started the survey. According to the Dutch medical research involving human subjects act, ethical approval was not required for the surveys because participants were not subjected to (medical) procedures or behavioural alternations and the survey was anonymous and limited in its burden (i.e., topics and length).
All interviews were recorded and transcribed verbatim. The transcripts were coded inductively to identify different themes in the data. One researcher (DH) conducted the primary analysis, which were checked by a second coder (NE). All survey data were analysed with IBM SPSS Statistics (version 28). Descriptive statistics were used to describe the demographic characteristics of the participants. Continuous data are expressed as a mean with standard deviation (SD) or as the median with interquartile range (IQR) when appropriate. Categorical data are presented as valid percent (i.e., percentages when missing data are excluded from the calculations), except for data deriving from multiple answer questions; here absolute frequencies were used. One-way ANOVA or Kruskal–Wallis tests were used (depending on the distribution of the data) to determine whether patients’ mean monitor and blunting scores on the TMSI were associated with patients’ preferences for wanting to know predictions.
Semi-structured interviews Seven CKD patients (four men, three women) with a mean age of 54 years (SD = 15) participated in the interviews. A total of five themes were identified in the data (shown in Table ). All illustrative quotations can be found in Additional file . More than half of the patients ( n = 5) understood the two predictions visualized in the mock-ups (theme one, understanding predictions about CKD progression). All but one patient indicated they wanted to know both predictions. Three patients preferred the prediction about the time until kidney failure (in years) over the KFRE, and two patients proposed combining them (theme two, preferences for predictions about CKD progression). In theme three ‘how predictions about CKD progression can help patients’, different reasons were mentioned why patients considered these predictions useful. Patients argued that the predictions could: 1) help them with life planning, 2), provide them with more clarity on the stage of their CKD), 3) help them focus on preserving their kidney function for as long as possible, and 4) provide them with comfort or consolation. Potential negative effects of discussing predictions about CKD progression (theme four) included: 1) the predictions could cause increased worrying, and 2) that individual trajectories may vary from the predictions. Lastly, patients indicated how to discuss predictions about CKD progression with patients (theme five). Several patients emphasised that these predictions can be very confrontational and stressed the importance of appropriate guidance and support when the predictions are discussed. Online surveys In total, 126 out of 407 patients responded to the survey invitation. This amounts to a response rate of 31%. Moreover, 50 out of 438 nephrologists responded to the survey invitation. This amounts to a response rate of 11%. The basic demographics of both the patients and nephrologists are presented in Table . The majority of patients ( n = 113, 90%) had been under nephrology care for at least 5 years. Most patients had undergone kidney transplantation ( n = 89, 71%) or were not yet on KRT ( n = 23, 19%). The SBSQ score for health literacy had a median of 4.7 (IQR = 0.7). Most patients ( n = 100, 79%) were highly educated. Mean scores on the TMSI for monitoring and blunting coping behaviours were comparable, with a mean of 19.4 and 18.6 respectively. At the time of the survey, the nephrologists had been practicing nephrology for a mean of 14.3 years (SD 9.1). Current use of, and experience with, CPMs Patients The majority of patients ( n = 111, 89%) reported that they had discussed predictions with their nephrologists. The most-commonly discussed predictions were: when they were expected to need KRT ( n = 81) and how rapidly their kidney function was expected to decline ( n = 68) illustrated in Fig. a. Only two patients indicated that, in retrospect, they would rather not have known these predictions. Patients indicated that discussing these predictions had helped them in the deliberation (pros vs cons) about their KRT options ( n = 77) and the realization that they had to make a KRT choice ( n = 71) (illustrated in Fig. b). Nephrologists Just over half of the nephrologists ( n = 26, 52%) indicated that they used CPMs at the time of the survey. Most nephrologists mentioned using a CPM predicting the risk of cardiovascular disease (CVD) ( n = 24), followed by a CPM predicting when patients will need KRT ( n = 8), a CPM predicting the risk of complications associated with different KRT modalities ( n = 3) and a CPM predicting how blood pressure affects kidney function ( n = 3). CPM’s predicting mortality before or after starting KRT were mentioned twice. Although a large proportion of nephrologists ( n = 21, 42%) did not use CPMs or did not know whether they had used them ( n = 3, 6%), all but two ( n = 48, 98%) discussed the expected kidney disease trajectory with patients. The majority ( n = 44, 92%) used graphs of the estimated glomerular filtration rate (eGFR) for this purpose. Nephrologists who did not use CPMs provided different reasons why. The most mentioned reason for not using CPMs was “not knowing any models” ( n = 11) followed by “not knowing enough about CPMs to use them” ( n = 6), “not knowing where to find them” ( n = 4), and “believing currently available CPMS are not reliable enough” ( n = 4). Less frequently mentioned were “not having enough time to use CPMs during consultations” ( n = 2), “believing currently available CPMs are impractical and difficult to use” ( n = 2) and “not seeing the point of using CPMs in providing CKD care” ( n = 1). Preferences for predictions in CKD Patients Most patients indicated that they wanted to know predictions about: 1) the risk of developing complications associated with the different KRT modalities ( n = 94, 78%), and 2) when they would need KRT ( n = 92, 77%) (illustrated in Fig. c). When asked to pick the most important prediction, the majority of patients chose “when I will need KRT in the future” ( n = 42, 61%). Predictions about the risk of dying before or after starting KRT were most frequently chosen as something patients did not want to know ( n = 27, 22%, and n = 26, 22%, respectively). Patients who wanted to know predictions had a significantly higher mean monitoring score compared to those who were neutral, or those who did not want to know these predictions. This was true for patients who desired knowing predictions concerning: 1) the risk of developing CVD (F (2,12) = [10.88], p = < 0.001), 2) when patients would need KRT (F (2,12) = [6.71], p = 0.002), and 3) the risk of dying before starting KRT (F (2,12) = [6.73], p = 0.002). The post hoc analyses are provided in Additional file . The mean monitoring scores of patients who wanted to know predictions about the risk of developing complications associated with the different KRT modalities, and the risk of dying after starting KRT did not significantly differ from mean monitoring scores of patients who were neutral, or who did not want to know these predictions. There were no significant differences between mean blunting scores as a function of patients’ preferences for wanting to know the different predictions in CKD. Regarding CPMs about CKD progression, 56 patients indicated that they perceived these predictions as confronting. Nevertheless, patients also agreed that such a prediction could help them to: 1) better know what they can expect (n = 75), 2) become better informed about their CKD ( n = 70), and 3) help with their (life) planning ( n = 65) (see Fig. d). When patients were shown the mock-up of the prediction from the KFRE, most patients considered it understandable ( n = 100, 80%). Likewise, most patients ( n = 105, 84%) understood the mock-up of the prediction in time to kidney failure (in years). The majority of patients wanted to know the prediction from the KFRE ( n = 89, 72%), 20 (16%) were neutral, and 14 (11%) did not want to know. Similarly, the majority of patients ( n = 96, 77%) wanted to know the prediction of time to kidney failure (in years), 10 (8%) were neutral, and 18 (15%) did not want to know. Fifty-four patients (45%) preferred the time to kidney failure (in years) prediction compared to 43 (36%) patients preferring the prediction from the KFRE; 24 patients (20%) were neutral. For both predictions, patients indicated that these could help them to: 1) better plan when they have to make a KRT decision, and 2) realize that a KRT decision needs to be made. Nephrologists The nephrologists indicated that they would most likely use a CPM to predict: 1) when CKD patients will need KRT, 2) how medication and blood pressure will affect a patient’s CKD trajectory, and 3) the risk of CVD in patients (illustrated in Fig. a). Twenty-three nephrologists (47%) picked a model predicting “when CKD patients will need KRT” as the most useful one. When the nephrologists were asked for what purpose they would want to develop a new CPM, 23 nephrologists (46%) chose “to better inform patients on the expected kidney function trajectory”. Other purposes for developing a new CPM included: “better being able to estimate the effects of treatment on slowing down kidney function deterioration” ( n = 15, 30%), “better being able to estimate when patients should start KRT education” ( n = 6, 12%), “better being able to estimate whether or not patients should start a certain kind of KRT” ( n = 4, 8%) and “better being able to estimate what the expected effects of a certain kind of KRT will be” ( n = 2, 4%). When they were asked whether they had already used the KFRE in the past, the majority ( n = 46, 92%) had not; mostly ( n = 38, 83%) because it was unknown to them. When they were asked whether they would use a CPM to predict the time to kidney failure in years (if available), more than half ( n = 28, 56%) indicated that they would. The prediction of time to kidney failure (in years) was preferred over the prediction from the KFRE by 31 nephrologists (62%). Four nephrologists explained that they expected patients would better understand a ‘time to’-prediction compared to a ‘risk of’-prediction. Barriers and facilitators for the adoption of CPMs in clinical practice Patients Sixty patients (49%) were neutral on the statement: “nephrologists should use CPMs during their consultations with patients”, 52 (41%) agreed, and 11 (9%) disagreed. Fifty-six patients (46%) wanted nephrologists to explain predictions during consultations, while 45 patients (37%) wanted to view predictions before their consultations so that they could discuss these with their nephrologist. Seventeen patients (14%) wanted to view predictions at any time, regardless of professional guidance. Nephrologists When the nephrologists were presented with statements arguing against the use of CPMs, the majority agreed that CPMs: 1) can give patients false expectations or a false sense of security ( n = 22, 50%), 2) don’t say anything about individual patients ( n = 20, 40%), and 3) are too time-consuming to use ( n = 18, 38%) (see Fig. b). Most nephrologists agreed ( n = 26, 52%) or completely agreed ( n = 11, 22%) that CPMs should only be used under professional guidance during consultations, rather than being available for patients at home. The nephrologists were asked to choose two factors from each of the domains of the MIDI (innovation, user, organisation) that they deemed most important in enabling successful use of a (new) prediction model (see Fig. c). For domain one (the innovation), the majority of nephrologists ( n = 25) considered the determinant “The prediction is clear and easily understandable for patients” as the most important determinant for successful adoption in clinical practice. For the second domain (the user), the majority ( n = 37) considered the determinant “If I believe the prediction from the CPM is clinically relevant” as the most important determinant. For the last domain (the organisation), most ( n = 33) considered the determinant “The CPM is integrated as a part of standard of care” as the most important determinant for adoption. All but two nephrologists ( n = 48, 96%) agreed that they would want to know the performance metrics of CPMs, such as confidence intervals, before they would consider using them. Twenty-three (46%) indicated that they would always discuss these performance metrics with their patients compared to 17 (34%) who would only discuss it with their patients if they believed the patients could understand these metrics and 9 (18%) who would refrain from discussing these metrics because they believed it would be too complicated for patients to understand. About two-thirds of the nephrologists ( n = 30, 60%) indicated that they would always discuss the uncertainty of an estimated prognosis with their patients, regardless of whether they would use a CPM to make these estimations. Eighteen nephrologists (36%) reported that they would discuss it “in most cases”, one nephrologist (2%) would discuss it “sometimes” and one (2%) would “never” discuss it with patients.
Seven CKD patients (four men, three women) with a mean age of 54 years (SD = 15) participated in the interviews. A total of five themes were identified in the data (shown in Table ). All illustrative quotations can be found in Additional file . More than half of the patients ( n = 5) understood the two predictions visualized in the mock-ups (theme one, understanding predictions about CKD progression). All but one patient indicated they wanted to know both predictions. Three patients preferred the prediction about the time until kidney failure (in years) over the KFRE, and two patients proposed combining them (theme two, preferences for predictions about CKD progression). In theme three ‘how predictions about CKD progression can help patients’, different reasons were mentioned why patients considered these predictions useful. Patients argued that the predictions could: 1) help them with life planning, 2), provide them with more clarity on the stage of their CKD), 3) help them focus on preserving their kidney function for as long as possible, and 4) provide them with comfort or consolation. Potential negative effects of discussing predictions about CKD progression (theme four) included: 1) the predictions could cause increased worrying, and 2) that individual trajectories may vary from the predictions. Lastly, patients indicated how to discuss predictions about CKD progression with patients (theme five). Several patients emphasised that these predictions can be very confrontational and stressed the importance of appropriate guidance and support when the predictions are discussed.
In total, 126 out of 407 patients responded to the survey invitation. This amounts to a response rate of 31%. Moreover, 50 out of 438 nephrologists responded to the survey invitation. This amounts to a response rate of 11%. The basic demographics of both the patients and nephrologists are presented in Table . The majority of patients ( n = 113, 90%) had been under nephrology care for at least 5 years. Most patients had undergone kidney transplantation ( n = 89, 71%) or were not yet on KRT ( n = 23, 19%). The SBSQ score for health literacy had a median of 4.7 (IQR = 0.7). Most patients ( n = 100, 79%) were highly educated. Mean scores on the TMSI for monitoring and blunting coping behaviours were comparable, with a mean of 19.4 and 18.6 respectively. At the time of the survey, the nephrologists had been practicing nephrology for a mean of 14.3 years (SD 9.1).
Patients The majority of patients ( n = 111, 89%) reported that they had discussed predictions with their nephrologists. The most-commonly discussed predictions were: when they were expected to need KRT ( n = 81) and how rapidly their kidney function was expected to decline ( n = 68) illustrated in Fig. a. Only two patients indicated that, in retrospect, they would rather not have known these predictions. Patients indicated that discussing these predictions had helped them in the deliberation (pros vs cons) about their KRT options ( n = 77) and the realization that they had to make a KRT choice ( n = 71) (illustrated in Fig. b). Nephrologists Just over half of the nephrologists ( n = 26, 52%) indicated that they used CPMs at the time of the survey. Most nephrologists mentioned using a CPM predicting the risk of cardiovascular disease (CVD) ( n = 24), followed by a CPM predicting when patients will need KRT ( n = 8), a CPM predicting the risk of complications associated with different KRT modalities ( n = 3) and a CPM predicting how blood pressure affects kidney function ( n = 3). CPM’s predicting mortality before or after starting KRT were mentioned twice. Although a large proportion of nephrologists ( n = 21, 42%) did not use CPMs or did not know whether they had used them ( n = 3, 6%), all but two ( n = 48, 98%) discussed the expected kidney disease trajectory with patients. The majority ( n = 44, 92%) used graphs of the estimated glomerular filtration rate (eGFR) for this purpose. Nephrologists who did not use CPMs provided different reasons why. The most mentioned reason for not using CPMs was “not knowing any models” ( n = 11) followed by “not knowing enough about CPMs to use them” ( n = 6), “not knowing where to find them” ( n = 4), and “believing currently available CPMS are not reliable enough” ( n = 4). Less frequently mentioned were “not having enough time to use CPMs during consultations” ( n = 2), “believing currently available CPMs are impractical and difficult to use” ( n = 2) and “not seeing the point of using CPMs in providing CKD care” ( n = 1).
The majority of patients ( n = 111, 89%) reported that they had discussed predictions with their nephrologists. The most-commonly discussed predictions were: when they were expected to need KRT ( n = 81) and how rapidly their kidney function was expected to decline ( n = 68) illustrated in Fig. a. Only two patients indicated that, in retrospect, they would rather not have known these predictions. Patients indicated that discussing these predictions had helped them in the deliberation (pros vs cons) about their KRT options ( n = 77) and the realization that they had to make a KRT choice ( n = 71) (illustrated in Fig. b).
Just over half of the nephrologists ( n = 26, 52%) indicated that they used CPMs at the time of the survey. Most nephrologists mentioned using a CPM predicting the risk of cardiovascular disease (CVD) ( n = 24), followed by a CPM predicting when patients will need KRT ( n = 8), a CPM predicting the risk of complications associated with different KRT modalities ( n = 3) and a CPM predicting how blood pressure affects kidney function ( n = 3). CPM’s predicting mortality before or after starting KRT were mentioned twice. Although a large proportion of nephrologists ( n = 21, 42%) did not use CPMs or did not know whether they had used them ( n = 3, 6%), all but two ( n = 48, 98%) discussed the expected kidney disease trajectory with patients. The majority ( n = 44, 92%) used graphs of the estimated glomerular filtration rate (eGFR) for this purpose. Nephrologists who did not use CPMs provided different reasons why. The most mentioned reason for not using CPMs was “not knowing any models” ( n = 11) followed by “not knowing enough about CPMs to use them” ( n = 6), “not knowing where to find them” ( n = 4), and “believing currently available CPMS are not reliable enough” ( n = 4). Less frequently mentioned were “not having enough time to use CPMs during consultations” ( n = 2), “believing currently available CPMs are impractical and difficult to use” ( n = 2) and “not seeing the point of using CPMs in providing CKD care” ( n = 1).
Patients Most patients indicated that they wanted to know predictions about: 1) the risk of developing complications associated with the different KRT modalities ( n = 94, 78%), and 2) when they would need KRT ( n = 92, 77%) (illustrated in Fig. c). When asked to pick the most important prediction, the majority of patients chose “when I will need KRT in the future” ( n = 42, 61%). Predictions about the risk of dying before or after starting KRT were most frequently chosen as something patients did not want to know ( n = 27, 22%, and n = 26, 22%, respectively). Patients who wanted to know predictions had a significantly higher mean monitoring score compared to those who were neutral, or those who did not want to know these predictions. This was true for patients who desired knowing predictions concerning: 1) the risk of developing CVD (F (2,12) = [10.88], p = < 0.001), 2) when patients would need KRT (F (2,12) = [6.71], p = 0.002), and 3) the risk of dying before starting KRT (F (2,12) = [6.73], p = 0.002). The post hoc analyses are provided in Additional file . The mean monitoring scores of patients who wanted to know predictions about the risk of developing complications associated with the different KRT modalities, and the risk of dying after starting KRT did not significantly differ from mean monitoring scores of patients who were neutral, or who did not want to know these predictions. There were no significant differences between mean blunting scores as a function of patients’ preferences for wanting to know the different predictions in CKD. Regarding CPMs about CKD progression, 56 patients indicated that they perceived these predictions as confronting. Nevertheless, patients also agreed that such a prediction could help them to: 1) better know what they can expect (n = 75), 2) become better informed about their CKD ( n = 70), and 3) help with their (life) planning ( n = 65) (see Fig. d). When patients were shown the mock-up of the prediction from the KFRE, most patients considered it understandable ( n = 100, 80%). Likewise, most patients ( n = 105, 84%) understood the mock-up of the prediction in time to kidney failure (in years). The majority of patients wanted to know the prediction from the KFRE ( n = 89, 72%), 20 (16%) were neutral, and 14 (11%) did not want to know. Similarly, the majority of patients ( n = 96, 77%) wanted to know the prediction of time to kidney failure (in years), 10 (8%) were neutral, and 18 (15%) did not want to know. Fifty-four patients (45%) preferred the time to kidney failure (in years) prediction compared to 43 (36%) patients preferring the prediction from the KFRE; 24 patients (20%) were neutral. For both predictions, patients indicated that these could help them to: 1) better plan when they have to make a KRT decision, and 2) realize that a KRT decision needs to be made. Nephrologists The nephrologists indicated that they would most likely use a CPM to predict: 1) when CKD patients will need KRT, 2) how medication and blood pressure will affect a patient’s CKD trajectory, and 3) the risk of CVD in patients (illustrated in Fig. a). Twenty-three nephrologists (47%) picked a model predicting “when CKD patients will need KRT” as the most useful one. When the nephrologists were asked for what purpose they would want to develop a new CPM, 23 nephrologists (46%) chose “to better inform patients on the expected kidney function trajectory”. Other purposes for developing a new CPM included: “better being able to estimate the effects of treatment on slowing down kidney function deterioration” ( n = 15, 30%), “better being able to estimate when patients should start KRT education” ( n = 6, 12%), “better being able to estimate whether or not patients should start a certain kind of KRT” ( n = 4, 8%) and “better being able to estimate what the expected effects of a certain kind of KRT will be” ( n = 2, 4%). When they were asked whether they had already used the KFRE in the past, the majority ( n = 46, 92%) had not; mostly ( n = 38, 83%) because it was unknown to them. When they were asked whether they would use a CPM to predict the time to kidney failure in years (if available), more than half ( n = 28, 56%) indicated that they would. The prediction of time to kidney failure (in years) was preferred over the prediction from the KFRE by 31 nephrologists (62%). Four nephrologists explained that they expected patients would better understand a ‘time to’-prediction compared to a ‘risk of’-prediction.
Most patients indicated that they wanted to know predictions about: 1) the risk of developing complications associated with the different KRT modalities ( n = 94, 78%), and 2) when they would need KRT ( n = 92, 77%) (illustrated in Fig. c). When asked to pick the most important prediction, the majority of patients chose “when I will need KRT in the future” ( n = 42, 61%). Predictions about the risk of dying before or after starting KRT were most frequently chosen as something patients did not want to know ( n = 27, 22%, and n = 26, 22%, respectively). Patients who wanted to know predictions had a significantly higher mean monitoring score compared to those who were neutral, or those who did not want to know these predictions. This was true for patients who desired knowing predictions concerning: 1) the risk of developing CVD (F (2,12) = [10.88], p = < 0.001), 2) when patients would need KRT (F (2,12) = [6.71], p = 0.002), and 3) the risk of dying before starting KRT (F (2,12) = [6.73], p = 0.002). The post hoc analyses are provided in Additional file . The mean monitoring scores of patients who wanted to know predictions about the risk of developing complications associated with the different KRT modalities, and the risk of dying after starting KRT did not significantly differ from mean monitoring scores of patients who were neutral, or who did not want to know these predictions. There were no significant differences between mean blunting scores as a function of patients’ preferences for wanting to know the different predictions in CKD. Regarding CPMs about CKD progression, 56 patients indicated that they perceived these predictions as confronting. Nevertheless, patients also agreed that such a prediction could help them to: 1) better know what they can expect (n = 75), 2) become better informed about their CKD ( n = 70), and 3) help with their (life) planning ( n = 65) (see Fig. d). When patients were shown the mock-up of the prediction from the KFRE, most patients considered it understandable ( n = 100, 80%). Likewise, most patients ( n = 105, 84%) understood the mock-up of the prediction in time to kidney failure (in years). The majority of patients wanted to know the prediction from the KFRE ( n = 89, 72%), 20 (16%) were neutral, and 14 (11%) did not want to know. Similarly, the majority of patients ( n = 96, 77%) wanted to know the prediction of time to kidney failure (in years), 10 (8%) were neutral, and 18 (15%) did not want to know. Fifty-four patients (45%) preferred the time to kidney failure (in years) prediction compared to 43 (36%) patients preferring the prediction from the KFRE; 24 patients (20%) were neutral. For both predictions, patients indicated that these could help them to: 1) better plan when they have to make a KRT decision, and 2) realize that a KRT decision needs to be made.
The nephrologists indicated that they would most likely use a CPM to predict: 1) when CKD patients will need KRT, 2) how medication and blood pressure will affect a patient’s CKD trajectory, and 3) the risk of CVD in patients (illustrated in Fig. a). Twenty-three nephrologists (47%) picked a model predicting “when CKD patients will need KRT” as the most useful one. When the nephrologists were asked for what purpose they would want to develop a new CPM, 23 nephrologists (46%) chose “to better inform patients on the expected kidney function trajectory”. Other purposes for developing a new CPM included: “better being able to estimate the effects of treatment on slowing down kidney function deterioration” ( n = 15, 30%), “better being able to estimate when patients should start KRT education” ( n = 6, 12%), “better being able to estimate whether or not patients should start a certain kind of KRT” ( n = 4, 8%) and “better being able to estimate what the expected effects of a certain kind of KRT will be” ( n = 2, 4%). When they were asked whether they had already used the KFRE in the past, the majority ( n = 46, 92%) had not; mostly ( n = 38, 83%) because it was unknown to them. When they were asked whether they would use a CPM to predict the time to kidney failure in years (if available), more than half ( n = 28, 56%) indicated that they would. The prediction of time to kidney failure (in years) was preferred over the prediction from the KFRE by 31 nephrologists (62%). Four nephrologists explained that they expected patients would better understand a ‘time to’-prediction compared to a ‘risk of’-prediction.
Patients Sixty patients (49%) were neutral on the statement: “nephrologists should use CPMs during their consultations with patients”, 52 (41%) agreed, and 11 (9%) disagreed. Fifty-six patients (46%) wanted nephrologists to explain predictions during consultations, while 45 patients (37%) wanted to view predictions before their consultations so that they could discuss these with their nephrologist. Seventeen patients (14%) wanted to view predictions at any time, regardless of professional guidance. Nephrologists When the nephrologists were presented with statements arguing against the use of CPMs, the majority agreed that CPMs: 1) can give patients false expectations or a false sense of security ( n = 22, 50%), 2) don’t say anything about individual patients ( n = 20, 40%), and 3) are too time-consuming to use ( n = 18, 38%) (see Fig. b). Most nephrologists agreed ( n = 26, 52%) or completely agreed ( n = 11, 22%) that CPMs should only be used under professional guidance during consultations, rather than being available for patients at home. The nephrologists were asked to choose two factors from each of the domains of the MIDI (innovation, user, organisation) that they deemed most important in enabling successful use of a (new) prediction model (see Fig. c). For domain one (the innovation), the majority of nephrologists ( n = 25) considered the determinant “The prediction is clear and easily understandable for patients” as the most important determinant for successful adoption in clinical practice. For the second domain (the user), the majority ( n = 37) considered the determinant “If I believe the prediction from the CPM is clinically relevant” as the most important determinant. For the last domain (the organisation), most ( n = 33) considered the determinant “The CPM is integrated as a part of standard of care” as the most important determinant for adoption. All but two nephrologists ( n = 48, 96%) agreed that they would want to know the performance metrics of CPMs, such as confidence intervals, before they would consider using them. Twenty-three (46%) indicated that they would always discuss these performance metrics with their patients compared to 17 (34%) who would only discuss it with their patients if they believed the patients could understand these metrics and 9 (18%) who would refrain from discussing these metrics because they believed it would be too complicated for patients to understand. About two-thirds of the nephrologists ( n = 30, 60%) indicated that they would always discuss the uncertainty of an estimated prognosis with their patients, regardless of whether they would use a CPM to make these estimations. Eighteen nephrologists (36%) reported that they would discuss it “in most cases”, one nephrologist (2%) would discuss it “sometimes” and one (2%) would “never” discuss it with patients.
Sixty patients (49%) were neutral on the statement: “nephrologists should use CPMs during their consultations with patients”, 52 (41%) agreed, and 11 (9%) disagreed. Fifty-six patients (46%) wanted nephrologists to explain predictions during consultations, while 45 patients (37%) wanted to view predictions before their consultations so that they could discuss these with their nephrologist. Seventeen patients (14%) wanted to view predictions at any time, regardless of professional guidance.
When the nephrologists were presented with statements arguing against the use of CPMs, the majority agreed that CPMs: 1) can give patients false expectations or a false sense of security ( n = 22, 50%), 2) don’t say anything about individual patients ( n = 20, 40%), and 3) are too time-consuming to use ( n = 18, 38%) (see Fig. b). Most nephrologists agreed ( n = 26, 52%) or completely agreed ( n = 11, 22%) that CPMs should only be used under professional guidance during consultations, rather than being available for patients at home. The nephrologists were asked to choose two factors from each of the domains of the MIDI (innovation, user, organisation) that they deemed most important in enabling successful use of a (new) prediction model (see Fig. c). For domain one (the innovation), the majority of nephrologists ( n = 25) considered the determinant “The prediction is clear and easily understandable for patients” as the most important determinant for successful adoption in clinical practice. For the second domain (the user), the majority ( n = 37) considered the determinant “If I believe the prediction from the CPM is clinically relevant” as the most important determinant. For the last domain (the organisation), most ( n = 33) considered the determinant “The CPM is integrated as a part of standard of care” as the most important determinant for adoption. All but two nephrologists ( n = 48, 96%) agreed that they would want to know the performance metrics of CPMs, such as confidence intervals, before they would consider using them. Twenty-three (46%) indicated that they would always discuss these performance metrics with their patients compared to 17 (34%) who would only discuss it with their patients if they believed the patients could understand these metrics and 9 (18%) who would refrain from discussing these metrics because they believed it would be too complicated for patients to understand. About two-thirds of the nephrologists ( n = 30, 60%) indicated that they would always discuss the uncertainty of an estimated prognosis with their patients, regardless of whether they would use a CPM to make these estimations. Eighteen nephrologists (36%) reported that they would discuss it “in most cases”, one nephrologist (2%) would discuss it “sometimes” and one (2%) would “never” discuss it with patients.
We conducted a national survey study to explore the current use of CPMs in Dutch CKD practice and to identify patients’ and nephrologists’ needs and preferences regarding the use of CPMs, as well as barriers and facilitators for the adoption of CPMs in clinical practice. Even though previous studies suggest that CPMs are used to a limited extent in clinical practice , more than half of the nephrologists who participated in our survey reported using CPMs. Likewise, the majority of patients reported that they had discussed predictions with their nephrologist in the past; mostly predictions about their risk of progression to kidney failure. On the contrary, nephrologists reported discussing a CPM for the risk of CVD in patients most frequently. This discrepancy could be explained by the fact that almost all nephrologists reported discussing expected kidney disease trajectories with their patients, and that most of them used graphs of their patients’ eGFR (not a CPM) for this purpose. Patients who participated in this study may have misinterpreted these extrapolations as predictions made with CPMs. For patients, knowing the details of the origin of the prediction might not matter much. However, nephrologists should be aware of this discrepancy when they discuss expected kidney disease trajectories with their patients, since both nephrologists and patients tend to overestimate the risk of progression to ESKD . The majority of both patients and nephrologists advocated for the use of CPMs in CKD practice. These findings are consistent with previous studies . Even though a large proportion of patients considered predictions confrontational (particularly predictions about CKD progression), almost none of them regretted discussing predictions with their nephrologists in the past. Reasons for nephrologists why they did not currently use CPMs were most often related to their limited knowledge about, or unfamiliarity with, existing models. Barriers relating to intrinsic motivation, user friendliness or reliability, as often mentioned in the literature , were infrequently reported. Perhaps these barriers are overvalued when implementation initiatives are formulated; hindering the widespread adoption of CPMs in CKD practice. Instead, we should focus more on the facilitators for the adoption of CPMs in clinical practice. In this study, facilitators for the adoption of CPMs related to presenting CPMs in a clear and understandable way, incorporating them as a part of standard care, and the CPMs being clinically relevant. Even though previous studies suggest that nephrologists and patients prioritise different treatment outcomes , both patients and nephrologists considered CPMs predicting CKD progression as the most relevant prediction, preferably predicting the time to KRT (in years) instead of a 2- and 5-year risk (in %). Patients indicated that this prediction could help them better plan when they have to make a KRT decision and realize that a KRT decision has to be made. The latter is an important enabler for patient empowerment in starting a shared decision-making process . When we explored patients’ normative beliefs about whether or not nephrologists should use CPMs during consultations, most were neutral or agreed that they should. However, it should be noted that there was a small proportion of patients who did not want to know any predictions when we explored their preferences for both CPMs in general, and CPMs related to CKD progression. This is especially relevant considering that the participating patients are potentially taking on a more active role in treatment decision-making compared to the general patient population (since they were highly educated, had high health literacy and were recruited from the Dutch Kidney Patients Association). The actual number of patients that do not want to know these predictions could potentially be higher in clinical practice. Although we did identify that higher monitor scores might be associated with wanting to know certain predictions, we did not find higher monitor scores in our study population when compared to their individual blunting scores, or to scores from other studies . Similar to others who studied patient preferences for receiving prognostic information , we propose that nephrologists simply ask, and provide patients with the opportunity to make their own decisions about whether or not they want predictive information to be shared with them. In addition to the highly educated patient population, the majority of the patients included in this study were patients who had received a kidney transplant and were under treatment for more than 5 years with their nephrologist. This affects generalization of the results towards the whole CKD population. Hypothetically, patients earlier in their disease phase might have different information needs regarding the use of CPMs. Additionally, participating patients might have discussed the predictions regarding CKD progression a longer time ago, increasing changes on recall bias. For the clinician’s survey, issues with generalization should also be noted; these survey results may not be indicative for all Dutch nephrologists. Since the response rate to the survey was low, we cannot exclude non-response bias. Nephrologists who were willing to fill in the survey may hold more positive attitudes towards CPMs than nephrologists who didn’t. We are among the first to provide quantitative data on what both patients and nephrologists prefer regarding the use and purpose of CPMs, and what predictions they prioritise. Moreover, we collected information on important determinants for the successful adoption of CPMs in clinical practice, which may be used to guide the implementation of CPMs. In addition, researchers and developers can use our findings for improving existing CPMs or for developing new CPMs. When the latter is considered, our study shows that patients and nephrologists prefer a ‘time to kidney failure’ prediction, rather than a ‘risk of progression to kidney failure’ prediction. This study focused on currently available CPMs in CKD. Future research may explore newly developed CPMs, such as CPMs predicting patient reported outcomes.
In this study, both nephrologists and the majority of patients want to discuss CPMs in Dutch CKD practice, especially those that predict CKD progression. Validated and freely available CPMs, that largely meet the needs and preferences expressed by patients and nephrologists in this study, already exist (e.g. the KFRE). However, these CPMs appear to be underused due to lack of knowledge regarding where to find them and how to use them meaningfully. We should focus on improving the accessibility of these CPMs and provide guidance on how to communicate the predictions effectively. Additionally, whether or not patients want to hear particular predictions varies among individual patients, and their preferences should therefore be explored during consultations.
Additional file 1: Figure S1. Mock-ups of two predictions of models predicting CKD progression (translated from Dutch). Additional file 2: Figure S2. Infographic explaining a clinical prediction model (in Dutch). Additional file 3: Table S1. Content of the online surveys for patients and nephrologists. Additional file 4: Table S2. Identified themes and illustrative quotes from patient interviews. Additional file 5: Box S1. Post-hoc analysis of coping strategies in relation to preferences regarding CPMs.
|
Diversity of bacterial community in the rhizosphere and bulk soil of | 5048523c-a147-4844-8a92-eb059089346c | 10035820 | Microbiology[mh] | Artemisia annua is widely grown in different parts of the world as a cheap source of the antimalarial compounds such as artemisinin, flavonoids, aromatic oils and polysaccharides . In Uganda, A . annua was introduced around 2003 and is mainly grown in Wakiso, Kaberamaido, Kapchorwa, Rukungiri, Kabarole and Kabale districts. Itis cultivated as a monocrop or intercropped with beans. The main districts where the crop is grown include. The content of active compounds varies greatly depending on the geographical location. The highland areas produce A . annua with more active compound than low land areas . For instance, they observed that artemisinin and total flavonoids levels were higher in samples obtained from high land areas (Kabarole and Kabale) compared to those obtained from lowland regions (wakiso) i.e. 0.8%, 0.5% Vs 0.4% and 2.6%, 2.55% Vs 1.5% respectively. However, the artemisinin concentration is very low compared to other parts of the world that produce A . annua with upto 2% artemisinin . Improving the concentrations of the antimalarial compounds in A . annua grown in Uganda is therefore an important area of investigation. Various plant growth promoting bacteria (PGPB) such as Azotobacter , Azospirillum , Bacillus and Pseudomonas have been reported to increase the concentration of artemisinin elsewhere. However, there is no study that highlights the rhizobacterial community of A . annua yet understanding the rhizobacterial community of a given plant species is a vital when considering the use of rhizobacteria as plant growth promoters . This study therefore aimed at profiling the diversity of bacterial community in the rhizosphere and bulk soil of A . annua grown in highlands of Uganda as basis for the use of microbial inoculants to enhance its antimalarial compounds.
Study sites and sample collection The study was conducted in 2 highlands of Uganda i.e. Kabale (South Western Uganda) and Kabarole (Western Uganda). These are areas producing large volumes of artemisinin ranging from 0.5 and 0.8% . The altitude of Kabarole and Kabale district is 1300–3800 meters and 2,000 meters above sea level respectively. Kabarole and Kabale district receive annual rainfall ranging from 1,200 mm– 1,500 mm and 800 mm– 1000 mm respectively . Soil samples were collected at the time of harvesting A . annua . Rhizosphere and bulk soils were sampled from the four existing cropping types of (i) intercrop of A . annua and beans (AA+B), (ii) beans alone (B), (iii) A . annua alone (A) and iv) Control-No crop grown, (N AA/B)). For the physicochemical analysis, composite samples (each 1.5 kg) were picked from the top soil (0–15 cm). For each of the 4 cropping type, 4 farms in each of the 2 districts having similar treatments were considered as replicates, thus bringing the total to 32 samples. The composite sample consisted of 10 cores obtained using zigzag technique. All 32 samples were placed in clean labelled sealable plastic bags and transported to the laboratory with in a day of collection. The soil was dried and sieved using a 2 mm sieve. For DNA extraction and molecular analyses, samples were put in collecting tubes (2.2g each) and were stored at −80°C. For each of the 4 cropping types, 4 farms in each of the 2 districts having similar treatments were considered as replicates and in each replicate 4 plants were selected to obtain rhizospheric sample (where A . annua plants were grown) or 4 cores were selected for obtaining bulk samples (where there was no A . annua ). Laboratory analysis Physicochemical analysis Various soil properties such as pH, texture, organic matter, N, P and K were analyzed. Soil pH was measured with a pH meter in water (1:10 w/v). Soil texture was determined using Bouyoucos method . The organic matter was determined using Walkley and Black method . The total nitrogen content was determined using Kjeldahl method after total digestion of soil sample. The soluble phosphorus was determined using Ascorbic acid method . Exchangeable potassium was extracted in ammonium acetate and measured using a flame photometer . DNA extraction Genomic DNA (gDNA) was isolated from each soil sample (0.25g) using Qiagen DNeasy ® Power soil ® kit (Germany). After each extraction, agarose electrophoresis was conducted to detect the presence of gDNA. Later, the concentration of gDNA was quantified using Qubit 3 Fluorometer (Singapore) and an average of 6.8 ng/μl was obtained. Thereafter, 14 μls of gDNA was sent to Macrogen for amplicon based metagenomics sequencing. Assembly . Longer assembled reads were obtained using the FLASH (1.2.11) program . On average for each sample, the results of assembly were 43,859,440 (total bases), 96,482 (read count), 0.00 (N, %), 56.61(GC, %), 97.43 (Q20, %) and 90.43 (Q30, %). Preprocessing and clustering . This was carried out using two programs i.e. CD-HIT-OTU MiSeq/FLX and rDNA tools-PACBIO . The steps involved identifying of chimeric reads and removing them, filtering out of short reads, trimming of extra-long tails. Filtered reads were clustered at 100% identity using CD-HIT-DUP. Secondary clusters were recruited into primary clusters. Noise sequences in clusters of size X or below were removed and X was statistically calculated. Remaining representative reads from non-chimeric clusters were clustered using a greedy algorithm into OTUs at a user-specified OTU cut off (e.g. 97% ID at species level). The results of preprocessing are summarized in . Taxonomic assignment and diversity statistics . This was carried out using the program QIIME . A reasonable number of reads was used in analysis since the rarefaction curves for the various samples became flatter to the right. Representative sequences from each OTUs were used to assign taxonomy. Furthermore, to identify differences between various treatments and the two study sites, data was analyzed statistically by one-way analysis of variance and independent student t- test (p<0.05) using SPSS 21.0 Software (SPSS, Chicago, IL, USA). In addition, Principal Component Analysis (PCA) was conducted to find the lowest number of factors which could account for the variability in the original variables that were associated with those factors. Genera abundance . A total of 626 bacterial genera (abundance ≥ 0.01%) were observed among the 3 most prevalent phylum (proteobacteria, acidobacteria and actinobacteria) common to both districts. However, PCA was conducted on 31 genera species that showed higher abundance (≥ 0.2%) and the physiochemical properties of soil. The suitability of data for factor analysis was assessed by running the correlation matrix which revealed the presence of many coefficients of 0.3 and above. Furthermore, since variables were of various units (%, mg/l), they were first normalized in order to bring the values of the different variables within the comparable range. This was done by subtracting the mean from the observed value and dividing by the standard deviation for each 42 variables using the following formula. Normalized value = Observed value − Mean Standard deviation Having standardized the data, weights were attached using Principal Component Analysis (PCA) for all the soil properties in STATA 15 Statistical Software for assigning the weights. The loadings from the first component of PCA are used as the weights for the indicators. The assigned weights varied between -1 and +1, sign with the magnitude of the weights describing the contribution of each variable to the overall value of the index for soil properties. Two statistical tests were first conducted in order to determine the suitability of PCA. First, the Kaisers-Meyer-Olkin (KMO) measure of sampling adequacy score whose value of 0.811 was above the minimum recommended level of 0.60 and Bartlett’s Test of sphericity was 0.000<0.01 implying that it was statistically significant thus supporting the factorability of the correlation matrix. Determination of the number of components to retain was based on Eigen value greater than 1.0 and the scree plot. Principal component Analysis revealed the presence of nine components with eigenvalues exceeding 1 and explained 88.32% as a cumulative of the variance. However, five components were first used in analysis and the components that contributed smaller amounts of the total variance were left out. Then finally, using Catell’s Scree plot , it was decided to retain two components for further investigation. The two-component solution explained a total of 56.34% of the variance, with Component 1 contributing most variance (47.61%) and Component 2 explaining the least (8.73%). To aid in the interpretation of these two components, varimax rotation was performed. Using information about how much of the variance is explained by each item (the communalities table), soil components that had small values less than 0.3 and thus failed to load on the components obtained, an indication that they did not fit well with other items in its component were discarded. Furthermore, a Rotated Factor Matrix table was constructed to tell us what the factor loadings looked like after rotation.
The study was conducted in 2 highlands of Uganda i.e. Kabale (South Western Uganda) and Kabarole (Western Uganda). These are areas producing large volumes of artemisinin ranging from 0.5 and 0.8% . The altitude of Kabarole and Kabale district is 1300–3800 meters and 2,000 meters above sea level respectively. Kabarole and Kabale district receive annual rainfall ranging from 1,200 mm– 1,500 mm and 800 mm– 1000 mm respectively . Soil samples were collected at the time of harvesting A . annua . Rhizosphere and bulk soils were sampled from the four existing cropping types of (i) intercrop of A . annua and beans (AA+B), (ii) beans alone (B), (iii) A . annua alone (A) and iv) Control-No crop grown, (N AA/B)). For the physicochemical analysis, composite samples (each 1.5 kg) were picked from the top soil (0–15 cm). For each of the 4 cropping type, 4 farms in each of the 2 districts having similar treatments were considered as replicates, thus bringing the total to 32 samples. The composite sample consisted of 10 cores obtained using zigzag technique. All 32 samples were placed in clean labelled sealable plastic bags and transported to the laboratory with in a day of collection. The soil was dried and sieved using a 2 mm sieve. For DNA extraction and molecular analyses, samples were put in collecting tubes (2.2g each) and were stored at −80°C. For each of the 4 cropping types, 4 farms in each of the 2 districts having similar treatments were considered as replicates and in each replicate 4 plants were selected to obtain rhizospheric sample (where A . annua plants were grown) or 4 cores were selected for obtaining bulk samples (where there was no A . annua ).
Physicochemical analysis Various soil properties such as pH, texture, organic matter, N, P and K were analyzed. Soil pH was measured with a pH meter in water (1:10 w/v). Soil texture was determined using Bouyoucos method . The organic matter was determined using Walkley and Black method . The total nitrogen content was determined using Kjeldahl method after total digestion of soil sample. The soluble phosphorus was determined using Ascorbic acid method . Exchangeable potassium was extracted in ammonium acetate and measured using a flame photometer . DNA extraction Genomic DNA (gDNA) was isolated from each soil sample (0.25g) using Qiagen DNeasy ® Power soil ® kit (Germany). After each extraction, agarose electrophoresis was conducted to detect the presence of gDNA. Later, the concentration of gDNA was quantified using Qubit 3 Fluorometer (Singapore) and an average of 6.8 ng/μl was obtained. Thereafter, 14 μls of gDNA was sent to Macrogen for amplicon based metagenomics sequencing. Assembly . Longer assembled reads were obtained using the FLASH (1.2.11) program . On average for each sample, the results of assembly were 43,859,440 (total bases), 96,482 (read count), 0.00 (N, %), 56.61(GC, %), 97.43 (Q20, %) and 90.43 (Q30, %). Preprocessing and clustering . This was carried out using two programs i.e. CD-HIT-OTU MiSeq/FLX and rDNA tools-PACBIO . The steps involved identifying of chimeric reads and removing them, filtering out of short reads, trimming of extra-long tails. Filtered reads were clustered at 100% identity using CD-HIT-DUP. Secondary clusters were recruited into primary clusters. Noise sequences in clusters of size X or below were removed and X was statistically calculated. Remaining representative reads from non-chimeric clusters were clustered using a greedy algorithm into OTUs at a user-specified OTU cut off (e.g. 97% ID at species level). The results of preprocessing are summarized in . Taxonomic assignment and diversity statistics . This was carried out using the program QIIME . A reasonable number of reads was used in analysis since the rarefaction curves for the various samples became flatter to the right. Representative sequences from each OTUs were used to assign taxonomy. Furthermore, to identify differences between various treatments and the two study sites, data was analyzed statistically by one-way analysis of variance and independent student t- test (p<0.05) using SPSS 21.0 Software (SPSS, Chicago, IL, USA). In addition, Principal Component Analysis (PCA) was conducted to find the lowest number of factors which could account for the variability in the original variables that were associated with those factors. Genera abundance . A total of 626 bacterial genera (abundance ≥ 0.01%) were observed among the 3 most prevalent phylum (proteobacteria, acidobacteria and actinobacteria) common to both districts. However, PCA was conducted on 31 genera species that showed higher abundance (≥ 0.2%) and the physiochemical properties of soil. The suitability of data for factor analysis was assessed by running the correlation matrix which revealed the presence of many coefficients of 0.3 and above. Furthermore, since variables were of various units (%, mg/l), they were first normalized in order to bring the values of the different variables within the comparable range. This was done by subtracting the mean from the observed value and dividing by the standard deviation for each 42 variables using the following formula. Normalized value = Observed value − Mean Standard deviation Having standardized the data, weights were attached using Principal Component Analysis (PCA) for all the soil properties in STATA 15 Statistical Software for assigning the weights. The loadings from the first component of PCA are used as the weights for the indicators. The assigned weights varied between -1 and +1, sign with the magnitude of the weights describing the contribution of each variable to the overall value of the index for soil properties. Two statistical tests were first conducted in order to determine the suitability of PCA. First, the Kaisers-Meyer-Olkin (KMO) measure of sampling adequacy score whose value of 0.811 was above the minimum recommended level of 0.60 and Bartlett’s Test of sphericity was 0.000<0.01 implying that it was statistically significant thus supporting the factorability of the correlation matrix. Determination of the number of components to retain was based on Eigen value greater than 1.0 and the scree plot. Principal component Analysis revealed the presence of nine components with eigenvalues exceeding 1 and explained 88.32% as a cumulative of the variance. However, five components were first used in analysis and the components that contributed smaller amounts of the total variance were left out. Then finally, using Catell’s Scree plot , it was decided to retain two components for further investigation. The two-component solution explained a total of 56.34% of the variance, with Component 1 contributing most variance (47.61%) and Component 2 explaining the least (8.73%). To aid in the interpretation of these two components, varimax rotation was performed. Using information about how much of the variance is explained by each item (the communalities table), soil components that had small values less than 0.3 and thus failed to load on the components obtained, an indication that they did not fit well with other items in its component were discarded. Furthermore, a Rotated Factor Matrix table was constructed to tell us what the factor loadings looked like after rotation.
Various soil properties such as pH, texture, organic matter, N, P and K were analyzed. Soil pH was measured with a pH meter in water (1:10 w/v). Soil texture was determined using Bouyoucos method . The organic matter was determined using Walkley and Black method . The total nitrogen content was determined using Kjeldahl method after total digestion of soil sample. The soluble phosphorus was determined using Ascorbic acid method . Exchangeable potassium was extracted in ammonium acetate and measured using a flame photometer .
Genomic DNA (gDNA) was isolated from each soil sample (0.25g) using Qiagen DNeasy ® Power soil ® kit (Germany). After each extraction, agarose electrophoresis was conducted to detect the presence of gDNA. Later, the concentration of gDNA was quantified using Qubit 3 Fluorometer (Singapore) and an average of 6.8 ng/μl was obtained. Thereafter, 14 μls of gDNA was sent to Macrogen for amplicon based metagenomics sequencing. Assembly . Longer assembled reads were obtained using the FLASH (1.2.11) program . On average for each sample, the results of assembly were 43,859,440 (total bases), 96,482 (read count), 0.00 (N, %), 56.61(GC, %), 97.43 (Q20, %) and 90.43 (Q30, %). Preprocessing and clustering . This was carried out using two programs i.e. CD-HIT-OTU MiSeq/FLX and rDNA tools-PACBIO . The steps involved identifying of chimeric reads and removing them, filtering out of short reads, trimming of extra-long tails. Filtered reads were clustered at 100% identity using CD-HIT-DUP. Secondary clusters were recruited into primary clusters. Noise sequences in clusters of size X or below were removed and X was statistically calculated. Remaining representative reads from non-chimeric clusters were clustered using a greedy algorithm into OTUs at a user-specified OTU cut off (e.g. 97% ID at species level). The results of preprocessing are summarized in . Taxonomic assignment and diversity statistics . This was carried out using the program QIIME . A reasonable number of reads was used in analysis since the rarefaction curves for the various samples became flatter to the right. Representative sequences from each OTUs were used to assign taxonomy. Furthermore, to identify differences between various treatments and the two study sites, data was analyzed statistically by one-way analysis of variance and independent student t- test (p<0.05) using SPSS 21.0 Software (SPSS, Chicago, IL, USA). In addition, Principal Component Analysis (PCA) was conducted to find the lowest number of factors which could account for the variability in the original variables that were associated with those factors. Genera abundance . A total of 626 bacterial genera (abundance ≥ 0.01%) were observed among the 3 most prevalent phylum (proteobacteria, acidobacteria and actinobacteria) common to both districts. However, PCA was conducted on 31 genera species that showed higher abundance (≥ 0.2%) and the physiochemical properties of soil. The suitability of data for factor analysis was assessed by running the correlation matrix which revealed the presence of many coefficients of 0.3 and above. Furthermore, since variables were of various units (%, mg/l), they were first normalized in order to bring the values of the different variables within the comparable range. This was done by subtracting the mean from the observed value and dividing by the standard deviation for each 42 variables using the following formula. Normalized value = Observed value − Mean Standard deviation Having standardized the data, weights were attached using Principal Component Analysis (PCA) for all the soil properties in STATA 15 Statistical Software for assigning the weights. The loadings from the first component of PCA are used as the weights for the indicators. The assigned weights varied between -1 and +1, sign with the magnitude of the weights describing the contribution of each variable to the overall value of the index for soil properties. Two statistical tests were first conducted in order to determine the suitability of PCA. First, the Kaisers-Meyer-Olkin (KMO) measure of sampling adequacy score whose value of 0.811 was above the minimum recommended level of 0.60 and Bartlett’s Test of sphericity was 0.000<0.01 implying that it was statistically significant thus supporting the factorability of the correlation matrix. Determination of the number of components to retain was based on Eigen value greater than 1.0 and the scree plot. Principal component Analysis revealed the presence of nine components with eigenvalues exceeding 1 and explained 88.32% as a cumulative of the variance. However, five components were first used in analysis and the components that contributed smaller amounts of the total variance were left out. Then finally, using Catell’s Scree plot , it was decided to retain two components for further investigation. The two-component solution explained a total of 56.34% of the variance, with Component 1 contributing most variance (47.61%) and Component 2 explaining the least (8.73%). To aid in the interpretation of these two components, varimax rotation was performed. Using information about how much of the variance is explained by each item (the communalities table), soil components that had small values less than 0.3 and thus failed to load on the components obtained, an indication that they did not fit well with other items in its component were discarded. Furthermore, a Rotated Factor Matrix table was constructed to tell us what the factor loadings looked like after rotation.
Physiochemical properties Results obtained from analysis of physical and chemical properties of soil are presented in . The soils in Kabale are clay loam and have neutral pH (6.64) while Kabarole soil are sandy loam and have slightly alkaline pH (7.48). The content of Nitrogen, Potassium, Sodium and Silt is similar in both Kabale and Kabarole. However, the content of Phosphorus, Organic Carbon, Magnesium and Calcium were higher in Kabarole soils than Kabale soils. Bulk soils had lower physiochemical properties than rhizospheric soils in Kabale. Phyla abundance The phylum Proteobacteria was the most prevalent followed by Acidobacteria and Actinobacteria ( ). Comparing the two districts, Kabale soils had higher diversity of bacteria than Kabarole and there was significant difference (P≤0.5) among all phyla except phylum Proteobacteria. Kabale soil samples contained more Acidobacteria, Bacteroidetes, Veruccomicrobia, Gemmatimonadetes and Firmicutes while Kabarole soil samples contained more Actinobacteria and other phyla like Plactomycetes. With respect to the various soil treatments, there was no significant difference (P≤ 0.05) among the abundance of various phyla in Kabale. However, in Kabarole, there was significant differences (P≤ 0.05) in the abundance of phyla actinobacteria (14.85%) and verrucomicrobia (7.01%) in soils with artemisia only and the other treatments. Comparing bulk and rhizospheric soils, Proteobacteria were dorminant in both soils but acidobacteria and actinobacteria were more dorminate in bulk soils of Kabale and Kabarole respectively. Genus abundance The abundance of various bacterial genera (≥ 0.2%) in the bulk and rhizospheric soil of the most prevalent phylum (Proteobacteria, acidobacteria and actinobacteria) are shown in . Most genera in Kabale soils did not show any significant differences. However, most genera in Kabarole soils showed significant differences. In bulk soil, the genera that showed significant differences were po, br, no, str, ga and sol. In rhizospheric soils, Many genera (sps, ac, ram, ma, rhi, aci, ed, si, oc, pa and pse) showed significant differences with the bulk soil but genera sps, ram, pa and pse were observed to show significant in differences in both Kabale and Kabarole soils. Results of PCA of various soil components (genera and physiochemical properties) are shown in . Seven genera (sps, Lu, he, st, ma, spm and pse) had positively higher loading with the second component. Twenty nine soil components had higher loading with the first component and 18 (sand, Ca, Mg, P, N, OCa, pH, Sol, ga, Str, Vi, acis, pe, az, ram, ral, P and Ly) had positive factor loading while 11 (ac, ni, ps, aci, rhi, ed, si, oc, pa, pa_a and clay) had negative factor loading.
Results obtained from analysis of physical and chemical properties of soil are presented in . The soils in Kabale are clay loam and have neutral pH (6.64) while Kabarole soil are sandy loam and have slightly alkaline pH (7.48). The content of Nitrogen, Potassium, Sodium and Silt is similar in both Kabale and Kabarole. However, the content of Phosphorus, Organic Carbon, Magnesium and Calcium were higher in Kabarole soils than Kabale soils. Bulk soils had lower physiochemical properties than rhizospheric soils in Kabale.
The phylum Proteobacteria was the most prevalent followed by Acidobacteria and Actinobacteria ( ). Comparing the two districts, Kabale soils had higher diversity of bacteria than Kabarole and there was significant difference (P≤0.5) among all phyla except phylum Proteobacteria. Kabale soil samples contained more Acidobacteria, Bacteroidetes, Veruccomicrobia, Gemmatimonadetes and Firmicutes while Kabarole soil samples contained more Actinobacteria and other phyla like Plactomycetes. With respect to the various soil treatments, there was no significant difference (P≤ 0.05) among the abundance of various phyla in Kabale. However, in Kabarole, there was significant differences (P≤ 0.05) in the abundance of phyla actinobacteria (14.85%) and verrucomicrobia (7.01%) in soils with artemisia only and the other treatments. Comparing bulk and rhizospheric soils, Proteobacteria were dorminant in both soils but acidobacteria and actinobacteria were more dorminate in bulk soils of Kabale and Kabarole respectively. Genus abundance The abundance of various bacterial genera (≥ 0.2%) in the bulk and rhizospheric soil of the most prevalent phylum (Proteobacteria, acidobacteria and actinobacteria) are shown in . Most genera in Kabale soils did not show any significant differences. However, most genera in Kabarole soils showed significant differences. In bulk soil, the genera that showed significant differences were po, br, no, str, ga and sol. In rhizospheric soils, Many genera (sps, ac, ram, ma, rhi, aci, ed, si, oc, pa and pse) showed significant differences with the bulk soil but genera sps, ram, pa and pse were observed to show significant in differences in both Kabale and Kabarole soils. Results of PCA of various soil components (genera and physiochemical properties) are shown in . Seven genera (sps, Lu, he, st, ma, spm and pse) had positively higher loading with the second component. Twenty nine soil components had higher loading with the first component and 18 (sand, Ca, Mg, P, N, OCa, pH, Sol, ga, Str, Vi, acis, pe, az, ram, ral, P and Ly) had positive factor loading while 11 (ac, ni, ps, aci, rhi, ed, si, oc, pa, pa_a and clay) had negative factor loading.
The abundance of various bacterial genera (≥ 0.2%) in the bulk and rhizospheric soil of the most prevalent phylum (Proteobacteria, acidobacteria and actinobacteria) are shown in . Most genera in Kabale soils did not show any significant differences. However, most genera in Kabarole soils showed significant differences. In bulk soil, the genera that showed significant differences were po, br, no, str, ga and sol. In rhizospheric soils, Many genera (sps, ac, ram, ma, rhi, aci, ed, si, oc, pa and pse) showed significant differences with the bulk soil but genera sps, ram, pa and pse were observed to show significant in differences in both Kabale and Kabarole soils. Results of PCA of various soil components (genera and physiochemical properties) are shown in . Seven genera (sps, Lu, he, st, ma, spm and pse) had positively higher loading with the second component. Twenty nine soil components had higher loading with the first component and 18 (sand, Ca, Mg, P, N, OCa, pH, Sol, ga, Str, Vi, acis, pe, az, ram, ral, P and Ly) had positive factor loading while 11 (ac, ni, ps, aci, rhi, ed, si, oc, pa, pa_a and clay) had negative factor loading.
The most abundant phylum of the A . annua rhizospheric bacterial community were Gemmatimonadetes, Acidobacteria and Proteobacteria. With the exception of acidobacteria, the results varied with what was reported by as they observed Chloroflexi, Cyanobacteria, and Planctomycetes as the most abundant in A . annua rhizosphere soil yet in this study, they were the least abundant. The variation may be stemming from the fact that plants were of different varieties and were also grown in different soil types at different altitudes. Proteobacteria have been reported to be dorminant in nutrient rich soil (copiotrophic) while acidobacteria are dominant in oligotrophic conditions and in soils with lower pH . Thus the results obtained tallied with what has been reported. Both bulk and rhizospheric soils had copiotrophic conditions as the ratio of proteobacteria to acidobacteria was high and also the nutrients were high ( ). Furthermore, bulk soils of Kabale had lower physiochemical properties and thus were expected to have more acidobacteria than rhizospheric soils and this was what was observed. On the other hand, more actinobacteria was observed in slightly alkaline Kabarole sandy loam soils that were rich in organic matter than in neutral Kabale soils, this observation tallied with reports of . According to PCA, the genera observed were categorized basing on pH and PCA1 was alkaline pH and PCA2 was neutral pH. This so because genus Pse (had the highest positive loading to PCA2) has been reported to grow at an optimum pH of 7.0–7.5 . Genera Rhi, aci and ed (had the highest negative loading to PCA1, greater than 0.9) have been reported to survive in acidic pH not alkaline pH i.e. genus ed (4.0–7.0) , aci (4.5–7.0) and rhi (5.5‒7.3) . On the other hand, also the genera with positive loading to PCA1 were not very high (had less than 0.9) as most of species of these genera have been observed to grow at moderately alkaline pH not at pH14.0 [ – ]. These results tallied with the findings of as they indicated that soil pH is vital in determining which genera present in agricultural soils.
In conclusion, the results show that the A . annua rhizosphere is a large reservoir of bacteria that may be capable of many roles. Most of the species observed in the rhizosphere are not among the most frequently mentioned non symbiotic PGPB ( Azospirillum sp., Azotobacter sp., Bacillus sp., Pseudomonas sp. etc) mentioned in various reports. However, many of the species belong to the phylum proteobacteria which constitutes most PGPB. Thus, use of selected bacteria especially proteobacteria may promote A . annua growth and increase its phytochemical contents.
|
Pharmacy students’ attitudes and intentions of pursuing postgraduate studies and training in pharmacogenomics and personalised medicine | fce8bcf5-b3e7-46b6-9594-d31448809d86 | 10035981 | Pharmacology[mh] | Personalised medicine (PM) and especially pharmacogenomics (PGx), are gaining momentum worldwide. PM constitutes a state-of-the-art approach able to revolutionise the way current medicine works . PGx is in the forefront of PM initiative, proposing an alternative treatment strategy in which individuals’ therapy scheme is tailored with respect of his/her genetic makeup. Currently, there are many clinical trials highlighting the clinical effectiveness of PGx in practice . Indeed, it is demonstrated that PGx-guided treatment will reduce the incidence rate of adverse drug reactions (ADRs), decrease mortality and morbidity level along with hospital admissions, while it can increase drug efficacy by adjusting either drug dosage or switching to another medication . Thanks to those advantages, overall disease and drug management can be improved to great extent. Apart from the clinical effectiveness of PGx, many studies conclude that PGx-guided treatment is a cost-effective strategy, and it will decrease healthcare expenditures. For all these points, PGx constitutes an important healthcare technology in drug management. Pharmacists are the most specialised healthcare professionals in terms of drug management, and hence they play a pivotal role in the advancement and implementation of PM interventions in the clinical setting . According to the latest American Society of Health System Pharmacists (ASHP) statements, pharmacists obtain the proper experience and education to interpret and apply PGx testing’s results . In addition, based on Genetics/Genomics Competency Center (G2C2), it seems that pharmacists share distinct responsibilities in terms of PGx implementation and they should understand the ethnicity effect in genetic variance and subsequently in drug response and highlight the existence of PGx guidelines for different medications . In other words, a pharmacist is authorised to adjust drug dosage, select a drug therapy, and monitor patients’ progress with regard to PGx results. Assisting other healthcare providers (i.e. physicians, nurses) in making clinical decision is also possible along with patient counselling related to their genetic results . Even if pharmacists’ contribution to PGx clinical implementation is vital, they have a minimal involvement in the field. As reported in many studies, a great percentage of pharmacists are not aware of PGx and its applications; some of them do not know how pharmacists can be involved in the field and what are their responsibilities, while others claim to have limited or inadequate training . Low participation rate can decelerate the adoption of PM applications. According to Koufaki et al., 2021, education is the key to face new challenges in such a constantly evolving environment and to achieve better results in terms of PGx adoption rate and pharmacists’ response . By initiating future generations of pharmacists to the principles and best practices of PGx at an early stage of their undergraduate education, it is possible to increase pharmacists’ self-confidence, disseminate new practices and thus enhance their participation in the sector . Admittedly, there are some studies that focus on understanding pharmacists’ and pharmacy students’ level of awareness, perceptions, attitudes and intentions about PGx applications in clinical practice. However, there is a significant gap in the literature regarding both pharmacists’ and pharmacy students’ intentions to continue their education in the field of PGx and PM . This gap may pose a barrier in the widespread adoption of clinical PGx. In this study, we aim to investigate the perceptions and opinions of undergraduate students of all years of studies from the Department of Pharmacy at the University of Patras in Greece. This department is the first pharmacy department in Greece that has a dedicated PGx module both in the undergraduate and graduate curriculum, towards some key aspects of PGx and PM. Using a questionnaire survey based on the Theory of Planned Behavior (TPB), we seek to evaluate the impact of selected factors and demographics on student’s intentions to pursue postgraduate training in PGx and PM. Research framework of the proposed model on students’ intention to pursue postgraduate training in PGx and PM and its predicting factors Literature review revealed that there was a lack of publications focussing on the intentions, and its predicting factors, of health science students' to pursue PGx and PM-related training after graduation. The Theory of Planned Behaviour (TPB) has been continuously applied in a great spectrum of healthcare issues, including the investigation of health science students’ intentions for further education . According to TPB, attitudes are considered as a critical factor affecting individuals’ intentions for future behaviour. Therefore, an extensive literature review was conducted regarding the attitudes, perceptions and intentions of health science students regarding PGx and PM. Prior research revealed that students’ attitudes towards PGx and PM application and adoption were related, among other factors, to their level of knowledge, their satisfaction from PM and PGx training and their self-confidence to apply PGx and PM in clinical practice . Literature review findings were discussed with pharmacy senior undergraduate and postgraduate (Master and PhD) students, as well as with experts in the field of PGx and PM. Experts were mainly coming from the Laboratory of PGx and Individualized Therapy, Department of Pharmacy, University of Patras, and had experience in research and survey development. Upon these discussions, the following factors were selected to probe their effect on the intention of pharmacy students to be further trained in the mentioned subjects/disciplines: 1) the self-confidence to apply PGx in clinical practice, 2) the satisfaction with the training regarding the application of PGx in clinical practice, 3) the prevailing attitudes-intentions for the application of PGx and PM and 4) the level of knowledge about PGx and PM. The proposed research framework of students’ intention and its predicting factors to continue training in PGx and PM after graduation is presented in Fig. . Mahmutovic et al. (2018) concluded that pharmacy students whose curriculum contained a course in PGx and PM hold more positive attitudes and were more willing to continue their training in this topic compared to students without previous relevant training . Moreover, Alzoubi et al. (2020) found that Jordanian medical students and physicians expressed rather positive attitudes towards PGx, and, at the same time, high interest in attending corresponding training sessions or workshops in the future . Additionally, according to the pharmacogenomics/genomics literacy framework for pharmacists (PGLP) developed by Rahma et al. (2021) attitudes played a positive influential role on pharmacists’ intentions to continuing medical education . Therefore, we assumed that attitudes influence positively students’ intentions to pursue postgraduate training in PGx and PM. Rahma et al. (2021) framework also pinpointed that both pharmacists’ knowledge and skills, often assessed by self-confidence, were correlated with their attitudes, and altogether affected their intentions to undergo a PGx training in the future . Grace et al. (2021) examined the impact of Personal Genomic educational testing (PGET) on student knowledge, comfort, and attitudes related to PGx in US students, and they noticed a simultaneous improvement in comfort with PGx clinical skills, PGx patient education, and attitudes towards PGx for all participants (receiving or not PGET) over the course of the study . These findings are in line with the results of other studies mentioning that students’ PGx knowledge is positively correlated with their self-confidence (readiness) and attitudes towards PGx implementation in clinical practice . Thus, we hypothesised that higher levels of self-confidence in PGx implementation in practice will be positively associated with favourable attitudes towards PGx and with greater intentions for PGx and PM postgraduate training. Accordingly, higher levels of PGx knowledge will be positively associated with higher self-confidence in PGx implementation, more favourable attitudes towards PGx and with greater intentions for PGx and PM postgraduate training. Both Mahmutovic et al. (2018) and Cheung et al. (2021) studies highlighted that the more students considered their study curriculum as well-designed for PGx understanding, the more they expressed positive attitudes towards PGx and were interested to get involved in this area of research or professional practice . Considering the above, we assumed that higher levels of training satisfaction will be positively associated with higher self-confidence in PGx implementation, more positive attitudes towards PGx and with greater intentions for PGx and PM postgraduate training. Survey development Survey’s tool development involved a multistage process. Initially, we combined relevant questionnaire surveys and the aforementioned discussions as the basis to develop questionnaire’s items associated with students’ intentions for postgraduate studies in PGx and PM and to determine the factors included in the proposed research model. At this stage, we decided to include another statement related to students’ evaluation of teaching methods used in PGx and PM education to facilitate the interpretation of the survey results. Next, the content validity of the selected items was examined by experts in the field to ascertain that all crucial items of each factor/construct were included. Draft questionnaire was then pretested through cognitive interviews with 25 pharmacy students of all years from the University of Patras to evaluate its clarity, content, length, and measurement scale . The final version of questionnaire was reviewed mainly about the wording and clarity of some items. It consisted of 38 questions divided into 7 different sections: (1) evaluation of teaching tools used for the PGx and PM training, (2) self-confidence to apply PGx in clinical practice, (3) satisfaction with the training regarding the application of PGx in clinical practice, (4) prevailing attitudes-intentions for the application of PGx and PM, (5), intentions for postgraduate training in PGx and PM, (6) level of knowledge about PGx and PM and (7) student demographics (such as gender, year of study, possession of another BSc degree, and if they or a family member are under chronic medication and have undergone a recent PGx or genetic test) (see Additional file ). All main sources considered, along with the measurement scale used to develop the survey, are demonstrated in Table . Moreover, an introductory cover page was attached in the beginning of the questionnaire to explain study’s purpose and objectives and to provide instructions on its completion. Additionally, definitions of PM and PGx were provided, and it was assured that participants anonymity and confidentiality would be preserved. The survey questionnaire required around 10 min to complete. Study sample Study sample consisted of 346 undergraduate students from the Department of Pharmacy of the University of Patras, Greece. Students of all academic years were asked to complete the questionnaire before or after their lectures during April (spring semester) 2022. Around two-thirds (64.5%) of the participants were women in line with the gender distribution of the Department’s student population (Table ). Fifth-year students (27.7%) were the most willing to participate in the survey, closely followed by the third year (24.3%), while fourth year (12.4%) were the least disposed to fill in the questionnaire. Notably, PGx principles and applications are taught in the course “Molecular Genetics and PGx” in the spring semester of the second study year. PGx-related material is also briefly covered in other modules within pharmacy curriculum. Moreover, 9% of the students stated they hold another BSc degree. Only 10% of the participants claimed that they or a relative of theirs have undergone a recent PGx or genetic test, while almost 67% stated that they or a family member was under chronic medication. Data analysis The research data were analysed by the SPSS and AMOS statistical programmes (both versions 28; IBM, NY, USA). Data analysis included frequencies and percentage of valid responses (valid %) and descriptive statistics (mean value, standard deviation (Std.Dev)). Mann–Whitney and Kruskal–Wallis tests were applied to identify the scale of differences among student groups based on their demographics (i.e. gender, study year, another BSc degree, PGx/genetic test, chronic medication). There is a debate whether parametric (t test and ANOVA) or nonparametric (Mann–Whitney and Kruskal–Wallis) tests should be used for group comparisons of item data selected with Likert scale. However, relevant studies concluded that both procedures, generally, have similar power to detect differences among groups . In our study, nonparametric tests were preferred, as the Kolmogorov–Smirnov test results indicated that the survey data were not normally distributed for almost all items. Structural equation modelling (SEM) analysis (IBM AMOS: generalised least squares (GLS) method for parameter estimation, regression coefficients estimation and statistical fits of the structural model evaluation) was utilised to discern the direct, indirect and total effects of both the examined factors and demographics on students’ intentions to pursue postgraduate training in PGx and PM . A direct effect refers to the direct impact of an independent (exogenous) variable on a dependent (endogenous) variable, an indirect effect is the effect of an independent variable on a dependent variable that goes through a mediator (the independent variable influences the mediator which then affects the dependent variable) and a total effect is the sum of the direct and indirect effects of an independent on a dependent variable . Prior to the structural model analysis, a measurement model analysis was conducted to examine the relationship between the latent variables and their measures, consisting of model fit statistics (CMIN/DF, CFI, TLI, NFI, IFI, RFI, RMSEA, SRMR) and the factor loadings of each factor . Moreover, Cronbach’s alpha and composite reliability (CR) test were used to measure the examined factors’ construct reliability. Average variance extracted (AVE) for each construct was also calculated to assess convergent validity. Discriminant validity was assessed by the heterotrait-monotrait ratio of correlations (HTMT) method. Additionally, the possible influence of common method variance (CMV) was measured by Harman’s single factor test . Literature review revealed that there was a lack of publications focussing on the intentions, and its predicting factors, of health science students' to pursue PGx and PM-related training after graduation. The Theory of Planned Behaviour (TPB) has been continuously applied in a great spectrum of healthcare issues, including the investigation of health science students’ intentions for further education . According to TPB, attitudes are considered as a critical factor affecting individuals’ intentions for future behaviour. Therefore, an extensive literature review was conducted regarding the attitudes, perceptions and intentions of health science students regarding PGx and PM. Prior research revealed that students’ attitudes towards PGx and PM application and adoption were related, among other factors, to their level of knowledge, their satisfaction from PM and PGx training and their self-confidence to apply PGx and PM in clinical practice . Literature review findings were discussed with pharmacy senior undergraduate and postgraduate (Master and PhD) students, as well as with experts in the field of PGx and PM. Experts were mainly coming from the Laboratory of PGx and Individualized Therapy, Department of Pharmacy, University of Patras, and had experience in research and survey development. Upon these discussions, the following factors were selected to probe their effect on the intention of pharmacy students to be further trained in the mentioned subjects/disciplines: 1) the self-confidence to apply PGx in clinical practice, 2) the satisfaction with the training regarding the application of PGx in clinical practice, 3) the prevailing attitudes-intentions for the application of PGx and PM and 4) the level of knowledge about PGx and PM. The proposed research framework of students’ intention and its predicting factors to continue training in PGx and PM after graduation is presented in Fig. . Mahmutovic et al. (2018) concluded that pharmacy students whose curriculum contained a course in PGx and PM hold more positive attitudes and were more willing to continue their training in this topic compared to students without previous relevant training . Moreover, Alzoubi et al. (2020) found that Jordanian medical students and physicians expressed rather positive attitudes towards PGx, and, at the same time, high interest in attending corresponding training sessions or workshops in the future . Additionally, according to the pharmacogenomics/genomics literacy framework for pharmacists (PGLP) developed by Rahma et al. (2021) attitudes played a positive influential role on pharmacists’ intentions to continuing medical education . Therefore, we assumed that attitudes influence positively students’ intentions to pursue postgraduate training in PGx and PM. Rahma et al. (2021) framework also pinpointed that both pharmacists’ knowledge and skills, often assessed by self-confidence, were correlated with their attitudes, and altogether affected their intentions to undergo a PGx training in the future . Grace et al. (2021) examined the impact of Personal Genomic educational testing (PGET) on student knowledge, comfort, and attitudes related to PGx in US students, and they noticed a simultaneous improvement in comfort with PGx clinical skills, PGx patient education, and attitudes towards PGx for all participants (receiving or not PGET) over the course of the study . These findings are in line with the results of other studies mentioning that students’ PGx knowledge is positively correlated with their self-confidence (readiness) and attitudes towards PGx implementation in clinical practice . Thus, we hypothesised that higher levels of self-confidence in PGx implementation in practice will be positively associated with favourable attitudes towards PGx and with greater intentions for PGx and PM postgraduate training. Accordingly, higher levels of PGx knowledge will be positively associated with higher self-confidence in PGx implementation, more favourable attitudes towards PGx and with greater intentions for PGx and PM postgraduate training. Both Mahmutovic et al. (2018) and Cheung et al. (2021) studies highlighted that the more students considered their study curriculum as well-designed for PGx understanding, the more they expressed positive attitudes towards PGx and were interested to get involved in this area of research or professional practice . Considering the above, we assumed that higher levels of training satisfaction will be positively associated with higher self-confidence in PGx implementation, more positive attitudes towards PGx and with greater intentions for PGx and PM postgraduate training. Survey’s tool development involved a multistage process. Initially, we combined relevant questionnaire surveys and the aforementioned discussions as the basis to develop questionnaire’s items associated with students’ intentions for postgraduate studies in PGx and PM and to determine the factors included in the proposed research model. At this stage, we decided to include another statement related to students’ evaluation of teaching methods used in PGx and PM education to facilitate the interpretation of the survey results. Next, the content validity of the selected items was examined by experts in the field to ascertain that all crucial items of each factor/construct were included. Draft questionnaire was then pretested through cognitive interviews with 25 pharmacy students of all years from the University of Patras to evaluate its clarity, content, length, and measurement scale . The final version of questionnaire was reviewed mainly about the wording and clarity of some items. It consisted of 38 questions divided into 7 different sections: (1) evaluation of teaching tools used for the PGx and PM training, (2) self-confidence to apply PGx in clinical practice, (3) satisfaction with the training regarding the application of PGx in clinical practice, (4) prevailing attitudes-intentions for the application of PGx and PM, (5), intentions for postgraduate training in PGx and PM, (6) level of knowledge about PGx and PM and (7) student demographics (such as gender, year of study, possession of another BSc degree, and if they or a family member are under chronic medication and have undergone a recent PGx or genetic test) (see Additional file ). All main sources considered, along with the measurement scale used to develop the survey, are demonstrated in Table . Moreover, an introductory cover page was attached in the beginning of the questionnaire to explain study’s purpose and objectives and to provide instructions on its completion. Additionally, definitions of PM and PGx were provided, and it was assured that participants anonymity and confidentiality would be preserved. The survey questionnaire required around 10 min to complete. Study sample consisted of 346 undergraduate students from the Department of Pharmacy of the University of Patras, Greece. Students of all academic years were asked to complete the questionnaire before or after their lectures during April (spring semester) 2022. Around two-thirds (64.5%) of the participants were women in line with the gender distribution of the Department’s student population (Table ). Fifth-year students (27.7%) were the most willing to participate in the survey, closely followed by the third year (24.3%), while fourth year (12.4%) were the least disposed to fill in the questionnaire. Notably, PGx principles and applications are taught in the course “Molecular Genetics and PGx” in the spring semester of the second study year. PGx-related material is also briefly covered in other modules within pharmacy curriculum. Moreover, 9% of the students stated they hold another BSc degree. Only 10% of the participants claimed that they or a relative of theirs have undergone a recent PGx or genetic test, while almost 67% stated that they or a family member was under chronic medication. The research data were analysed by the SPSS and AMOS statistical programmes (both versions 28; IBM, NY, USA). Data analysis included frequencies and percentage of valid responses (valid %) and descriptive statistics (mean value, standard deviation (Std.Dev)). Mann–Whitney and Kruskal–Wallis tests were applied to identify the scale of differences among student groups based on their demographics (i.e. gender, study year, another BSc degree, PGx/genetic test, chronic medication). There is a debate whether parametric (t test and ANOVA) or nonparametric (Mann–Whitney and Kruskal–Wallis) tests should be used for group comparisons of item data selected with Likert scale. However, relevant studies concluded that both procedures, generally, have similar power to detect differences among groups . In our study, nonparametric tests were preferred, as the Kolmogorov–Smirnov test results indicated that the survey data were not normally distributed for almost all items. Structural equation modelling (SEM) analysis (IBM AMOS: generalised least squares (GLS) method for parameter estimation, regression coefficients estimation and statistical fits of the structural model evaluation) was utilised to discern the direct, indirect and total effects of both the examined factors and demographics on students’ intentions to pursue postgraduate training in PGx and PM . A direct effect refers to the direct impact of an independent (exogenous) variable on a dependent (endogenous) variable, an indirect effect is the effect of an independent variable on a dependent variable that goes through a mediator (the independent variable influences the mediator which then affects the dependent variable) and a total effect is the sum of the direct and indirect effects of an independent on a dependent variable . Prior to the structural model analysis, a measurement model analysis was conducted to examine the relationship between the latent variables and their measures, consisting of model fit statistics (CMIN/DF, CFI, TLI, NFI, IFI, RFI, RMSEA, SRMR) and the factor loadings of each factor . Moreover, Cronbach’s alpha and composite reliability (CR) test were used to measure the examined factors’ construct reliability. Average variance extracted (AVE) for each construct was also calculated to assess convergent validity. Discriminant validity was assessed by the heterotrait-monotrait ratio of correlations (HTMT) method. Additionally, the possible influence of common method variance (CMV) was measured by Harman’s single factor test . Students’ perceptions on the examined factors affecting their intentions for postgraduate training in PGx and PM Students assessed the teaching tools used in PGx modules to be moderately to rather useful (Fig. and Table ). Lectures and supplementary material in e-class were perceived as the most useful tools (mean around 5.2), followed by laboratory training (mean 4.7). Books and formative assignments were considered as moderately helpful (mean around 4.2). Study year was the only demographic variable to influence students’ evaluation of lectures, laboratory training, e-class material and optional assignments (Additional file : Tables S1–S5). Indeed, second-year students were indicated to be the most pleased with available teaching tools in contrast to fifth-year students. Second-year students were very satisfied with lectures, laboratory training and e-class material (means from 5.3 to 5.9), while third-year students found lectures and e-class material quite valuable (means around 5.6). Students’ satisfaction related to training received in PGx clinical implementation is in line with their evaluation of teaching tools. They appear to be moderately to rather satisfied with theoretical and laboratory training along with department’s curriculum, as 45% to 50% of participants agreed with the relevant statements and means ranged from 4.3 to 4.6 (Fig. and Table ). Again, study year was the only demographic variable affecting students’ views (Additional file : Tables S1–S5), with second-year students being the most satisfied (mean 5.3) and fifth-year the least (mean 4.0). Third-year students were rather satisfied with their theoretical training (mean 4.8), but moderately with their laboratory training (mean 3.7), which is attributed mainly to the special circumstances that laboratory exercises were conducted during the last two years because of COVID-19 restrictions. Students claimed to be moderately confident in implementing PGx in clinical practice, as the means of all relevant statements ranged from 3.5 to 3.9 (Fig. and Table ). Study year only affected students’ self-confidence, mainly due to significant divergences between first and second-year students (Additional file : Tables S1–S5). Students shared very positive attitudes towards PGx and its implementation in clinical practice. Around 80% agreed that PGx should be an important part of pharmacy students’ curriculum and they would recommend PGx testing to patients or a family member (Fig. and Table ). Moreover, almost 60% agreed that they intended to keep up with future updates in the field of PGx and incorporate PGx testing in patient care. Female students were more positive than their male classmates concerning the significance of PGx in their curriculum, while female students intended to be updated about future advances in the field and to integrate PGx in patient care. Students (or a family member of theirs) receiving chronic medication were more inclined to recommend PGx testing to patients or a family member, whereas those holding another BSc degree were less willing to undergo a PGx testing in the future. First-year, closely followed by second-year, students held more positive attitudes and were more disposed to adopt PGx in clinical practice and keep up with future updates. Individuals’ level of knowledge on PGx and PM was measured by an index, (0 to 10), concerning the total number of correct answers given at the 10 relevant statements (Fig. ). Students’ level of knowledge in PGx and PM was found moderately satisfactory (Figs. , and ). Approximately 75% of them gave 4 to 7 correct answers, whereas only 4% of the respondents got a score of 9 out of 10 (mean = 5.7, median = mode = 6). Students (or a family member of theirs) receiving chronic medication scored on average 0.5 units higher than others. Moreover, students of the last three study years presented an almost equal level of knowledge in PGx and PM, but much higher than those of the first and second year (Additional file : Tables S1–S5). Students claimed to be moderately willing to pursue postgraduate studies related to PGx and PM, Almost 45% of them agreed with the pertinent clause and one-quarter of them were neutral (mean 4.2). However, they were quite disposed to attend relevant certified training or information programmes, since about two-thirds of them agreed with the relevant statement and only 17% disagreed (mean 5.03) (Fig. and Table ). Female students were more inclined to improve their knowledge and skills on PGx and PM, while those possessing another BSc degree stated a lower interest. Students’ interest for postgraduate training in PGx and PM declined as they ascended the study year. Thus, while students of the first 3 years appeared quite willing for postgraduate training, fourth- and, especially, fifth-year students were, to some extent, reluctant, particularly to attend a relevant MSc course (Additional file : Tables S1–S5). SEM analysis of key determinants on student intentions for postgraduate training in PGx and PM The measurement model of the four factors (Training Satisfaction, Self-Confidence, Attitudes and Intentions) was tested by Confirmatory Factor Analysis (CFA). The level of knowledge on PGx and PM was measured by an index, of a scale from 0 to 10; therefore, it was not considered in the measurement model analysis. All values of the model-fit measures calculated to assess the model’s overall goodness of fit were within the corresponding acceptance levels. Specifically, the values computed by AMOS were: CMIN/DF = 1.512 < 3–5, CFI = 0.986 ≥ 0.90, TLI = 0.981 ≥ 0.90, NFI = 0.960 ≥ 0.90, IFI = 0.986 ≥ 0.90, RFI = 0.945 ≥ 0.90, RMSEA = 0.39 (0.023 – 0.052) < 0.08 and SRMR = 0.041 < 0.08 (Collier 2020; Kline, 2016). According to Table , the factor loadings of each item on the respective constructs are well above the suggested threshold of 0.5; therefore, all items were eligible for the next stages of analysis. However, ATT6 item ( I would recommend pharmacogenomic testing to a family member ) was removed from the analysis because of its high correlation with ATT5 item (0.848). Cronbach’s alpha and composite reliability (CR) verified the construct reliability of each model construct (Table ). All Cronbach’s alpha values were quite higher than the acceptance level of 0.70 , and CRs, ranging from 0.81 to 0.86, were well above the threshold value of 0.70 . The average variance extracted (AVE) for all constructs was found higher than the benchmark of 0.50 (Table ), indicating the convergent validity of the factors included in model . The discriminant validity was confirmed by the heterotrait-monotrait ratio of correlations (HTMT) method, as all ratios ranged between 0.202 and 0.797 (Table ), below the suggested critical value of 0.85 . Finally, in our measurement model the Harman’s single factor test showed that the total variance for a single factor was 36.77%, considerably lower than the recommended threshold of 50%, and hence, we assumed that the common method bias did not constitute a significant threat to the validity of our research findings . Mann–Whitney and Kruskal–Wallis tests revealed significant differences on some of students’ answers according to gender, study year, possession of another BSc degree and receiving chronic medication. Thus, these demographics were included in SEM analysis as control variables. A binary variable was used for “study year” (1st–3rd: 0; 4th–5th: 1) to simplify the SEM analysis. Third-year students were grouped together with their first- and second-year classmates, as they expressed relatively similar intentions for postgraduate training in PGx and PM. This may be explained by the fact that fourth and fifth-year students are approaching to the completion of their studies, and consequently they are more settled regarding their intentions for postgraduate training including PGx and PM. Figure outlines the statistically significant effects of key determinants on student intentions. More details are presented in Figure S1 and Tables and . The values of the global fit statistics (CMIN/DF < 2–5, CFI ≥ 0.90, TLI ≥ 0.90, NFI ≥ 0.90, IFI ≥ 0.90, RFI ≥ 0.90, RMSEA < 0.08, SRMR < 0.08) indicated a good model fit . Moreover, the coefficient of determination ( R 2 ) was 0.70 signifying, as well, a good model fit. Attitudes towards PGx implementation were the only factor which directly influenced students’ intentions to pursue postgraduate training in PGx and PM. The high value of the pertinent standardised regression weight, 0.76, indicated a very strong positive impact of attitudes on intentions for postgraduate training. Self-confidence and training satisfaction affected students’ intentions only indirectly, as they exhibited a moderate positive influence on attitudes (total effects: 0.35 and 0.31, respectively). Furthermore, training satisfaction exerted a relatively low (0.25) positive effect on self-confidence. However, self-confidence was found to have a rather low negative (−0.11), but still significant, impact solely on students’ intentions to attend certified training or information programmes related to PGx and PM (INT2). Therefore, the positive total, and indirect, effects of both self-confidence and training satisfaction on intentions for postgraduate training were relatively low: 0.26 and 0.24, respectively. The level of knowledge was the only examined factor of the proposed research model that had neither direct nor indirect impact on students’ intentions for postgraduate training, in general. However, it appeared to have a quite low positive impact (0.11) on students’ intentions to attend certified training or information programmes related to PGx and PM (INT2). Although no significant relationship was established between the level of knowledge and the training satisfaction construct, a low negative relationship was found between the level of knowledge and the satisfaction with both the laboratory training (STF2) and the departments’ curriculum (STF3) (−0.19 and −0.11, respectively). Concerning control variables, study year, primarily, and gender, to a less degree, influenced students’ intentions for postgraduate training both directly and indirectly. Specifically, in line with the Kruskal–Wallis test results, students of the last two study years appeared to be considerably less willing to pursue postgraduate training in PGx and PM, less zealous to implement PGx and PM in clinical practice and less satisfied with their training than their classmates of the first 3 years. Female students were more disposed to pursue postgraduate training and hold more favourable attitudes towards PGx and PM than their male classmates. Students assessed the teaching tools used in PGx modules to be moderately to rather useful (Fig. and Table ). Lectures and supplementary material in e-class were perceived as the most useful tools (mean around 5.2), followed by laboratory training (mean 4.7). Books and formative assignments were considered as moderately helpful (mean around 4.2). Study year was the only demographic variable to influence students’ evaluation of lectures, laboratory training, e-class material and optional assignments (Additional file : Tables S1–S5). Indeed, second-year students were indicated to be the most pleased with available teaching tools in contrast to fifth-year students. Second-year students were very satisfied with lectures, laboratory training and e-class material (means from 5.3 to 5.9), while third-year students found lectures and e-class material quite valuable (means around 5.6). Students’ satisfaction related to training received in PGx clinical implementation is in line with their evaluation of teaching tools. They appear to be moderately to rather satisfied with theoretical and laboratory training along with department’s curriculum, as 45% to 50% of participants agreed with the relevant statements and means ranged from 4.3 to 4.6 (Fig. and Table ). Again, study year was the only demographic variable affecting students’ views (Additional file : Tables S1–S5), with second-year students being the most satisfied (mean 5.3) and fifth-year the least (mean 4.0). Third-year students were rather satisfied with their theoretical training (mean 4.8), but moderately with their laboratory training (mean 3.7), which is attributed mainly to the special circumstances that laboratory exercises were conducted during the last two years because of COVID-19 restrictions. Students claimed to be moderately confident in implementing PGx in clinical practice, as the means of all relevant statements ranged from 3.5 to 3.9 (Fig. and Table ). Study year only affected students’ self-confidence, mainly due to significant divergences between first and second-year students (Additional file : Tables S1–S5). Students shared very positive attitudes towards PGx and its implementation in clinical practice. Around 80% agreed that PGx should be an important part of pharmacy students’ curriculum and they would recommend PGx testing to patients or a family member (Fig. and Table ). Moreover, almost 60% agreed that they intended to keep up with future updates in the field of PGx and incorporate PGx testing in patient care. Female students were more positive than their male classmates concerning the significance of PGx in their curriculum, while female students intended to be updated about future advances in the field and to integrate PGx in patient care. Students (or a family member of theirs) receiving chronic medication were more inclined to recommend PGx testing to patients or a family member, whereas those holding another BSc degree were less willing to undergo a PGx testing in the future. First-year, closely followed by second-year, students held more positive attitudes and were more disposed to adopt PGx in clinical practice and keep up with future updates. Individuals’ level of knowledge on PGx and PM was measured by an index, (0 to 10), concerning the total number of correct answers given at the 10 relevant statements (Fig. ). Students’ level of knowledge in PGx and PM was found moderately satisfactory (Figs. , and ). Approximately 75% of them gave 4 to 7 correct answers, whereas only 4% of the respondents got a score of 9 out of 10 (mean = 5.7, median = mode = 6). Students (or a family member of theirs) receiving chronic medication scored on average 0.5 units higher than others. Moreover, students of the last three study years presented an almost equal level of knowledge in PGx and PM, but much higher than those of the first and second year (Additional file : Tables S1–S5). Students claimed to be moderately willing to pursue postgraduate studies related to PGx and PM, Almost 45% of them agreed with the pertinent clause and one-quarter of them were neutral (mean 4.2). However, they were quite disposed to attend relevant certified training or information programmes, since about two-thirds of them agreed with the relevant statement and only 17% disagreed (mean 5.03) (Fig. and Table ). Female students were more inclined to improve their knowledge and skills on PGx and PM, while those possessing another BSc degree stated a lower interest. Students’ interest for postgraduate training in PGx and PM declined as they ascended the study year. Thus, while students of the first 3 years appeared quite willing for postgraduate training, fourth- and, especially, fifth-year students were, to some extent, reluctant, particularly to attend a relevant MSc course (Additional file : Tables S1–S5). The measurement model of the four factors (Training Satisfaction, Self-Confidence, Attitudes and Intentions) was tested by Confirmatory Factor Analysis (CFA). The level of knowledge on PGx and PM was measured by an index, of a scale from 0 to 10; therefore, it was not considered in the measurement model analysis. All values of the model-fit measures calculated to assess the model’s overall goodness of fit were within the corresponding acceptance levels. Specifically, the values computed by AMOS were: CMIN/DF = 1.512 < 3–5, CFI = 0.986 ≥ 0.90, TLI = 0.981 ≥ 0.90, NFI = 0.960 ≥ 0.90, IFI = 0.986 ≥ 0.90, RFI = 0.945 ≥ 0.90, RMSEA = 0.39 (0.023 – 0.052) < 0.08 and SRMR = 0.041 < 0.08 (Collier 2020; Kline, 2016). According to Table , the factor loadings of each item on the respective constructs are well above the suggested threshold of 0.5; therefore, all items were eligible for the next stages of analysis. However, ATT6 item ( I would recommend pharmacogenomic testing to a family member ) was removed from the analysis because of its high correlation with ATT5 item (0.848). Cronbach’s alpha and composite reliability (CR) verified the construct reliability of each model construct (Table ). All Cronbach’s alpha values were quite higher than the acceptance level of 0.70 , and CRs, ranging from 0.81 to 0.86, were well above the threshold value of 0.70 . The average variance extracted (AVE) for all constructs was found higher than the benchmark of 0.50 (Table ), indicating the convergent validity of the factors included in model . The discriminant validity was confirmed by the heterotrait-monotrait ratio of correlations (HTMT) method, as all ratios ranged between 0.202 and 0.797 (Table ), below the suggested critical value of 0.85 . Finally, in our measurement model the Harman’s single factor test showed that the total variance for a single factor was 36.77%, considerably lower than the recommended threshold of 50%, and hence, we assumed that the common method bias did not constitute a significant threat to the validity of our research findings . Mann–Whitney and Kruskal–Wallis tests revealed significant differences on some of students’ answers according to gender, study year, possession of another BSc degree and receiving chronic medication. Thus, these demographics were included in SEM analysis as control variables. A binary variable was used for “study year” (1st–3rd: 0; 4th–5th: 1) to simplify the SEM analysis. Third-year students were grouped together with their first- and second-year classmates, as they expressed relatively similar intentions for postgraduate training in PGx and PM. This may be explained by the fact that fourth and fifth-year students are approaching to the completion of their studies, and consequently they are more settled regarding their intentions for postgraduate training including PGx and PM. Figure outlines the statistically significant effects of key determinants on student intentions. More details are presented in Figure S1 and Tables and . The values of the global fit statistics (CMIN/DF < 2–5, CFI ≥ 0.90, TLI ≥ 0.90, NFI ≥ 0.90, IFI ≥ 0.90, RFI ≥ 0.90, RMSEA < 0.08, SRMR < 0.08) indicated a good model fit . Moreover, the coefficient of determination ( R 2 ) was 0.70 signifying, as well, a good model fit. Attitudes towards PGx implementation were the only factor which directly influenced students’ intentions to pursue postgraduate training in PGx and PM. The high value of the pertinent standardised regression weight, 0.76, indicated a very strong positive impact of attitudes on intentions for postgraduate training. Self-confidence and training satisfaction affected students’ intentions only indirectly, as they exhibited a moderate positive influence on attitudes (total effects: 0.35 and 0.31, respectively). Furthermore, training satisfaction exerted a relatively low (0.25) positive effect on self-confidence. However, self-confidence was found to have a rather low negative (−0.11), but still significant, impact solely on students’ intentions to attend certified training or information programmes related to PGx and PM (INT2). Therefore, the positive total, and indirect, effects of both self-confidence and training satisfaction on intentions for postgraduate training were relatively low: 0.26 and 0.24, respectively. The level of knowledge was the only examined factor of the proposed research model that had neither direct nor indirect impact on students’ intentions for postgraduate training, in general. However, it appeared to have a quite low positive impact (0.11) on students’ intentions to attend certified training or information programmes related to PGx and PM (INT2). Although no significant relationship was established between the level of knowledge and the training satisfaction construct, a low negative relationship was found between the level of knowledge and the satisfaction with both the laboratory training (STF2) and the departments’ curriculum (STF3) (−0.19 and −0.11, respectively). Concerning control variables, study year, primarily, and gender, to a less degree, influenced students’ intentions for postgraduate training both directly and indirectly. Specifically, in line with the Kruskal–Wallis test results, students of the last two study years appeared to be considerably less willing to pursue postgraduate training in PGx and PM, less zealous to implement PGx and PM in clinical practice and less satisfied with their training than their classmates of the first 3 years. Female students were more disposed to pursue postgraduate training and hold more favourable attitudes towards PGx and PM than their male classmates. To the best of our knowledge, this is the first study to examine the effects of selected factors (e.g. attitudes, knowledge, self-confidence, training satisfaction) on pharmacy undergraduate students’ intentions to pursue PGx-related studies in the future using TPB. Our study showed that the upcoming generations of professional pharmacists were aspired by the advantages of PGx and shared a positive attitude towards its integration in clinical practice. In parallel, it was observed that they were willing to pursue postgraduate studies in this field, despite the fact that they expressed a moderate satisfaction from their department’s curriculum, and they did not feel well-prepared to clinically implement PGx. Furthermore, SEM analysis’ results indicated a good fit of the proposed model, pinpointing that some of the examined factors such as attitudes, training satisfaction and self-confidence exerted a significant impact on students’ intentions to pursue further education in PGx and PM field. Results from SEM analysis also demonstrated that attitudes were the most influential factor, directly affecting respondents’ intention to continue their studies in the sector, a fact that is congruent with what Koufaki et al., 2022 showed in a study among Malaysian and Greek healthcare students . Students mentioned that lectures, supplementary material and laboratory training were the most useful educational tools. This tendency is observed across all academic years. Second-year students claimed that lectures were the most valuable education tool. PGx- and PM-dedicated module takes place at spring semester of second year as in 21% of pharmacy schools in the USA and therefore corresponding students were affected by it. As far as the self-confidence concerns, most of the respondents did not consider themselves as adequately prepared to apply PGx in clinical practice except of first-year students. It is likely that first-year students overestimated their competencies and readiness in PGx implementation. Second-year students were the least self-confident probably because they realised their limited knowledge on PGx and PM subjects, while the students of the last three study years were, slightly more self-confident. In other studies individuals also characterised their self-confidence in clinically applying PGx as weak . Moreover, according to de Denus et al., 2013, only a 7.7% of Canadian pharmacists claimed to feel comfortable to integrate PGx in their daily practice while only a quarter of USA pharmacists stated to be able to interpret PGx results to their patients’ treatment . Moreover, Bank et al., 2018 found that an almost 28% of pharmacy students thought that they were adequately trained and prepared to interpret PGx testing data and change medication or dosage . Moreover, participants characterised their curriculum’s satisfaction as intermediate, and they weren’t very satisfied. Final-year students shared the lowest level of satisfaction regarding their studies which may be attributed, inter alia, to their internships at community and hospital pharmacies. These results are consistent with other studies in which 60.3% of individuals believed that their university education was insufficient and only a third of them supposed that lectures were comprehensive and clear . This descending trend highlighted the insufficiency of PGx-related material in latent modules and the students’ demand for a more tailored educational scheme that focuses more on learner’s skills, needs and wants. Admittedly, new pedagogies have incorporated evidence-based learning as an alternative teaching approach with the objective to better prepare students for their future professional career . Furthermore, participants from all academic years expressed a very positive attitude towards PGx implementation in clinical practice, a strong indication of being aware and convinced of PGx benefits. Students’ positive attitude was also highlighted by their willingness to undergo a PGx testing or to recommend it to patients or friends. Our results are in line with another study occurred in Greece by Siamoglou et al., 2021 in which the majority of health students had a positive opinion about the topic and they mentioned that they would perform a genetic test in the future. Other studies agreed with our findings as well and pinpointed that positive attitude was associated with a person’s intention to adopt a relevant clinical intervention . Positive attitude is also translated into positive intention, as it was hypothesised in our proposed model. Students’ majority were prone to pursue postgraduate studies in the future or continue getting updated about the newest advances in the field. This may be also attributed to the fact that in Greece, all faculties–departments of medicine and pharmacy include courses focussing on PGx and PM in their undergraduate curricula, according to the information provided by their webpages. Additionally, there are three postgraduate programmes with a concentration on this field taught in the biggest Universities of Greece and the relevant academic research laboratories offer doctorate programmes, as well. These aspects highlight the high interest of Greek academics in promoting PGx and PM and raising students’ awareness about the importance of their clinical implementation. Consistent with previous studies, this observation is important and encouraging. In particular, Mahmutovic et al., 2018 mentioned that 60% of the students asked would like to continue their studies in the field, Schwartz et al. 2017 demonstrated that 97% of the hospital pharmacists expressed interest in continuing their training, while in Indonesia, 94.2% of pharmacists were interested in future education along with Jordan where a 75% of pharmacists expressed their will to participate in postgraduate training activities or seminars . Based on Pop et al., 2022, and Filiptsova et al., 2015, new generation of pharmacists were more progressive and willing to implement innovative technologies and interventions in their clinical practice with the objective to improve patient experience and provide them with better quality of services . To achieve it, it is fundamental to receive official training and accreditation. It is true that in many studies, respondents, especially, professionals pinpointed that pharmacists should be knowledgeable in terms of PGx interventions since one of their main duties is patients’ education and counselling on their medication . Despite the fact that students were willing to pursue a PGx training in the future, their level of knowledge was moderate and had no significant impact on their intentions to read up with PGx and PM in the future. The PGx and PM knowledge index calculated in this study accrued from the total number of correct answers given to 10 statements already used and validated in other studies . Moreover, the knowledge index, employed, managed to adequately classify participants in terms of their level of PGx and PM knowledge, as the relevant diagram of the overall knowledge assessment yielded, approximately, a normal distribution curve. Surprisingly, most students failed to answer three knowledge questions related to PGx clinical implementation. Indeed, only 20% of respondents knew that PGx clinical guidelines weren’t available for most of the medications, a fact that highlights students’ need for less theoretical sessions and better information about PGx testing availability in a commercial level. Students of the last three academic years demonstrated almost equal level of knowledge in the discipline, but much higher than those of the first and second year, also underpinning the necessity for more PGx-relevant lectures. Pop et al., 2022 made a relevant observation about information gap among Romanian pharmacists . They pointed out that even if Romanian pharmacists had a relatively high level of theoretical PGx knowledge, most of them weren’t aware that PGx testing was commercially available in their country for many medications . Low level of knowledge and lack of information are significant barriers to PGx clinical implementation . The fact that knowledge did not affect students’ intentions for postgraduate training may be attributed, inter alia, to the low pace of PGx and PM implementation in clinical practice. It was found that students’ intentions for postgraduate training in PGx and PM faded approaching to the completion of their studies. In fact, fifth-year students were to some extent, reluctant to attend a relevant MSc course. By the completion of the questionnaire, almost all graduates have attended internships at community and hospital pharmacies and quite probably realised the limited PGx and PM adoption in the clinical setting. Therefore, they likely aspire to postgraduate training in other more promising fields that would facilitate their initial career steps. The low level of students’ self-confidence coupled with their moderate training satisfaction rate, as well, raise several questions about the efficacy of current pharmacy curriculum. Evidently, school’s curriculum needs a change, especially considering that our study results revealed that both student self-confidence and training satisfaction affected positively their attitudes towards PGx implementation in practice, and consequently, their intentions to read up with PGx and PM. By incorporating case studies, publications, or even journal club sessions dedicated to PM and its clinical interventions to the existing courses, students would better understand the correlation of all modules and gain a deeper knowledge of PM. This could also be improved by the addition of an elective course about the application of PM and PGx in a pharmacist’s clinical practice, to wit, a change from knowledge-based PGx training to a more practical one. This kind of addition would bear fruitful results because pharmacists are more interested in receiving formal training related to drug therapy recommendations, results interpretation, patient counselling and education, future advances as many studies claimed . Our study has a few limitations. The study sample consisted only of undergraduate pharmacy students from the University of Patras, and it didn’t represent all pharmacy schools in Greece. Survey’s questionnaire couldn’t include all factors that potentially affect students’ intentions to pursue postgraduate studies in PGx. Finally, study results may be biased by the COVID-19 pandemic restrictions to teaching processes. Pharmacists play a determinant role in clinical PGx but their presence is still limited. Future generations of pharmacists opted to apply new technologies such as PGx in their practice but they considered themselves as not adequately prepared and seek for postgraduate opportunities to broaden their horizons. This phenomenon is global . The low satisfaction level from pharmacy department’s curriculum along with low self-confidence indicates the need for curriculum modifications to tailor students’ needs. Hence, it is urgent for pharmacy schools to update and upgrade their curricula accordingly and introduce more PGx- and PM-related sessions. Study’s questionnaire proved to be a reliable and promising instrument to interpret pharmacy students’ intentions to pursue postgraduate studies in PGx and PM. Given the close collaboration of physicians and pharmacists in the clinical implementation of PGx, this survey instrument would also be suitable for understanding the underlying factors of medical students’ aspirations to further develop their knowledge and skills in PGx and PM. Additional file 1: Table S1 . Gender’s impact on students’ answers; Table S2 Recent PGx/genetic test impact on students’ answers; Table S3 Chronic medication’s impact on students’ answers; Table S4 Another BSc degree’s impact on students’ answers; Table S5 Study year’s impact on students’ answers; Fig. S1 SEM diagram of the factors influencing students’ intention to pursue postgraduate training in PGx and PM. Additional file 2 : Survey Questionnaire. |
Investigation of the effect of music listened to by patients with moderate dental anxiety during restoration of posterior occlusal dental caries | 5e70d880-7874-419b-b4af-c4f98bf12b1f | 10036243 | Dental[mh] | Dental anxiety is excessive fear created by the thought of visiting the dentist for dental treatments . Factors such as the patient’s age, gender, educational level, profession, previous experiences, socio-economic status, and geography may each play a role in its etiology . Although there have been many innovations and developments in the field of dentistry in recent years, some patients do not know how to cope with dental anxiety, and therefore, they delay dental treatments. For this reason, it has been determined that anxious dental patients are more likely to have dental caries, gingival diseases, and bad oral hygiene . If dental anxiety is controlled, the oral and dental health of patients will improve significantly. As a result of anxiety, mydriasis (dilation in pupils), increased heart rate (tachycardia), high blood pressure (hypertension), high blood sugar (hyperglycemia), increased body temperature, high cholesterol, increased cortisol secretion, and decreased oxygen saturation values (desaturation) may be seen [ – ]. To reduce dental anxiety and therefore these symptoms, premedication, sedation, distraction, hypnosis, and music therapy can be used . Music therapy is a non-invasive, economical, effective intervention method to reduce the anxiety level of patients . It may exert its effect on subjective anxiety scales and objective physiologic parameters, such as heart rate, blood pressure, oxygen saturation, body temperature, and cortisol hormones [ – ]. In dentistry, music can reduce dental anxiety and pain, which are linked closely . Music listening might also activate imagery, offering an escape from the stressful condition. Anxiety often increases before the dental treatment starts. Visit purpose and waiting room experience, specifically the amount of time waiting for the treatment and the waiting room environment, are the factors which may cause dental anxiety prior to restorative treatment . Thus, on a psychosocial level, music provides the patient an aesthetic experience that can offer comfort and peace while awaiting and during dental treatment . The Modified Dental Anxiety Scale (MDAS) is a questionnaire with five questions to classify dentally anxious participants. Each question has five different response options ranging from a score of 1 (not anxious) to 5 (extremely anxious). The MDAS sum score has a minimum of 5 and a maximum of 25 . For scores 5–9, participants are classified as not dentally anxious or slightly anxious, and thus, they do not need any intervention . The people whose scores are higher than 19 are deemed highly anxious, and they may be treated with more complex interventions [ – ]. Participants with scores between 10 and 18 are defined as having moderate dental anxiety. It is especially expressed that individuals with moderate dental anxiety may benefit principally from noninvasive methods, such as listening to music . Restorative dentistry has a strong correlation with both pre-treatment dental anxiety and fear of pain during treatment. The patient’s subjective reactions to restorative dentistry procedures such as local anesthesia, cavity preparation, the pressure applied to the tooth, the vibrations and noise recorded, the heat and smell generated at the interface between the tooth and the bur, and the time taken cause anxiety . Although studies have examined the effects of music therapy in various fields of dentistry [ , , ], no study has looked at its effects on patients with moderate dental anxiety during restorative treatment. For these reasons, the aim of this study is to examine the effect of listening to music during restorative dental treatments on patients with moderate dental anxiety. The results obtained will shed light on the effect of music therapy as an anxiety reduction method.
The present study is a parallel group randomized controlled clinical trial. This study was approved for its medical ethics with protocol number 2020/41 by the Karadeniz Technical University Rectorate KTU Faculty of Medicine Scientific Research Ethics Committee. The study was carried out between January 2021 and January 2022. It was conducted in the Karadeniz Technical University Faculty of Dentistry (Department of Restorative Dentistry). The study followed the Declaration of Helsinki on medical protocol and ethics. The methodology was performed within the parameters of the CONSORT statement guidelines (Fig. ) and was registered at www.clinicaltrials.gov . Subjects and sample size The GPower program was used to calculate the sample size. The sample size was calculated based on the results of the previous study , by taking into consideration the salivary cortisol levels in terms of differences in control and experimental groups. A sample calculation was made by providing an alpha error = 0.05, beta error = 0.20, and the effect size as 0.74. The minimum required sample was calculated as 30 people for each group. Moreover, regarding the potential data loss (approximately 20%), the final total sample size was obtained as 70 ( n = 35 for each group). Using the MDAS in Turkish, individuals with moderate dental anxiety scores (between 10 and 18) were determined. Intraoral examinations of all patients had been performed, and their panoramic radiographies had been taken before the day of dental treatment. All participants were informed about the aim of the present study. Both male and female patients who were older than 18 years of age, did not need an emergency dental treatment, had at least one case of occlusal dental caries, did not use any medication, and did not have a systemic, chronic, endocrine, and mental disease were included in the study. Smokers were also included in the study. Those patients with hearing loss, having any physical disability, who were pregnant, and who were lactating were excluded. Patients who had any prior experience with treating dentist were excluded from the study. Moreover, teeth with dental cavities in the adjacent teeth where the rubber dam would not be placed and dental caries closer than 1 mm to the pulp tissue were not included in the study. Patients who wanted to take part in the research voluntarily signed an informed consent form. They were divided into two groups by an allocation concealment process (using sealed/opaque envelopes). The control group contained the ones who did not listen to music. The experimental group contained the ones who did listen to music. The patients were asked to quit smoking 2 h before their dental treatment . They were also asked to stop drinking tea, coffee, and alcohol 24 h before their dental treatment [ – ]. Since heavy exercise may affect salivary cortisol, patients were told to come to the appointment without having done any heavy exercise 24 h before the treatment . Treatment of patients in the control group All cases and measurements were performed by the same restorative treatment specialist (E. K.) who had 4 years of experience. Patients did not have any prior experience with treating dentist before. Patients who came to the clinic on the day of their treatment rested in the dentist chair for 5 min . Their gender, marital status, educational status, age, height, and weight were recorded. Blood pressure, heart rate, body temperature, oxygen saturation, and salivary cortisol levels of the patients were measured before their treatments. During the treatment session, the dentist (E. K.) communicated with the patient only for the procedure and cooperation. The dentist (E. K.) recorded the started treatment time on a paper. While posterior superior alveolar nerve block anesthesia was performed for maxillary molars, inferior alveolar nerve block anesthesia was performed for mandibular molars in a standardized manner. The sixth mandibular and maxillary molars were supplemented by infiltration injections. No positive aspirations, which affected the results, occurred. Only one carpule of local anesthetic without vasoconstrictor (Safecaine, Vem İlaç San. ve Tic. Anonim Şirketi, Turkey) was applied in total by a disposable dental syringe to the patients. The rubber dam was placed on the tooth. After the tooth decay was cleaned and the appropriate form was given to the cavity, the heart rate, body temperature, systolic/diastolic blood pressure, and oxygen saturation were re-measured . Then, the restoration phase began. Acid etching (OptiBond, Kerr, California, USA) was applied to the enamel of the teeth for 30 s using selective etching and then washed. The cavity was dried. Dental bonding (Single Bond Universal Adhesive, 3M ESPE, USA) was applied and light cured according to the manufacturer’s recommendation. Dental composite resin (Filtek P60, 3M ESPE, USA) was cured for 20 s with a light curing unit (Elipar S10, 1200 mW/cm 2 , 3M ESPE, Saint Paul, USA). After finishing and polishing the restoration, the rubber dam was removed. The restoration duration was recorded. The blood pressure, heart rate, body temperature, oxygen saturation, and salivary cortisol levels were re-measured, and the patients were asked to answer the MDAS again. Routine recommendations were given, and the patients’ gender, marital status, educational status, restored teeth number, age, height, and weight were recorded. Treatment of patients in the experimental group Patients who came to the clinic on the day of their treatment rested in the dentist chair for 5 min . Within this period, the patients were asked their favorite music. In line with the patients’ request, their music was selected and started to play for them without headphones. It was confirmed whether they could clearly hear the sound. The sound level was adjusted according to the patients’ request so that they could hear it easily without being disturbed. They were informed that once they were ready, the dental treatment would begin. Then, the dentist (E. K.) started the treatment by measuring the parameters. Physiological measurements and restoration steps for the control group were repeated in the same way for the patients in the experimental group. Music listening continued all over the session. Measurement of physiologic parameters While blood pressure, oxygen saturation, heart rate, and body temperature were measured three times, salivary cortisol levels and MDAS responses were measured twice due to the rubber dam applied to the patients and the possibility of blood in the patients’ saliva that could cause erroneous results in the salivary cortisol measurements . To observe heart rate, systolic and diastolic blood pressures, oxygen saturation, body temperature changes of patients, and changes in anxiety between the stages of dental restoration, three measurements were made : once before, once during, and once after the restorative treatment. Systolic and diastolic blood pressure were measured with a manual blood pressure monitor (Perfect, Erka, Germany). The blood pressure measurements were performed on the right arm of the patients at the heart level and recorded as mmHg on a paper. Oxygen saturation and heart rate were measured with a finger-type portable pulse oximeter device (VZN, Ankara, Turkey) on the right index finger. The body temperature was measured on the forehead by an infrared temporal non-contact thermometer (TG8818 N, Hunan Tuogao Medical Technology, Republic of China) and recorded in Celsius. Saliva samples were collected from each patient just before and immediately after the dental treatments. The patients were told to spit into the sterile and disposable tubes (Sali-Tubes 100 [SLV-4158], DRG, Germany). These tubes were used for the collection and storage of the unstimulated saliva samples. Half of the tubes were filled with saliva without blood contamination. The saliva samples were collected between 9:00 and 12:00 a.m. so that the measurements were not affected by the circadian rhythm and cortisol fluctuations . After the saliva samples were collected, they were placed in a −80 °C refrigerator. Laboratory measurement of cortisol levels of saliva samples The samples were analyzed by a blind researcher after the treatments of all patients were finished and at the same time for each group. They were defrosted and centrifuged at 3.000×g for 10 min. An enzyme-linked immunosorbent assay kit (DRG, Germany, REF: SLV-4635) was used to determine the amount of cortisol in the saliva samples. Absorbance readings were taken at 450-nm wavelengths, and the results were recorded as nanograms per milliliter. Statistical analysis The data was analyzed with IBM SPSS V23. Compliance with the normal distribution was evaluated with the Kolmogorov-Smirnov test. Except heart rate, all of the data were non-normally distributed. A chi-square test was used to compare categorical variables according to the groups. An independent two-sample t -test was used to compare the normally distributed data (heart rate) according to the paired groups, and a Mann-Whitney U test was used to compare the non-normally distributed data. A Wilcoxon test was used to compare data that was not normally distributed according to the pairwise time within the group. A repeated analysis of variance was used to compare normally distributed data three or more times, and the Friedman test was used to compare data that was not normally distributed. Analysis results were presented as mean ± standard deviation for the quantitative data. The categorical data as a deviation and median (minimum–maximum) was presented as the frequency (percentage). The significance level was taken as p < 0.05.
The GPower program was used to calculate the sample size. The sample size was calculated based on the results of the previous study , by taking into consideration the salivary cortisol levels in terms of differences in control and experimental groups. A sample calculation was made by providing an alpha error = 0.05, beta error = 0.20, and the effect size as 0.74. The minimum required sample was calculated as 30 people for each group. Moreover, regarding the potential data loss (approximately 20%), the final total sample size was obtained as 70 ( n = 35 for each group). Using the MDAS in Turkish, individuals with moderate dental anxiety scores (between 10 and 18) were determined. Intraoral examinations of all patients had been performed, and their panoramic radiographies had been taken before the day of dental treatment. All participants were informed about the aim of the present study. Both male and female patients who were older than 18 years of age, did not need an emergency dental treatment, had at least one case of occlusal dental caries, did not use any medication, and did not have a systemic, chronic, endocrine, and mental disease were included in the study. Smokers were also included in the study. Those patients with hearing loss, having any physical disability, who were pregnant, and who were lactating were excluded. Patients who had any prior experience with treating dentist were excluded from the study. Moreover, teeth with dental cavities in the adjacent teeth where the rubber dam would not be placed and dental caries closer than 1 mm to the pulp tissue were not included in the study. Patients who wanted to take part in the research voluntarily signed an informed consent form. They were divided into two groups by an allocation concealment process (using sealed/opaque envelopes). The control group contained the ones who did not listen to music. The experimental group contained the ones who did listen to music. The patients were asked to quit smoking 2 h before their dental treatment . They were also asked to stop drinking tea, coffee, and alcohol 24 h before their dental treatment [ – ]. Since heavy exercise may affect salivary cortisol, patients were told to come to the appointment without having done any heavy exercise 24 h before the treatment .
All cases and measurements were performed by the same restorative treatment specialist (E. K.) who had 4 years of experience. Patients did not have any prior experience with treating dentist before. Patients who came to the clinic on the day of their treatment rested in the dentist chair for 5 min . Their gender, marital status, educational status, age, height, and weight were recorded. Blood pressure, heart rate, body temperature, oxygen saturation, and salivary cortisol levels of the patients were measured before their treatments. During the treatment session, the dentist (E. K.) communicated with the patient only for the procedure and cooperation. The dentist (E. K.) recorded the started treatment time on a paper. While posterior superior alveolar nerve block anesthesia was performed for maxillary molars, inferior alveolar nerve block anesthesia was performed for mandibular molars in a standardized manner. The sixth mandibular and maxillary molars were supplemented by infiltration injections. No positive aspirations, which affected the results, occurred. Only one carpule of local anesthetic without vasoconstrictor (Safecaine, Vem İlaç San. ve Tic. Anonim Şirketi, Turkey) was applied in total by a disposable dental syringe to the patients. The rubber dam was placed on the tooth. After the tooth decay was cleaned and the appropriate form was given to the cavity, the heart rate, body temperature, systolic/diastolic blood pressure, and oxygen saturation were re-measured . Then, the restoration phase began. Acid etching (OptiBond, Kerr, California, USA) was applied to the enamel of the teeth for 30 s using selective etching and then washed. The cavity was dried. Dental bonding (Single Bond Universal Adhesive, 3M ESPE, USA) was applied and light cured according to the manufacturer’s recommendation. Dental composite resin (Filtek P60, 3M ESPE, USA) was cured for 20 s with a light curing unit (Elipar S10, 1200 mW/cm 2 , 3M ESPE, Saint Paul, USA). After finishing and polishing the restoration, the rubber dam was removed. The restoration duration was recorded. The blood pressure, heart rate, body temperature, oxygen saturation, and salivary cortisol levels were re-measured, and the patients were asked to answer the MDAS again. Routine recommendations were given, and the patients’ gender, marital status, educational status, restored teeth number, age, height, and weight were recorded.
Patients who came to the clinic on the day of their treatment rested in the dentist chair for 5 min . Within this period, the patients were asked their favorite music. In line with the patients’ request, their music was selected and started to play for them without headphones. It was confirmed whether they could clearly hear the sound. The sound level was adjusted according to the patients’ request so that they could hear it easily without being disturbed. They were informed that once they were ready, the dental treatment would begin. Then, the dentist (E. K.) started the treatment by measuring the parameters. Physiological measurements and restoration steps for the control group were repeated in the same way for the patients in the experimental group. Music listening continued all over the session.
While blood pressure, oxygen saturation, heart rate, and body temperature were measured three times, salivary cortisol levels and MDAS responses were measured twice due to the rubber dam applied to the patients and the possibility of blood in the patients’ saliva that could cause erroneous results in the salivary cortisol measurements . To observe heart rate, systolic and diastolic blood pressures, oxygen saturation, body temperature changes of patients, and changes in anxiety between the stages of dental restoration, three measurements were made : once before, once during, and once after the restorative treatment. Systolic and diastolic blood pressure were measured with a manual blood pressure monitor (Perfect, Erka, Germany). The blood pressure measurements were performed on the right arm of the patients at the heart level and recorded as mmHg on a paper. Oxygen saturation and heart rate were measured with a finger-type portable pulse oximeter device (VZN, Ankara, Turkey) on the right index finger. The body temperature was measured on the forehead by an infrared temporal non-contact thermometer (TG8818 N, Hunan Tuogao Medical Technology, Republic of China) and recorded in Celsius. Saliva samples were collected from each patient just before and immediately after the dental treatments. The patients were told to spit into the sterile and disposable tubes (Sali-Tubes 100 [SLV-4158], DRG, Germany). These tubes were used for the collection and storage of the unstimulated saliva samples. Half of the tubes were filled with saliva without blood contamination. The saliva samples were collected between 9:00 and 12:00 a.m. so that the measurements were not affected by the circadian rhythm and cortisol fluctuations . After the saliva samples were collected, they were placed in a −80 °C refrigerator.
The samples were analyzed by a blind researcher after the treatments of all patients were finished and at the same time for each group. They were defrosted and centrifuged at 3.000×g for 10 min. An enzyme-linked immunosorbent assay kit (DRG, Germany, REF: SLV-4635) was used to determine the amount of cortisol in the saliva samples. Absorbance readings were taken at 450-nm wavelengths, and the results were recorded as nanograms per milliliter.
The data was analyzed with IBM SPSS V23. Compliance with the normal distribution was evaluated with the Kolmogorov-Smirnov test. Except heart rate, all of the data were non-normally distributed. A chi-square test was used to compare categorical variables according to the groups. An independent two-sample t -test was used to compare the normally distributed data (heart rate) according to the paired groups, and a Mann-Whitney U test was used to compare the non-normally distributed data. A Wilcoxon test was used to compare data that was not normally distributed according to the pairwise time within the group. A repeated analysis of variance was used to compare normally distributed data three or more times, and the Friedman test was used to compare data that was not normally distributed. Analysis results were presented as mean ± standard deviation for the quantitative data. The categorical data as a deviation and median (minimum–maximum) was presented as the frequency (percentage). The significance level was taken as p < 0.05.
There were no statistically significant differences between the distributions of gender, marital status, educational status, and restored teeth between the groups ( p > 0.05) (Table ). There were no statistically significant differences between the median values of age, height, weight, BMI, and restoration duration between the groups ( p > 0.05) (Table ). When the intra-group comparison results of the blood pressures were examined (Table ), a statistically significant difference was found between the systolic ( p < 0.001) and diastolic median values ( p = 0.004) in the experimental group over time. Looking at intergroup comparisons, at the end of the treatment, the diastolic blood pressure decreased in the experimental group significantly ( p < 0.042). A statistically significant difference was found between the heart rate values over time (Table ) in the control group ( p = 0.001) and in the experimental group ( p < 0.001). There were no significant differences between the groups at different times throughout the treatment ( p > 0.05). A statistically significant difference was found for the cortisol values (Table ) before ( p < 0.001) and after the dental treatments ( p < 0.001) between the groups. The cortisol levels of the experimental group decreased significantly ( p = 0.001), while there was no significant difference in the control group over time ( p = 0.512). At the end of the treatment, cortisol levels of the experimental group were statistically higher than the control group. Although there was no difference between the groups before the dental treatment in terms of the MDAS score (Table ), after the dental treatment, it decreased in both groups. This decrease was higher for the experimental group ( p < 0.001) than for the control group ( p = 0.003). After the dental treatment, there were statistically significant differences between the MDAS scores ( p = 0.001). There was no statistically significant difference between the distributions of the other variables (body temperature and oxygen saturation) (Table ) at different times for intra- and between-group comparisons ( p > 0.05).
Dental anxiety is a complex inhibitor that could affect patients’ oral health. Patients with this anxiety may avoid making dental appointments, causing delayed treatment to early carious lesions. Patients may avoid dental treatments despite experiencing pain. Dentists can diagnose dental anxiety and offer the best treatment options. However, this is impossible when dental check-ups are evaded. For dentists to provide the best treatment options , it is important for them to analyze the anxiety levels of patients. In this respect, many studies have evaluated the effect of music therapy on anxiety during endodontic treatment [ , , ], dental extractions, implant surgery [ , , – ], and restorative treatment . However, none of these studies has evaluated moderate dental anxiety. Conducting this type of study is important because restorative treatments are among the most performed dental therapies, and to our knowledge, this is the first article to investigate the effect of music on patients with moderate dental anxiety during restorative treatment. In this study, although there was a statistically significant decrease in systolic and diastolic blood pressure, heart rate, salivary cortisol, and the MDAS scores in the experimental group during dental treatment, at the end of the treatment, it was only the MDAS scores that were significantly lower than the control group. Kritsidima et al. stated that the MDAS showed both immediate and anticipatory anxiety. Humphries et al. found that the MDAS was significantly correlated with the avoidance of dental treatments. In light of these findings and the results of the present study, it can be concluded that although music therapy does not have a significant effect on physiological parameters of dental anxiety during restorative dental treatment, patients do find the effect of music on reducing dental anxiety to be positive, which could lower anxiety levels about follow-up dental treatments. This could lead to more frequent dental visits, early treatment, and better oral health and contribute to increased patient satisfaction and less pain. Thus, music therapy may be an option for patients with moderate dental anxiety during restorative dental treatments for the stated positive results, considering its ease of use and the absence of any known deleterious effects. The MDAS is simple, easy, and practical. It is not time-consuming, and it can be implemented quickly . Its suitability for the Turkish population has been proven in studies, and the validity and reliability of the Turkish version were high . Moreover, the MDAS has been described as a useful tool to predict sympathetic nervous activity before the start of treatment . For these reasons, we preferred the MDAS in this study. However, at the beginning of the study, the MDAS scores of both groups were 12. Because of their mild to moderate levels of anxiety, the patients may have had more control over their dental treatments and were able to manage their anxiety without music. This could be why the patients were not able to fully benefit from music therapy in this study. Salivary cortisol is a reliable marker of dental-related anxiety. Collecting cortisol hormones from saliva is an easy and safe method for patients and dentists. The procedure is non-invasive and practical . Salivary cortisol is highly correlated with serum cortisol levels . However, as cortisol is a physiologic parameter for body fluids, various factors, such as socioeconomic status, day-to-day events, previous dental experiences, and other fear-inducing factors may affect its amount . Mejia-Rubalcava et al. state that initially, both groups in their study had similar levels of salivary cortisol. However, in the second measurement, it was statistically lower in the experimental group. Wazzan et al. found no significant difference in terms of salivary cortisol at the beginning and end of the study between both groups. In contrast to these, in the present study, the cortisol levels of the experimental group were higher than in the control group at the beginning and end of the study. This could be related to the experimental group containing participants who visited dentists less frequently and had more traumatic dental experiences than the others. Toward the end of the treatment, they relaxed. Another possibility is that initially, listening to music was perceived by patients as distressing. However, the significant decrease in cortisol toward the end of the treatment could indicate that they had become accustomed to the dental treatment with music therapy. The higher anxiety of the patients in the experimental group may have affected the results and caused the patients not to benefit from the music adequately. Heart rate, blood pressure, oxygen saturation, and body temperature are physiological parameters affected by anxiety . However, there is no consensus about the effect of music therapy on these parameters. Di Nasso et al. state that music therapy decreased systolic blood pressure, diastolic blood pressure, and heart rate during root canal treatments. In contrast to that study, Kupeli and Gulnahar found no significant difference between the groups regarding heart rate and mean arterial pressure. They concluded that this could be because of the individual differences in the response of the parasympathetic nervous system during the stimulation of the sympathetic nervous system. Moreover, Mejia-Rubalcava found that significant differences were registered in the systolic and diastolic pressure, heart rate, and body temperature for the experimental group, as the oxygen saturation did not change. Looking at anxiety parameters in this study, we can estimate that the patients had the highest degree of dental anxiety at the beginning of the treatment. Other than a small decrease in diastolic blood pressure in the experimental group, there was no statistically significant difference between the two groups in terms of heart rate, systolic blood pressure, oxygen saturation, and body temperature at the end of the treatment. This small decrease could also be because the patients got used to the treatment and were relieved when it was over. The second and third measurements of the pulse values in the control and experimental groups being similar also supports this. There are some limitations to this in vivo study. The small sample size comprises mostly females and young people. Thus, it may not be a population-representative sample. The patients were not asked if they had undergone a previous dental treatment or traumatic visit before. It was also impossible to evaluate their existing psychological state, which might affect the results. Moreover, it was not confirmed if the measurement tools were valid and reliable. The patients were determined based on one questionnaire, the MDAS, and they may have exaggerated their answers. Among the anxiety parameters that can be detected in saliva, only cortisol measurements were evaluated, and the other parameters were ignored. The different salivary cortisol levels between the two groups at the beginning of the treatment may have affected the results, and during the restoration process, it was impossible to blind the patients and the researcher who performed the dental treatment. Only occlusal dental cavities on the molar teeth were restored, and the music frequency was not evaluated. Listening to the same music at different frequencies could cause different psychological effects . The place and measurements of the physiological parameters could have also caused more stress in the patients. The study was performed in a dental hospital during the COVID-19 pandemic, which could have impacted anxiety levels during treatments. Moreover, since it is a cross-sectional study, it may not be reliable in determining the cause-effect relationship. Confusing situations may not be determined clearly. It is suggested that future studies are conducted in consideration of these factors with substantial follow-ups to determine the effectiveness of music therapy on lowering anxiety levels.
Music therapy has not a significant effect on physiologic parameters of patients with moderate dental anxiety during restorative treatment. According to this study, it may be concluded that music therapy may have an effect on lowering self-reported dental anxiety levels of the patients.
|
A comparative analysis of human and AI performance in forensic estimation of physical attributes | 0164229c-5e2b-4fcc-ac5c-939d68bad121 | 10036317 | Forensic Medicine[mh] | Despite recent advances in artificial intelligence (AI) promising to revolutionise automated decision making, concerns are now being raised regarding fairness and efficacy across a range of high-impact fields, including the criminal justice system. The increasing use of algorithms in incarceration and rehabilitation has been widely scrutinized, ranging from policing , to criminal sentencing and pretrial detention . Use of these automated approaches has raised serious concerns regarding civil liberties and due process rights . The COMPAS algorithm for predicting recidivism, for example, has been found to not only reinforce problematic racial and social biases , but also perform no more accurately than untrained humans . Similarly, in 2018, Buolamwini and Gebru found that popular facial verification and identification technologies— the use of which within law enforcement remains largely unregulated —produced disproportionately higher error rates for racial minorities . It is, of course, appropriate to consider replacing or augmenting potentially error-prone human judgement and analysis with the goal of a more equitable criminal justice system. Here we focus on the growing trend of citizen policing in which, with a high-resolution camera in every hand, every-day citizens are playing an increasingly vital role in documenting everything from major global events to human-rights violations, police misconduct, and neighborhood crimes. At the same time, advances in artificial intelligence have made identifying individuals in images easier. And yet, reliable forensic identification is riddled with bias and errors , . The National Registry of Exonerations, for example, reports that between 1989 and 2019, flawed forensic techniques contributed to almost one quarter of wrongful convictions in the US. Some effort has gone into documenting and trying to address these issues in AI-based face recognition , but less attention has been paid to basic forensic identification based on physical traits like height and weight. To illustrate this point, in 2008 George Powell was identified as a suspect in a string of armed robberies. A store clerk initially identified the robber as 5 [12pt]{minimal}
$$^$$ ′ 6 [12pt]{minimal}
$$^{}$$ ″ tall, and eventually identified Powell in a lineup. Powell stands at 6 [12pt]{minimal}
$$^$$ ′ 3 [12pt]{minimal}
$$^{}$$ ″ . From video surveillance, an expert measured the robber to be 6 [12pt]{minimal}
$$^$$ ′ 1 [12pt]{minimal}
$$^{}$$ ″ . Powell was convicted and sentenced to 28 years in prison. After his conviction, two new experts concluded the robber was less than 5 [12pt]{minimal}
$$^$$ ′ 10 [12pt]{minimal}
$$^{}$$ ″ , after which the original expert adjusted his estimate to a range of 6 [12pt]{minimal}
$$^$$ ′ 1 [12pt]{minimal}
$$^{}$$ ″ to 5 [12pt]{minimal}
$$^$$ ′ 10 [12pt]{minimal}
$$^{}$$ ″ . Due in part to these inconsistencies, Powell’s conviction was vacated in 2018, and he was granted a new trial. Because physical attributes like height, weight, age, and race are fundamental to forensic identification, it is essential to validate the accuracy of new and traditional tools. Height and weight estimation could also play a crucial role in increasing the reliability of photographic identification. If, for example, weight can be estimated to within an accuracy of [12pt]{minimal}
$$5\%$$ 5 % , then based on the distribution of US adult male weights , some [12pt]{minimal}
$$90\%$$ 90 % of men could be eliminated from consideration from this single measurement. Despite its seeming simplicity, many factors make it challenging to accurately estimate height and weight from a single image. Due to spinal compression, for example, height fluctuates daily by up to 1.9 cm ; due to body pose, apparent height in an image can vary by up to 6 cm ; and shoes, hair, and headwear further obscure a person’s true height. Recent advances in AI and computer vision have led to spectacular leaps in image understanding and modeling of the human form (e.g., , ). We evaluate the accuracy with which AI-based tools—and for comparison—expert photogrammetrists and non-experts can estimate a person’s height and weight from a single image.
Data set A total of 58 participants (33 women and 25 men) were recruited from the UC Berkeley campus and photographed in two settings: (1) a studio setting with a fixed white background and artificial lighting with a tripod-mounted DSLR camera (4000 [12pt]{minimal}
$$$$ × 6000 pixels); and (2) an in-the-wild setting emulating a CCTV-like scene in which a narrow corridor was photographed by a ceiling-mounted GoPro camera (5184 [12pt]{minimal}
$$$$ × 3888 pixels). Each participant was assigned an anonymized identifier and photographed in the studio setting in eight neutral poses, Fig. a, six dynamic poses, Fig. b, and one neutral pose while standing next to a reference object (the same stool was used for all participants), Fig. c. Each participant was photographed in the wild in two static, Fig. d, and three dynamic poses. This process yielded a total of 812 no-reference studio images, 58 reference studio images, and 290 in-the-wild images. Each participant’s height and weight was measured and recorded alongside their anonymized identifier. The collected female/male heights are normally distributed with a mean of 161.1/176.1 cm and a standard deviation of 5.3/8.3 cm; the average US adult female/male height is 161/175 cm with a standard deviation of 7.0/7.4 cm . The collected female/male weights are 60.9/78.4 kg with a standard deviation of 11.4/12.9 kg; the average US adult female/male weight is 78.7/90.8 kg with a standard deviation of 19.7/19.8 kg . While our participants’ heights closely follows the national average, our participants weighed approximately [12pt]{minimal}
$$20\%$$ 20 % less than the national average and are less variable (presumably because they were drawn primarily from a University student population). Each participant was paid [12pt]{minimal}
$$\$20$$ $ 20 . AI Recent advances in machine learning and computer vision have led to impressive results for estimating body shape and pose from a single image . We previously extended this system to yield state-of-the-art body shape and pose estimation , . Here we briefly describe this system. A full-body, 3D model is fit to an image of a person using an augmented version of SMPLify-X . The original SMPLify-X extracts 2D keypoints from the body and face, from which a 3D model is automatically fit. Although this model can accurately capture complex body poses, it does not incorporate body shape. This is because the model fitting relies only on the extracted 2D skeletal keypoints and does not consider the body shape depicted in the image. An augmented version of this model incorporates into the 3D modeling an additional parameter that captures the overall body shape, yielding more accurate estimates of body shape and size, Fig. e. Although the 3D body model is estimated in real-world units, this metric reconstruction is highly inaccurate , even while the overall body pose and shape are well estimated. We, therefore, adopt a different approach that scales the estimated 3D model based on a gender-specific average inter-pupillary distance (IPD). The IPD is relatively consistent, with an average adult IPD for women/men of 6.17/6.40 cm with a standard deviation of 0.36/0.34 cm . Because our 3D models do not have pupils, the pupil center is specified as the midway point between the left and right corners of the eye. Once scaled, the 3D model is reposed into a neutral, upright pose, from which the person’s height is measured as the distance from the top of the head to a plane formed by three points on the bottom of the feet. The person’s weight is measured as the volume of the 3D model, converted to kilograms by multiplying by 1023 kg/m [12pt]{minimal}
$$^3$$ 3 , corresponding to a gender-agnostic average body fat of [12pt]{minimal}
$$34\%$$ 34 % . Experts We recruited 10, US-based, certified photogrammetrists (certification requires a minimum of between four and six years of experience depending on the governing body). Each expert was provided with a random subset of five in-the-wild images (each image depicted a different person) and asked to estimate the person’s height and weight (one expert declined to estimate weight). Each expert was provided with a schematic diagram of the scene with two real-world measurements consisting of the width of the back door into the hallway and the distance between the back door and the top of the stairs. Non-experts We recruited 325 participants from Amazon’s Mechanical Turk platform. Unlike the experts described in the previous section, who made height and weight estimates from only the in-the-wild images, our non-experts were tasked with making estimates from the no-reference studio images, the referenced studio-images, or the in-the-wild images. A representative subset of 290 (out of 812) no-reference studio images were partitioned into five non-overlapping sets of 58 images in which each photographed participant appeared only once. The 290 in-the-wild images were similarly partitioned into five non-overlapping sets of 58 images each. The 58 reference studio images were placed into a single set. On entry into the study, each participant was assigned a random set from the above 11 possible subsets. Shown one image at a time, in random order, participants were asked to estimate the height and weight of the person depicted in the photo. Unlike the experts and AI, no additional information was provided to these non-experts. Randomly interspersed within the 58 images were four catch trials consisting of stock photos clearly annotated with the subject’s height and weight. If a participant failed any of the catch trials, their entire set of responses were excluded. A total of 65 out of 325 participants failed to correctly complete the catch trials, and another 24 failed to complete the study, yielding a total of 236 valid responses. Participants were paid $5.00, but were not paid if they failed any of the catch trials. Each image was analyzed by an average of 22 non-experts. Denoting the estimated height from non-expert j for image i as [12pt]{minimal}
$$_{i,j}$$ h ~ i , j with true height [12pt]{minimal}
$$h_{i}$$ h i , the median individual accuracy is computed as [12pt]{minimal}
$$_j( |_{i,j} - h_{i} |)$$ median j | h ~ i , j - h i | ; the median crowd accuracy is computed as [12pt]{minimal}
$$|_j(_{i,j}) - h_{i} |$$ | median j ( h ~ i , j ) - h i | . The individual and crowd weight errors are estimated in the same way. The median error across all images are reported in Table in both absolute units (cm/kg) and as a percent of base height and weight. A median (as compared to a mean) is employed because responses within and across images are not normally distributed. Human subjects All data collection was approved by the UC Berkeley Committee for Protection of Human Subjects (2022-01-14999). All participants provided informed consent prior to their participation, and data collection was performed in accordance with relevant guidelines and regulations.
A total of 58 participants (33 women and 25 men) were recruited from the UC Berkeley campus and photographed in two settings: (1) a studio setting with a fixed white background and artificial lighting with a tripod-mounted DSLR camera (4000 [12pt]{minimal}
$$$$ × 6000 pixels); and (2) an in-the-wild setting emulating a CCTV-like scene in which a narrow corridor was photographed by a ceiling-mounted GoPro camera (5184 [12pt]{minimal}
$$$$ × 3888 pixels). Each participant was assigned an anonymized identifier and photographed in the studio setting in eight neutral poses, Fig. a, six dynamic poses, Fig. b, and one neutral pose while standing next to a reference object (the same stool was used for all participants), Fig. c. Each participant was photographed in the wild in two static, Fig. d, and three dynamic poses. This process yielded a total of 812 no-reference studio images, 58 reference studio images, and 290 in-the-wild images. Each participant’s height and weight was measured and recorded alongside their anonymized identifier. The collected female/male heights are normally distributed with a mean of 161.1/176.1 cm and a standard deviation of 5.3/8.3 cm; the average US adult female/male height is 161/175 cm with a standard deviation of 7.0/7.4 cm . The collected female/male weights are 60.9/78.4 kg with a standard deviation of 11.4/12.9 kg; the average US adult female/male weight is 78.7/90.8 kg with a standard deviation of 19.7/19.8 kg . While our participants’ heights closely follows the national average, our participants weighed approximately [12pt]{minimal}
$$20\%$$ 20 % less than the national average and are less variable (presumably because they were drawn primarily from a University student population). Each participant was paid [12pt]{minimal}
$$\$20$$ $ 20 .
Recent advances in machine learning and computer vision have led to impressive results for estimating body shape and pose from a single image . We previously extended this system to yield state-of-the-art body shape and pose estimation , . Here we briefly describe this system. A full-body, 3D model is fit to an image of a person using an augmented version of SMPLify-X . The original SMPLify-X extracts 2D keypoints from the body and face, from which a 3D model is automatically fit. Although this model can accurately capture complex body poses, it does not incorporate body shape. This is because the model fitting relies only on the extracted 2D skeletal keypoints and does not consider the body shape depicted in the image. An augmented version of this model incorporates into the 3D modeling an additional parameter that captures the overall body shape, yielding more accurate estimates of body shape and size, Fig. e. Although the 3D body model is estimated in real-world units, this metric reconstruction is highly inaccurate , even while the overall body pose and shape are well estimated. We, therefore, adopt a different approach that scales the estimated 3D model based on a gender-specific average inter-pupillary distance (IPD). The IPD is relatively consistent, with an average adult IPD for women/men of 6.17/6.40 cm with a standard deviation of 0.36/0.34 cm . Because our 3D models do not have pupils, the pupil center is specified as the midway point between the left and right corners of the eye. Once scaled, the 3D model is reposed into a neutral, upright pose, from which the person’s height is measured as the distance from the top of the head to a plane formed by three points on the bottom of the feet. The person’s weight is measured as the volume of the 3D model, converted to kilograms by multiplying by 1023 kg/m [12pt]{minimal}
$$^3$$ 3 , corresponding to a gender-agnostic average body fat of [12pt]{minimal}
$$34\%$$ 34 % .
We recruited 10, US-based, certified photogrammetrists (certification requires a minimum of between four and six years of experience depending on the governing body). Each expert was provided with a random subset of five in-the-wild images (each image depicted a different person) and asked to estimate the person’s height and weight (one expert declined to estimate weight). Each expert was provided with a schematic diagram of the scene with two real-world measurements consisting of the width of the back door into the hallway and the distance between the back door and the top of the stairs.
We recruited 325 participants from Amazon’s Mechanical Turk platform. Unlike the experts described in the previous section, who made height and weight estimates from only the in-the-wild images, our non-experts were tasked with making estimates from the no-reference studio images, the referenced studio-images, or the in-the-wild images. A representative subset of 290 (out of 812) no-reference studio images were partitioned into five non-overlapping sets of 58 images in which each photographed participant appeared only once. The 290 in-the-wild images were similarly partitioned into five non-overlapping sets of 58 images each. The 58 reference studio images were placed into a single set. On entry into the study, each participant was assigned a random set from the above 11 possible subsets. Shown one image at a time, in random order, participants were asked to estimate the height and weight of the person depicted in the photo. Unlike the experts and AI, no additional information was provided to these non-experts. Randomly interspersed within the 58 images were four catch trials consisting of stock photos clearly annotated with the subject’s height and weight. If a participant failed any of the catch trials, their entire set of responses were excluded. A total of 65 out of 325 participants failed to correctly complete the catch trials, and another 24 failed to complete the study, yielding a total of 236 valid responses. Participants were paid $5.00, but were not paid if they failed any of the catch trials. Each image was analyzed by an average of 22 non-experts. Denoting the estimated height from non-expert j for image i as [12pt]{minimal}
$$_{i,j}$$ h ~ i , j with true height [12pt]{minimal}
$$h_{i}$$ h i , the median individual accuracy is computed as [12pt]{minimal}
$$_j( |_{i,j} - h_{i} |)$$ median j | h ~ i , j - h i | ; the median crowd accuracy is computed as [12pt]{minimal}
$$|_j(_{i,j}) - h_{i} |$$ | median j ( h ~ i , j ) - h i | . The individual and crowd weight errors are estimated in the same way. The median error across all images are reported in Table in both absolute units (cm/kg) and as a percent of base height and weight. A median (as compared to a mean) is employed because responses within and across images are not normally distributed.
All data collection was approved by the UC Berkeley Committee for Protection of Human Subjects (2022-01-14999). All participants provided informed consent prior to their participation, and data collection was performed in accordance with relevant guidelines and regulations.
Shown in Table is a summary of the height/weight estimation errors for AI, expert, non-expert, and baseline from 1160 images across our three data sets (Fig. ). Shown in Fig. are the error distributions annotated with the median and [12pt]{minimal}
$$95\%$$ 95 % confidence intervals computed from 1000 bootstrap iterations. The baseline estimator corresponds to simply using a gender-specific average US adult height/weight for every image (see “ ” in “ ”). With a median height error of only 4.2 cm, this baseline predictor is surprisingly good, outperformed only by the non-expert crowd. With a median weight error of 17.5 kg, however, the baseline is the worst performing. This asymmetry is due to the fact that gendered adult heights have relatively low variance as compared to weight. At first glance, the non-expert crowd is more accurate than all others even in the no-reference studio images in which height/weight estimates are made in the absence of any contextual information (Fig. a, b). Of the 290 in-the-wild images, we obtained height/weight estimates from all groups for 50/44 images (one expert declined to estimate weight). From this subset, a 5-way Friedman test reveals a significant difference in the error distribution of height ( [12pt]{minimal}
$$p = 3.5 10^{-6}$$ p = 3.5 × 10 - 6 ) and weight ( [12pt]{minimal}
$$p = 9.8 10^{-6}$$ p = 9.8 × 10 - 6 ). Following this, we performed 10 Wilcoxon two-sided rank tests on all pairs of height/weight estimates. Shown in the lower portion of Fig. are the resulting p -values where statistical significance is set at [12pt]{minimal}
$$p < 0.005$$ p < 0.005 , incorporating a Bonferroni correction to adjust the baseline p-value of 0.05 by the 10 pairwise comparisons. The AI-based height estimator is no more accurate than experts, non-experts, or baseline (guessing a gender-specific average height). Experts are no more accurate than individual non-experts, and are less accurate than the non-expert crowd and baseline. Neither the non-expert crowd nor individual are more accurate than baseline. The AI-based weight estimator is no more accurate than experts and individual non-experts and is less accurate then the non-expert crowd; and experts are no more accurate than non-experts. Unlike height, baseline weight is less accurate than all other groups. This asymmetry is due to the fact that the variance in adult weight is much higher than in height. What is particularly surprising about these results is that both the AI and experts had access to explicit metric measurements (IPD and door/hallway measurements, respectively), whereas the non-experts were not provided this information. It can be argued that these results only hold for our particular AI-based estimator. However, other state of the art AI estimators are as, or less, accurate than ours . We contend, therefore, that the problem of accurate height and weight estimation may be out of reach of current AI systems.
A group of two dozen non-experts outperforms AI and expert height/weight estimation even when the non-experts are provided with less information. This underwhelming performance by experts and AI should give significant pause as to how—or even if—it is reasonable to rely on these methods for forensic identification based on basic physical attributes. With a median AI-based height error of [12pt]{minimal}
$$4.4\%$$ 4.4 % , for example, a man standing at 183 cm ( [12pt]{minimal}
$$6^$$ 6 ′ ) will be estimated to within a range of 175–191 cm ( [12pt]{minimal}
$$5^ 9^{}-6^ 3^{}$$ 5 ′ 9 ″ - 6 ′ 3 ″ ), capturing a quarter of all US adult men. Our experiments were not designed to evaluate gender or racial bias, however, we qualitatively find that height and weight errors are similar for women and men; we did not have enough diversity in our data set to determine if there are any racial biases. As with any forensic identification, it will be important to determine if any such racial (or other) bias exists. The troubling state of human-based forensic identification needs critical attention , . Simply deploying AI-based tools, however, provides no guarantee that critical decision-making in criminal investigations will be any more fair or accurate, and—as our results reveal—they may make things worse. As with other automated techniques designed to replace or augment human decision making, it is critical to carefully evaluate the accuracy and potential bias in any such proposed systems. Most AI and computer-vision systems, however, are typically evaluated against previously published systems and are not directly compared to human performance. As it pertains to the criminal justice system, a machine-to-human comparison is critical to ensure that replacing or augmenting humans will not, in fact, lead to worse outcomes. One advantage of the AI-based system evaluated here is that it explicitly estimates a person’s body shape and pose, from which height and weight can be explainably determined. By contrast, purely machine-learning based approaches take a more opaque approach, attempting to learn the relationship between an image of a person and their physical attributes. In the work of , for example, the neural-network based system achieves a mean absolute height error of 8.4 cm for neutral poses and 12.1 cm for non-neutral poses; significantly worse than those reported in Table . In addition to the poor performance, this approach is not particularly explainable which—we contend—can be problematic in the criminal justice system where experts, attorneys, and judges should be able to scrutinize the inner workings of any forensic technique being used in such a potentially high-stakes setting. We have focused on forensic identification based on height and weight. Even this most basic of measurements appears to be out of reach of modern AI-based systems, casting significant doubt as to the feasibility of AI-based forensic identification based on more complex measurements or features.
|
Concordance of immunohistochemistry for predictive and prognostic factors in breast cancer between biopsy and surgical excision: a single-centre experience and review of the literature | ace62b9f-2e48-498c-b367-2024766345fe | 10036406 | Anatomy[mh] | Assessment of breast cancer biomarkers has become a staple of routine histopathology for every colleague working in this field. Assessment and quantification of oestrogen receptor (ER), progesterone receptor (PR), and c-erbB2/HER2 are used daily by the clinicians making fundamental therapeutic choices for the patients. On the other hand, Ki-67 has struggled to join this established trifecta in the routine management and risk stratification of breast cancer patients and its prognostic and predictive value is restricted to very specific settings in breast cancer; recently, the results from the monarchE study has suggested a prognostic role for Ki-67 ≥ 20% in patients with early breast cancer treated with cyclin-dependent kinase 4 and 6 (CDK4/6) inhibitors in combination with endocrine therapy . Recently, changes in the stratification of ER positivity have been put forward by ASCO/CAP through an update of the recommendations for ER and PR determination, with the formal introduction of the ER-low-positive (ER-LP, defined as 1–10% of nuclear positivity in the tumour) ; this category has been known for some time to share more similarities with basal-type triple-negative breast cancer than with the luminal group , in morphological aspects, molecular signature and clinical behaviour. In the same way, the results of the DESTINY-Breast03 trial , and the identification of the HER2-low category as a new group of patients who may significantly benefit from anti-HER2 target therapy administration, have underlined once again the importance of the c-erbB2 status and its assessment with both immunohistochemistry and molecular techniques. Core needle biopsy (CNB) is the most common method for diagnosis of breast cancer, and it has been demonstrated to be a reliable indicator of the surgical specimen results . Assessment of biomarkers on the preoperative specimen is required to administer neoadjuvant therapy to the selected patients that benefit from it and, in case of a complete pathological response, it represents the only sample of that tumour available; also for metastatic patients, who are not eligible for surgical resection, the bioptical sample is the only one that can be used to take life-changing clinical decisions. It is clear, then, the importance of a reliable assessment of these biomarkers on biopsy. Given the steadily increasing request for precise molecular characterization of breast cancer to ensure correct patient management, we aimed to retrospectively evaluate our patient cohort for reproducibility of the biopsy results for ER, PR, c-erbB2, and Ki-67, focusing also on the recently defined ER-LP and HER2-low groups. We also review the recent literature on the topic to validate our results in the context of the international available results.
Patients and clinicopathological characteristics We reviewed the records of 1654 breast cancer patients who underwent both biopsy and surgical excision at the Department of Breast Surgery, San Matteo Hospital, between January 2014 and December 2020. We excluded patients for which ER, PR, c-erBb2 or Ki-67 values were missing, patients with in situ or microinvasive-only breast cancer, patients with multifocal tumours, metastatic disease and patients who underwent neoadjuvant chemotherapy. Our final cohort comprised 923 patients and their clinical characteristics are summarized in Table . The study was conducted according to the guidelines of the Declaration of Helsinki. Pathology evaluation Samples were fixed in 10% neutral buffered formalin and embedded in paraffin before histopathological evaluation. 4-to-5 µm-thick sections were cut and stained with hematoxylin and eosin (HE), and unstained sections were used for immunohistochemistry with antibodies anti-ER (clone EP1, Dako Omnis), anti-PR (clone PgR 1294, Dako Omnis), c-erbB2 (clone A0485, Dako Omnis) and Ki-67 (clone MIB-1, Dako Omnis). All immunoreactions were carried out on a Dako Omnis platform (Dako, Glostrup, Denmark). All cases were seen by at least one pathologist (M.L. and/or C.R.) expert in breast pathology, who also revised all the discrepant cases prior to the final diagnosis. ER and PR were defined positive when ≥ 1% of the tumour cell nuclei showed immunostaining, according to the 2010 ASCO/CAP guidelines . ER was further stratified into LP and positive using the 10% cut-off, according to the 2020 ASCO/CAP guidelines update . Ki-67 was scored ‘high’ when ≥ 20% of the tumour nuclei were positive, taking into account the cut-off clinically used to define the Luminal B class according to the 2013 St Gallen International Breast Cancer Conference experts Panel opinion . Cells positive for Ki-67 were scored over 100 cells in both ‘cold’ and ‘hot’ tumour areas, and the final value represented an average between those of the different areas. This method is similar to the one recommended by the International Ki-67 in Breast Cancer Working Group (IKWG) in their 2021 updated recommendations , with the only difference that the recommended online scoring app was not used. c-erbB2 was scored according to the ASCO/CAP 2013 guidelines and 2018 Focused Update as 0, 1 +, 2 + or 3 + depending on intensity and completeness of the membrane staining. FISH was performed in all the equivocal cases, but the results are not reported in this paper since we focus on the immunohistochemical evaluation alone. Cohen’s kappa (κ) was used to measure the interobserver agreement between biopsy and surgical specimen; weighted κ was used when concordance between more than one result was evaluated to account for close matches. κ values < 0.20 were interpreted as poor agreement, 0.21–0.40 fair agreement, 0.41–0.60 moderate agreement, 0.61–0.80 good agreement, 0.81–0.99 very good agreement and 1 perfect agreement. P-values were calculated with Fisher’s exact test and chi-squared test, when appropriated, on GraphPad Prism 5, and p -values < 0.05 were deemed significant.
We reviewed the records of 1654 breast cancer patients who underwent both biopsy and surgical excision at the Department of Breast Surgery, San Matteo Hospital, between January 2014 and December 2020. We excluded patients for which ER, PR, c-erBb2 or Ki-67 values were missing, patients with in situ or microinvasive-only breast cancer, patients with multifocal tumours, metastatic disease and patients who underwent neoadjuvant chemotherapy. Our final cohort comprised 923 patients and their clinical characteristics are summarized in Table . The study was conducted according to the guidelines of the Declaration of Helsinki.
Samples were fixed in 10% neutral buffered formalin and embedded in paraffin before histopathological evaluation. 4-to-5 µm-thick sections were cut and stained with hematoxylin and eosin (HE), and unstained sections were used for immunohistochemistry with antibodies anti-ER (clone EP1, Dako Omnis), anti-PR (clone PgR 1294, Dako Omnis), c-erbB2 (clone A0485, Dako Omnis) and Ki-67 (clone MIB-1, Dako Omnis). All immunoreactions were carried out on a Dako Omnis platform (Dako, Glostrup, Denmark). All cases were seen by at least one pathologist (M.L. and/or C.R.) expert in breast pathology, who also revised all the discrepant cases prior to the final diagnosis. ER and PR were defined positive when ≥ 1% of the tumour cell nuclei showed immunostaining, according to the 2010 ASCO/CAP guidelines . ER was further stratified into LP and positive using the 10% cut-off, according to the 2020 ASCO/CAP guidelines update . Ki-67 was scored ‘high’ when ≥ 20% of the tumour nuclei were positive, taking into account the cut-off clinically used to define the Luminal B class according to the 2013 St Gallen International Breast Cancer Conference experts Panel opinion . Cells positive for Ki-67 were scored over 100 cells in both ‘cold’ and ‘hot’ tumour areas, and the final value represented an average between those of the different areas. This method is similar to the one recommended by the International Ki-67 in Breast Cancer Working Group (IKWG) in their 2021 updated recommendations , with the only difference that the recommended online scoring app was not used. c-erbB2 was scored according to the ASCO/CAP 2013 guidelines and 2018 Focused Update as 0, 1 +, 2 + or 3 + depending on intensity and completeness of the membrane staining. FISH was performed in all the equivocal cases, but the results are not reported in this paper since we focus on the immunohistochemical evaluation alone. Cohen’s kappa (κ) was used to measure the interobserver agreement between biopsy and surgical specimen; weighted κ was used when concordance between more than one result was evaluated to account for close matches. κ values < 0.20 were interpreted as poor agreement, 0.21–0.40 fair agreement, 0.41–0.60 moderate agreement, 0.61–0.80 good agreement, 0.81–0.99 very good agreement and 1 perfect agreement. P-values were calculated with Fisher’s exact test and chi-squared test, when appropriated, on GraphPad Prism 5, and p -values < 0.05 were deemed significant.
ER ER was positive in 829/923 (89%) biopsies and in 831/923 (90%) surgical resection specimens, with a concordance of 97.83% ( n = 903/923), a Cohen’s κ of 0.880 (very good agreement) and a p -value < 0.0001. The concordant and discrepant results are reported in Table , with 9 positive results (9/923, 1%) later reported as negative on the surgical specimen and 11 negative results (11/923, 1.2%) that were upgraded to positive on the final pathology report. These changes would have had a clinical impact with either withholding or administration of endocrine therapy in these patients, if the decision was based only on the biopsy results. When the positive results were further stratified into ER-LP and ER-positive (Table ) the overall agreement was still very good (97.18%, weighted κ = 0.924, p value < 0.0001), but the concordance for the ER-LP category itself was low ( n = 11/23, 47.8%). Of the 12 discordant ER-LP results on biopsy, 8/12 (66.7%) were ER-negative on the surgical specimen and 4/12 (33.3%) were ER-positive. Of these four cases, three cases were only slightly above the cut-off for ER-LP (15%), whilst one showed positivity in 60% of the cells (Fig. ). PR PR was positive in 759/923 (82%) biopsies and in 778/923 (84%) surgical resection specimens. Concordance was 94.26% ( n = 870/923), Cohen’s κ was 0.794 (good agreement) and p value was < 0.0001. Most ( n = 36/53, 67.9%) of the discordant results were biopsies reported as PR negative that were later reported as PR positive on the surgical specimen. The complete results are summarized in Table . c-erbB2/HER2 c-erbB2 was found to be concordant in 68% of cases ( n = 631/923), and Table details the results for each category. Cohen’s weighted κ was 0.675 (good agreement) and p -value was < 0.0001. Breaking down the results for each category, 1 + was the least concordant group (37% vs 83%, 79% and 97% for 0, 2 + and 3+ , respectively), with 72% ( n = 136/188) of the discordant results being diagnosed as 0 on the surgical specimen. According to current clinical practice, only four (4/923, 0.4%) patients had changes in the diagnosis that could significantly impact their treatment choice (two 1 + biopsies later upgraded to 3 and two 3 + biopsies downgraded to 1 +). They were all, except for one locally advanced cancer, early breast cancers that would have not been candidates for neoadjuvant treatment. However, on the account of the results of the biopsy alone, they would have been denied a potentially life-saving treatment or subjected to the toxicities of an ineffective one. ki-67 Ki-67 was high in 256/923 (28%) biopsies and in 357/923 (38%) surgical resection specimens. Concordance was 86.13% ( n = 759/923), Cohen’s κ was 0.686 (good agreement), and p value was < 0.0001. Concordance and discordance are reported in Table ; most of the discordant results ( n = 107/128, 84%) were biopsies in which Ki-67 was reported as low that were later upgraded on the excision specimen.
ER was positive in 829/923 (89%) biopsies and in 831/923 (90%) surgical resection specimens, with a concordance of 97.83% ( n = 903/923), a Cohen’s κ of 0.880 (very good agreement) and a p -value < 0.0001. The concordant and discrepant results are reported in Table , with 9 positive results (9/923, 1%) later reported as negative on the surgical specimen and 11 negative results (11/923, 1.2%) that were upgraded to positive on the final pathology report. These changes would have had a clinical impact with either withholding or administration of endocrine therapy in these patients, if the decision was based only on the biopsy results. When the positive results were further stratified into ER-LP and ER-positive (Table ) the overall agreement was still very good (97.18%, weighted κ = 0.924, p value < 0.0001), but the concordance for the ER-LP category itself was low ( n = 11/23, 47.8%). Of the 12 discordant ER-LP results on biopsy, 8/12 (66.7%) were ER-negative on the surgical specimen and 4/12 (33.3%) were ER-positive. Of these four cases, three cases were only slightly above the cut-off for ER-LP (15%), whilst one showed positivity in 60% of the cells (Fig. ).
PR was positive in 759/923 (82%) biopsies and in 778/923 (84%) surgical resection specimens. Concordance was 94.26% ( n = 870/923), Cohen’s κ was 0.794 (good agreement) and p value was < 0.0001. Most ( n = 36/53, 67.9%) of the discordant results were biopsies reported as PR negative that were later reported as PR positive on the surgical specimen. The complete results are summarized in Table .
c-erbB2 was found to be concordant in 68% of cases ( n = 631/923), and Table details the results for each category. Cohen’s weighted κ was 0.675 (good agreement) and p -value was < 0.0001. Breaking down the results for each category, 1 + was the least concordant group (37% vs 83%, 79% and 97% for 0, 2 + and 3+ , respectively), with 72% ( n = 136/188) of the discordant results being diagnosed as 0 on the surgical specimen. According to current clinical practice, only four (4/923, 0.4%) patients had changes in the diagnosis that could significantly impact their treatment choice (two 1 + biopsies later upgraded to 3 and two 3 + biopsies downgraded to 1 +). They were all, except for one locally advanced cancer, early breast cancers that would have not been candidates for neoadjuvant treatment. However, on the account of the results of the biopsy alone, they would have been denied a potentially life-saving treatment or subjected to the toxicities of an ineffective one.
Ki-67 was high in 256/923 (28%) biopsies and in 357/923 (38%) surgical resection specimens. Concordance was 86.13% ( n = 759/923), Cohen’s κ was 0.686 (good agreement), and p value was < 0.0001. Concordance and discordance are reported in Table ; most of the discordant results ( n = 107/128, 84%) were biopsies in which Ki-67 was reported as low that were later upgraded on the excision specimen.
ER In the published literature ER concordance between biopsy and surgical excision averages at 93.3% (range 78.7–99.1%, Table ). Our concordance was 97.83%, slightly higher than the average value but in keeping with the known results. This variance can also be attributed to the different cut-offs and scoring methods used in the single laboratories, reported in the table that impair their universal reproducibility. Several factors are recognized to influence concordance between biopsy and surgical specimen, with the pre-analytical phase being the most important to standardize intra-laboratory results, trying to avoid under-fixation . The choice of the antibody clone and the immunostaining platform has also been demonstrated to be relevant in the over- or underestimation of the ER percentage , especially when comparing the widely used Dako and Ventana clones. The high percentage (66.7% of ER-LP cases) of ER-LP that tested ER-negative on the surgical specimen suggests that ER-LP results more frequently represent an “overcalling” of a non-luminal carcinoma, in line with the current understanding of ER-LP tumour biology, that relates them more closely to this group. Recent published data suggest also that artefactual reduced intensity of staining of ER in normal breast tissue adjacent to the neoplasia may concur to report these cases as ER-LP , therefore underlying the importance of the presence of internal and/or on-slide controls in the assessment of these borderline cases. In summary, our results suggest that caution should be taken when calling ER-LP on biopsy and for subsequent decision-making, and further studies are needed to define if this category can be safely defined on a biopsy specimen. PR Our final concordance for PR was 94.26%. From a review of the literature, the average reported concordance is 87.3% (range 73.5–95%, Table ), and our results fall on the upper side of this range. The lower concordance of PR assessment on biopsy and surgical specimen reflects its naturally occurring dishomogeneity in breast normal tissue and tumours, owing to his nature as a down-stream ER effector and therefore requiring an intact ER pathway to be strongly expressed. Our results with a higher proportion of upgrades rather than downgrades on the surgical specimen (67.9% vs 32.1%) suggest that indeed tumour heterogeneity, with a negative spot sampled on biopsy, may be a significant issue in PR assessment. Given this heterogeneity, PR concordance is especially sensitive to sampling artefacts, especially undersampling of the target lesion. From the current literature, four represents the minimum number of biopsy cores that should be retrieved to ensure a correct preoperative evaluation . c-erbB2/HER2 In the published literature, concordance for c-erbB2 when evaluated with immunohistochemistry alone averages at 85.4% (range 56–98.8%, Table ); our series demonstrates a concordance in 68% of cases, lower than the reported values, but still in the published range. Breaking down the results, the 1 + category was the one with the lowest concordance, with only 37% of results getting confirmation on the surgical specimens. Whilst we had previously reported that this discordance for the 1 + category was not likely to have a significant therapeutic impact for the patients , the recent introduction of the HER2-low category as a subset of patients that could benefit from the administration of targeted anti-HER2 therapy will radically change that in the upcoming years. The 2013 ASCO/CAP guidelines define 1 + as incomplete membrane staining that is faint or barely perceptible and within > 10% of tumour cells and readily appreciated in a contiguous population using a low-power objective, whereas 0 is defined as absence of staining or incomplete membrane staining that is faint/barely perceptible and within ≤ 10% of tumour cells . This diagnosis must be made on immunohistochemistry alone, with no help coming from ancillary molecular studies. However, in everyday practice, this distinction is not an easy one to make especially on biopsy, where crushing or technical artefacts and tumour heterogeneity make it a particularly challenging task. In the light of the growing importance of distinguishing between 0 and 1 +, we stress the importance of a dedicated, up-to-date breast pathologist examining these specimens, especially in those difficult cases in which the diagnosis is not immediately clear. Ki-67 Ki-67 represents an unfulfilled promise in the field of breast cancer; despite being widely used as a marker of proliferation, it has failed time and time again to reach prognostic and predictive significance. This is at least in part ascribable to the still unclear nature of this biomarker that, despite the best efforts that has been put into it, still does not have a fixed biological cut-off; guidelines for what qualifies as ‘high’ and ‘low’ Ki-67, and whether these definitions truly represent different biological entities, are still unclear. Recent recommendations from the IKWG report that sufficient levels of evidence of the prognostic value of Ki-67 exists only in the setting of ER-positive early-stage breast cancer, where levels ≥ 5% and ≥ 30%, respectively, may favour withholding or administration of chemotherapy. Historically, the 2013 St Gallen International Breast Cancer Conference suggested a 20% cut-off for the definition of ‘high’ Ki-67 in the definition of the surrogate intrinsic subtypes of breast cancer , but still advised that different cut-offs could be adopted by single laboratories. As reported in literature , most laboratories use either a 14% or a 20% cut-off. Our concordance was in line with the values reported in literature (Table ), especially with those reported by You et al. , whose cut-off and population are similar to those of this work. Similar values are reported also with a 14% cut-off by Kim et al., , Ahn et al., and Meattini et al. , whilst lower concordance levels are reported by studies in which a Ventana antibody is used for immunohistochemistry irrespectively of the cut-off used. To our knowledge, this work represents one of the largest single-centre series present in literature and the first one where issues within the ER-LP and HER2-low categories were specifically addressed. Moreover, all the immunohistochemistry reactions were carried out with the same antibodies on the same platform, and a breast pathologist (M.L. and/or C.R.) was involved in the diagnosis of all cases, ensuring a high degree of homogeneity in the test results. Possible limitations include the retrospective nature of the study and the lack of data regarding the HER2 2 + amplification status; however, this study aimed to evaluate only the immunohistochemical concordance for HER2, with particular attention to the HER2-low categories. We confirmed a very good agreement for ER assessment on biopsy, and a still satisfactory, albeit lower, concordance for PR that falls just short of the very good agreement cut-off. Ki-67 evaluation instead was confirmed to be slightly less reliable than ER and PR, with a significantly lower concordance, and could warrant evaluation also on the surgical specimen, if available, in the light of potential treatment options . ER-LP analysis revealed that, even if the global ER concordance is still satisfactory, the concordance for the specific category is still low and must be further investigated to define the boundaries within which it can be safely assessed on a biopsy sample. The low concordance for the 1 + c-erbB2 immunohistochemistry is particularly relevant in the context of the new therapeutic advances involving it and highlights the technical difficulties in consistently implementing the current diagnostic criteria available for the diagnosis. In the light of the quick and exciting evolution of the therapeutic landscape of breast cancer, it is important that caution is taken in evaluating biopsy samples, especially for those predictive factors that may significantly impact the therapeutic options of the patient. Although only a very small number of patients in our series would have received an inappropriate treatment based on biopsy results alone, our data underline the relevance of surgical sample retesting, at least in selected cases, including large tumours and cases with discrepant histological characteristics. Specific training needs should be addressed by the national and international pathology societies, especially in those still grey areas of ER-LP and HER2-low categories, where the difference between a negative and a positive result may withhold a target therapy, or cause patients to undergo unnecessary, and often quite burdensome, aggressive therapy.
In the published literature ER concordance between biopsy and surgical excision averages at 93.3% (range 78.7–99.1%, Table ). Our concordance was 97.83%, slightly higher than the average value but in keeping with the known results. This variance can also be attributed to the different cut-offs and scoring methods used in the single laboratories, reported in the table that impair their universal reproducibility. Several factors are recognized to influence concordance between biopsy and surgical specimen, with the pre-analytical phase being the most important to standardize intra-laboratory results, trying to avoid under-fixation . The choice of the antibody clone and the immunostaining platform has also been demonstrated to be relevant in the over- or underestimation of the ER percentage , especially when comparing the widely used Dako and Ventana clones. The high percentage (66.7% of ER-LP cases) of ER-LP that tested ER-negative on the surgical specimen suggests that ER-LP results more frequently represent an “overcalling” of a non-luminal carcinoma, in line with the current understanding of ER-LP tumour biology, that relates them more closely to this group. Recent published data suggest also that artefactual reduced intensity of staining of ER in normal breast tissue adjacent to the neoplasia may concur to report these cases as ER-LP , therefore underlying the importance of the presence of internal and/or on-slide controls in the assessment of these borderline cases. In summary, our results suggest that caution should be taken when calling ER-LP on biopsy and for subsequent decision-making, and further studies are needed to define if this category can be safely defined on a biopsy specimen.
Our final concordance for PR was 94.26%. From a review of the literature, the average reported concordance is 87.3% (range 73.5–95%, Table ), and our results fall on the upper side of this range. The lower concordance of PR assessment on biopsy and surgical specimen reflects its naturally occurring dishomogeneity in breast normal tissue and tumours, owing to his nature as a down-stream ER effector and therefore requiring an intact ER pathway to be strongly expressed. Our results with a higher proportion of upgrades rather than downgrades on the surgical specimen (67.9% vs 32.1%) suggest that indeed tumour heterogeneity, with a negative spot sampled on biopsy, may be a significant issue in PR assessment. Given this heterogeneity, PR concordance is especially sensitive to sampling artefacts, especially undersampling of the target lesion. From the current literature, four represents the minimum number of biopsy cores that should be retrieved to ensure a correct preoperative evaluation .
In the published literature, concordance for c-erbB2 when evaluated with immunohistochemistry alone averages at 85.4% (range 56–98.8%, Table ); our series demonstrates a concordance in 68% of cases, lower than the reported values, but still in the published range. Breaking down the results, the 1 + category was the one with the lowest concordance, with only 37% of results getting confirmation on the surgical specimens. Whilst we had previously reported that this discordance for the 1 + category was not likely to have a significant therapeutic impact for the patients , the recent introduction of the HER2-low category as a subset of patients that could benefit from the administration of targeted anti-HER2 therapy will radically change that in the upcoming years. The 2013 ASCO/CAP guidelines define 1 + as incomplete membrane staining that is faint or barely perceptible and within > 10% of tumour cells and readily appreciated in a contiguous population using a low-power objective, whereas 0 is defined as absence of staining or incomplete membrane staining that is faint/barely perceptible and within ≤ 10% of tumour cells . This diagnosis must be made on immunohistochemistry alone, with no help coming from ancillary molecular studies. However, in everyday practice, this distinction is not an easy one to make especially on biopsy, where crushing or technical artefacts and tumour heterogeneity make it a particularly challenging task. In the light of the growing importance of distinguishing between 0 and 1 +, we stress the importance of a dedicated, up-to-date breast pathologist examining these specimens, especially in those difficult cases in which the diagnosis is not immediately clear.
Ki-67 represents an unfulfilled promise in the field of breast cancer; despite being widely used as a marker of proliferation, it has failed time and time again to reach prognostic and predictive significance. This is at least in part ascribable to the still unclear nature of this biomarker that, despite the best efforts that has been put into it, still does not have a fixed biological cut-off; guidelines for what qualifies as ‘high’ and ‘low’ Ki-67, and whether these definitions truly represent different biological entities, are still unclear. Recent recommendations from the IKWG report that sufficient levels of evidence of the prognostic value of Ki-67 exists only in the setting of ER-positive early-stage breast cancer, where levels ≥ 5% and ≥ 30%, respectively, may favour withholding or administration of chemotherapy. Historically, the 2013 St Gallen International Breast Cancer Conference suggested a 20% cut-off for the definition of ‘high’ Ki-67 in the definition of the surrogate intrinsic subtypes of breast cancer , but still advised that different cut-offs could be adopted by single laboratories. As reported in literature , most laboratories use either a 14% or a 20% cut-off. Our concordance was in line with the values reported in literature (Table ), especially with those reported by You et al. , whose cut-off and population are similar to those of this work. Similar values are reported also with a 14% cut-off by Kim et al., , Ahn et al., and Meattini et al. , whilst lower concordance levels are reported by studies in which a Ventana antibody is used for immunohistochemistry irrespectively of the cut-off used. To our knowledge, this work represents one of the largest single-centre series present in literature and the first one where issues within the ER-LP and HER2-low categories were specifically addressed. Moreover, all the immunohistochemistry reactions were carried out with the same antibodies on the same platform, and a breast pathologist (M.L. and/or C.R.) was involved in the diagnosis of all cases, ensuring a high degree of homogeneity in the test results. Possible limitations include the retrospective nature of the study and the lack of data regarding the HER2 2 + amplification status; however, this study aimed to evaluate only the immunohistochemical concordance for HER2, with particular attention to the HER2-low categories. We confirmed a very good agreement for ER assessment on biopsy, and a still satisfactory, albeit lower, concordance for PR that falls just short of the very good agreement cut-off. Ki-67 evaluation instead was confirmed to be slightly less reliable than ER and PR, with a significantly lower concordance, and could warrant evaluation also on the surgical specimen, if available, in the light of potential treatment options . ER-LP analysis revealed that, even if the global ER concordance is still satisfactory, the concordance for the specific category is still low and must be further investigated to define the boundaries within which it can be safely assessed on a biopsy sample. The low concordance for the 1 + c-erbB2 immunohistochemistry is particularly relevant in the context of the new therapeutic advances involving it and highlights the technical difficulties in consistently implementing the current diagnostic criteria available for the diagnosis. In the light of the quick and exciting evolution of the therapeutic landscape of breast cancer, it is important that caution is taken in evaluating biopsy samples, especially for those predictive factors that may significantly impact the therapeutic options of the patient. Although only a very small number of patients in our series would have received an inappropriate treatment based on biopsy results alone, our data underline the relevance of surgical sample retesting, at least in selected cases, including large tumours and cases with discrepant histological characteristics. Specific training needs should be addressed by the national and international pathology societies, especially in those still grey areas of ER-LP and HER2-low categories, where the difference between a negative and a positive result may withhold a target therapy, or cause patients to undergo unnecessary, and often quite burdensome, aggressive therapy.
|
Aetiology of long bone chronic osteomyelitis: an analysis of the current situation in one region in Egypt | cf89960a-bd06-4372-bde6-0133b0cce164 | 10036446 | Debridement[mh] | Osteomyelitis is characterized by low-grade inflammation caused by persistent pathogenic microorganisms with bone destruction and necrosis. It is still one of the most challenging conditions due to its long course, complex treatment, and risk of recurrence . Chronic osteomyelitis is intensely challenging to treat due to the poor blood supply, the devitalized tissues, the poor antibiotic penetration and the soft tissue envelope in some regions . Biofilm-producing organisms adhere resiliently to the surface of the bone and implanted material, which makes them resistant to host defences and antibiotics . A definitive diagnosis of chronic osteomyelitis depends on identifying causative organisms through microbiological methods (culture and sensitivity) and histopathological examination of tissues . Gram-positive organisms (specifically Staph aureus) remained the most commonly reported organisms to cause chronic osteomyelitis in long bones. However, other reports revealed paradoxical results with prevailing Gram-negative bacilli (GNB). Gram-negative osteomyelitis is hard to control and has increased multidrug resistance (MDR) and increased rates of recurrence [ – ]. This study aimed to analyse and discuss the microbiological results of intraoperatively collected samples of patients diagnosed with chronic osteomyelitis in one centre. With very few available published data from low resources regions , we planned to investigate the current situation to figure out the best way to control this debilitating disease and update our infection control measures.
Study design This prospective clinical study was done on a consecutive group of 33 patients who underwent surgical debridement for long bones chronic osteomyelitis in a single centre (Benha University Hospitals). The patients had their initial diagnosis and previous surgeries elsewhere in the region (Qalyubia governorate) and were referred to our centre later due to persistent infection. All patients had an established diagnosis of COM (clinical and radiological) in the appendicular system. Acute osteomyelitis or septic arthritis, periprosthetic infection, diabetic foot infection, or vertebral osteomyelitis were excluded from the study. Preoperative assessment All patients included in the study were evaluated clinically, laboratory, and radiologically. Laboratory investigation included a complete blood picture, erythrocyte sedimentation rate, C-reactive protein, and other relevant preoperative investigations. A detailed history of previous surgeries and the metalwork used was also obtained. Radiologically, anteroposterior and lateral plain X-rays were obtained to evaluate bone quality, sequestrated bone, sclerosis, the union of fractures, and loosening around metalwork. 18F-Fluorodeoxyglucose (FDG)-positron emission tomography (FDG PET-CT) scan was done on 70% of patients, aiming to provide a three-dimensional analysis of the infected area, helping the preoperative planning and determine the extent of resection during surgical debridement. Surgical debridement and samples collection Surgical index debridement was done for all cases by the first author (AE) between February 2019 and February 2022. The extent of debridement was planned preoperatively. Antibiotic therapy was stopped 14 days before the index debridement surgery, and no routine antibiotic prophylaxis was given until bone/tissue biopsies were collected. Three to six deep bone/tissue samples were collected in leak-proof sterile containers using different instruments and immediately transferred to the Microbiology department . Microbiological examination Standard bacteriological techniques were applied to isolate and identify the causative organisms. Specimens were cultured on primary culture media (nutrient, blood, and MacConkey agars), incubated at 37 °C for 24–48 h. Liquid samples were inoculated in blood culture broths incubated in Bact/alert system. Positive vials were cultured on the previous culture media. Colonial growth on the agar surface was inspected for colonial morphology, haemolysis on blood agar, swarming, or any other characteristic odour or pigment. Where mixed bacterial growth was observed, subculture was done. No molecular typing or PCR was performed. Antimicrobial sensitivity testing used an automated bacterial identification system (Vitek 2 and Phoenix) that analyses MIC patterns and detects phenotypes for most tested organisms. Smear from the bacterial growth was prepared and fixed with one or two drops of methanol, stained with Gram stain and examined microscopically. Histopathological examination One bone/soft tissue sample was sent for histopathological examination; the sample was collected and transferred immediately with no preservative fluid. The histopathologists were not aware of the microbiological results of the patients. The characteristic histopathological findings of COM are the persistence of dilated blood vessels, granulation tissue, inflammatory cell infiltration, devitalized bone, reactive new bone formation, and suppuration. A definitive diagnosis of COM was made based on deep tissue surgical cultures and histopathological analysis. Analysis of the results and statistical methods Data management and statistical analysis were done using SPSS version 28 (IBM, Armonk, New York, USA). Quantitative data were assessed for normality using the Shapiro–Wilk test and direct data visualization methods. According to normality testing, numerical data were summarized as means and standard deviations or medians and ranges. Categorical data were summarized as numbers and percentages. The microbiology, histopathology, and culture results were plotted together to conclude and investigate the causative organisms and antibiotic culture and sensitivity.
This prospective clinical study was done on a consecutive group of 33 patients who underwent surgical debridement for long bones chronic osteomyelitis in a single centre (Benha University Hospitals). The patients had their initial diagnosis and previous surgeries elsewhere in the region (Qalyubia governorate) and were referred to our centre later due to persistent infection. All patients had an established diagnosis of COM (clinical and radiological) in the appendicular system. Acute osteomyelitis or septic arthritis, periprosthetic infection, diabetic foot infection, or vertebral osteomyelitis were excluded from the study.
All patients included in the study were evaluated clinically, laboratory, and radiologically. Laboratory investigation included a complete blood picture, erythrocyte sedimentation rate, C-reactive protein, and other relevant preoperative investigations. A detailed history of previous surgeries and the metalwork used was also obtained. Radiologically, anteroposterior and lateral plain X-rays were obtained to evaluate bone quality, sequestrated bone, sclerosis, the union of fractures, and loosening around metalwork. 18F-Fluorodeoxyglucose (FDG)-positron emission tomography (FDG PET-CT) scan was done on 70% of patients, aiming to provide a three-dimensional analysis of the infected area, helping the preoperative planning and determine the extent of resection during surgical debridement.
Surgical index debridement was done for all cases by the first author (AE) between February 2019 and February 2022. The extent of debridement was planned preoperatively. Antibiotic therapy was stopped 14 days before the index debridement surgery, and no routine antibiotic prophylaxis was given until bone/tissue biopsies were collected. Three to six deep bone/tissue samples were collected in leak-proof sterile containers using different instruments and immediately transferred to the Microbiology department .
Standard bacteriological techniques were applied to isolate and identify the causative organisms. Specimens were cultured on primary culture media (nutrient, blood, and MacConkey agars), incubated at 37 °C for 24–48 h. Liquid samples were inoculated in blood culture broths incubated in Bact/alert system. Positive vials were cultured on the previous culture media. Colonial growth on the agar surface was inspected for colonial morphology, haemolysis on blood agar, swarming, or any other characteristic odour or pigment. Where mixed bacterial growth was observed, subculture was done. No molecular typing or PCR was performed. Antimicrobial sensitivity testing used an automated bacterial identification system (Vitek 2 and Phoenix) that analyses MIC patterns and detects phenotypes for most tested organisms. Smear from the bacterial growth was prepared and fixed with one or two drops of methanol, stained with Gram stain and examined microscopically.
One bone/soft tissue sample was sent for histopathological examination; the sample was collected and transferred immediately with no preservative fluid. The histopathologists were not aware of the microbiological results of the patients. The characteristic histopathological findings of COM are the persistence of dilated blood vessels, granulation tissue, inflammatory cell infiltration, devitalized bone, reactive new bone formation, and suppuration. A definitive diagnosis of COM was made based on deep tissue surgical cultures and histopathological analysis.
Data management and statistical analysis were done using SPSS version 28 (IBM, Armonk, New York, USA). Quantitative data were assessed for normality using the Shapiro–Wilk test and direct data visualization methods. According to normality testing, numerical data were summarized as means and standard deviations or medians and ranges. Categorical data were summarized as numbers and percentages. The microbiology, histopathology, and culture results were plotted together to conclude and investigate the causative organisms and antibiotic culture and sensitivity.
This study was conducted on 33 patients (30 males and three females) suffering from chronic osteomyelitis (COM). There were 18 smokers, one EX smoker, and 14 nonsmokers. The mean age at the index debridement surgery was 37.1 years (range 7–73). Patient characteristics are detailed in Table . Laboratory and culture findings The mean WBCs were 7.1 ± 1.1. The mean ESR was 48.8 ± 12.2. The mean CRP was 22.3 ± 9.7. The mean number of samples collected for microbiology was 5 ± 1. Growth was reported in 28 patients (84.8%). GNB were the most common causative organisms in 51.5% of the cohort, slightly higher than Gram-positive (48.5%). Looking at the spectrum of organisms, cultures grew isolated GNB in 36.4%, gram-positive in 33.3%, mixed in 15.2%, and no growth in 15.2%. Monomicrobial infection was found in 18 patients (54.5%), while polymicrobial infection (different identified organisms) was evident in a third of the cohort (ten patients, 30.3%) (Table ). Organism identity and sensitivity MRSA was the most frequently identified single organism to grow in all cultures (isolated or mixed) (36.4%). Klebsiella was the most frequent single GNB (15.2%); However, there were some results of Gram-negative cultures (12.1%) that could not identify the organism specifically (Table ). MDR Klebsiella ( n = 5) was only 20–40% sensitive to gentamycin, amikacin, ciprofloxacin, and trimethoprim–sulphamethoxazole. Pseudomonas ( n = 3) was 100% sensitive towards piperacillin/tazobactam, gentamycin, amikacin, ciprofloxacin, and levofloxacin and 66% sensitive to cefipime and imipenem. E. Coli ( n = 3) was sensitive to piperacillin–tazobactam, ciprofloxacin, and levofloxacin in 66.7% and sensitive to amoxicillin–clavulanic acid and trimethoprim–sulphamethoxazole in only 33.3%. Proteus ( n = 3) was 67.3% sensitive to piperacillin/tazobactam, meropenem, ceftazidime, and amikacin.
The mean WBCs were 7.1 ± 1.1. The mean ESR was 48.8 ± 12.2. The mean CRP was 22.3 ± 9.7. The mean number of samples collected for microbiology was 5 ± 1. Growth was reported in 28 patients (84.8%). GNB were the most common causative organisms in 51.5% of the cohort, slightly higher than Gram-positive (48.5%). Looking at the spectrum of organisms, cultures grew isolated GNB in 36.4%, gram-positive in 33.3%, mixed in 15.2%, and no growth in 15.2%. Monomicrobial infection was found in 18 patients (54.5%), while polymicrobial infection (different identified organisms) was evident in a third of the cohort (ten patients, 30.3%) (Table ).
MRSA was the most frequently identified single organism to grow in all cultures (isolated or mixed) (36.4%). Klebsiella was the most frequent single GNB (15.2%); However, there were some results of Gram-negative cultures (12.1%) that could not identify the organism specifically (Table ). MDR Klebsiella ( n = 5) was only 20–40% sensitive to gentamycin, amikacin, ciprofloxacin, and trimethoprim–sulphamethoxazole. Pseudomonas ( n = 3) was 100% sensitive towards piperacillin/tazobactam, gentamycin, amikacin, ciprofloxacin, and levofloxacin and 66% sensitive to cefipime and imipenem. E. Coli ( n = 3) was sensitive to piperacillin–tazobactam, ciprofloxacin, and levofloxacin in 66.7% and sensitive to amoxicillin–clavulanic acid and trimethoprim–sulphamethoxazole in only 33.3%. Proteus ( n = 3) was 67.3% sensitive to piperacillin/tazobactam, meropenem, ceftazidime, and amikacin.
In this study, intra-operative deep tissue and bone specimens accurately identified causative bacteria in 84.8% of patients. All cases underwent index debridement and sample collection by the same surgeon, with three to six collected specimens. There was an increased incidence of G-negative osteomyelitis in 51.5% of the whole cohort, either as isolated GNB (36.4%) or mixed gram-positive and negative (15.2%). Chronic osteomyelitis is a challenge for orthopaedic surgeons. Appropriate diagnosis and treatment of chronic osteomyelitis require microbiologic cultures of the infected bone and proper antibiotics according to culture and sensitivity. Bone provides a unique harbour for microorganisms that produce biofilms, allowing them to attach resiliently to biologic and implanted surfaces, remaining insusceptible to host defences and antibiotics . Several studies analysed the bacteriological profile in long bone osteomyelitis [ , – , – ]. Reports from low resources settings and developing countries are sparse [ , , , , ]. Nevertheless, other studies investigated the aetiology and bacteriology of COM in a diabetic foot or joint replacement infection or mixed group of patients [ – ]. We did not reflect on these studies due to the particular pathology in each situation which is different from implant-related long bone osteomyelitis. Sheehy et al. and Dudareva et al. reported 66% and 67.7% culture-positive results, respectively, compared to 84.4% in this study . Sheehy et al. reported a range of samples from 0 to 10, with no samples analysed in ten patients out of 166 (6%). Marais et al. . recorded higher culture-positive results (90.9%); however, this result represents only 11 of 26 patients who underwent surgical debridement and intra-operative bone biopsies. The lack of growth in the collected samples of any study has some explanations. Antibiotics should be stopped for 14 days in clinically stable patients before surgery. Samples should be collected from the deep bone and soft tissues. Multiple samples should be collected with different instruments. Proper containers and a timely chain of transfer should be applied. However, we applied all these precautions in this study, with 15% of samples failing to grow organisms . Other factors are related to microbiological methods and techniques. Pretreatment of samples with specific chemicals could release the organism from the biofilm. Also, physical methods like sonication offer similar effects. Extended cultures also provide extra opportunities to grow weak organisms . This study could not apply such methods due to unavailability or logistic problems. Although sinus tract swabs are still one of the diagnostic tools of COM and may provide the accurate causative organism, tissue specimen remains the gold standard for diagnosing COM and providing accurate causative organisms. Vemu et al. compared tissue specimens and sinus tract swabs; positive culture in tissue specimens was 60.7% compared to 37.7% in sinus swabs . Contamination with nonpathogenic bacteria or other naturally occurring random organisms during the sampling process could explain this outcome. The current study showed that chronic osteomyelitis was the highest in middle age (median 37y), which agrees with other studies [ , , , ]. Road traffic accidents frequently occur among those age groups; such high-energy trauma often leads to open fractures and soft tissue trauma and potential risk factors for traumatic osteomyelitis. In the current study, the male/female ratio was 10:1, which is very similar to previous studies [ , , , , ]. The aetiology of COM in our study was post-operative in half of the patients. In other studies, haematogenous osteomyelitis or OM following open fractures were the leading cause [ , , ]. Post-operative COM was only 29.3%, 9.9%, and 2.8% of the whole cohorts, respectively [ , , ]. This finding is very critical, we had difficulty analysing this high rate, but there were some clues. Patients were operated on previously in different hospitals. Also, there is no universal MRSA screen or urine dipstick screening. We could not document the general health status or the soft tissue injury. Additionally, antibiotic and follow-up regimens were different, with a tendency to give strong antibiotics for a long time and to do daily dressing on the post-operative wounds. Finally, we do not have the number of surgeries at each hospital to get an actual incidence of surgical site infection. In the current study, the tibia (48.5%) was the most frequent site of chronic osteomyelitis. Many studies had similar findings [ , , , , , ]. Two other studies reported the femur as the most frequent infected bone ; in our report, the femur was the second affected bone (30.3%). GNB organisms were the most common cultured bacteria in 51.5% of patients (alone or mixed with Gram + ve organisms). In the literature, GNB bacteria were prevalent in a few studies. GNB incidence has evolved; a historical paper by Meyers et al. reported a 28% incidence of GNB in 1973, increasing from 15% only in 1966 . Recently, Marais et al. reported 26 patients in South Africa with COM in two groups: palliative versus curative. Their cohort had nearly equal numbers of haematogenous, post-operative and open fractures, and the tibia was the most commonly affected bone. GNB infection was recorded in 61.5% ( n = 8) compared to 38.5% ( n = 5) Gram + ve . In Egypt, two studies analysed the bacteriological profile in their centres within the context of reporting the surgical outcomes of single-stage debridement . El-Moatassem examined 16 patients with tibial COM, who grew Gram + ve bacteria (MSSA and MRSA) in 12 (75%) and Pseudomonas in one; three more did not grow any organism . Badie and Arafa reported the results of 30 cases with long bone COM; half had tibial involvement. They had 56% haematogenous, 20% post-operative infection, and the rest were open fractures. Cultures were positive in 93% of cases, and GNB grew in 32% . Both studies emphasized the surgical aspects; we could not retrieve further details on sensitivity or resistance patterns. Agaja and Ayorinde reported similar trends in their cohort of 107 patients from Nigeria, with GNB (43%) slightly higher than Gram + ve (37.4%). However, 81.3% of their patients had haematogenous osteomyelitis . Mousa et al. studied 134 patients from Iraq in four groups: haematogenous, exogenous (open fractures), post-operative, and mastoiditis. GNB (Pseudomonas species and Klebsiella species) were prevalent in the exogenous and post-operative osteomyelitis groups . Carvalho et al. published a study that included 101 patients from Brazil who presented with isolated GNB osteomyelitis; 43% COM and 32% were associated with open fractures, and the remaining were post-operative. Clinical remission was achieved in 60% of patients. Enterobacter and Acinetobacter baumannii were the commonest. Pseudomonas came third, with sensitivity ranging from 70 to 80% towards the GNB targeting antibiotics . Jorge et al. investigated 193 patients diagnosed with osteomyelitis following fractures (closed or open) in Brazil. They reported GNB in 51.8% of patients. Pseudomonas aeruginosa and A. baumannii were the most frequent . Koutserimpas et al. analysed 14 patients with MDR and extended drug resistance (XDR) in one centre in Greece over ten years; most had post-operative osteomyelitis. Acinetobacter baumannii was the most frequent, followed by Escherichia coli and Klebsiella pneumoniae . They highlighted that A. baumanii was repeatedly a leading cause of hospital-acquired infection in their region . In our cohort, Klebsiella was the commonest, followed by Pseudomonas , Proteus and Ecoli . Acinetobacter was found in one patient with no Enterobacter. In the present study, MRSA remains the most frequent single isolate at 36.4% ( n = 12), in agreement with many previous studies, which reported an incidence ranging between 20 and 48% [ , , , , , ]. However, two studies from the same team ten years apart reported very similar incidences of staphylococcus infections (31.3%–37.5%) in both cohorts, yet, there was a declining proportion of MRSA from 30.8 to 11.4% of all staph infections, respectively . Strict applied hospital infection prevention policies could have resulted in this reduction, including preoperative MRSA screening and decolonization with topical nasal mupirocin and chlorhexidine shower. In the current study, 85.7% of S. aureus strains were resistant to methicillin (MRSA). In addition, all MRSA were 100% sensitive to amikacin, ciprofloxacin, erythromycin, linezolid, vancomycin, and clindamycin. This spectrum of sensitivities draws attention to the current problem and should attract more practical steps to reduce MRSA osteomyelitis. GNB sensitivity profile also represents a challenge; GNB is more likely to be resistant to multiple drugs, e.g. Klebsiella , which is 40% sensitive only to gentamicin and amikacin and 20% sensitive to ciprofloxacin and sulphamethoxazole. In this study, we aimed to raise voices about the differences in causes and bacteriological profiles of COM in low resources settings. The management of chronic osteomyelitis in low resources countries had more obstacles compared to the developed world . However, the high-quality literature about the bacteriological spectrum and sensitivities profile was driven by developed countries . Consensus and recommendations for antibiotic management were also reported from specialized centres in high-income countries . The contradictory findings in this study would change the current practice of choice of antibiotic prophylaxis and treatment in low resources settings.
This study has some limitations due to the relatively small sample size, heterogeneity of cases and single hospital location. The lack of data about the previous surgeries in different hospitals made tracing challenging. We could not identify the causes behind the increased post-operative infection which is a major preventable source in our cohort. More work needs to be done to analyse the GNB endemic and how to fight it.
There may be a shift in our region’s aetiologies and causative organisms; closed fractures turn into COM postoperatively, several unsuccessful attempts, delayed index debridement, and more GNB organisms. Plans need to be applied to break the cycle and improve outcomes.
|
A realized facilitation cascade mediated by biological soil crusts in a sagebrush steppe community | 3f5a26d0-1ab8-4b1e-bcee-ba6bc1a3087d | 10036522 | Microbiology[mh] | Biological soil crusts cover much of the soil surface in the arid and semi-arid regions of the world and have been described an “organizing principle” – . Soil crusts are highly variable and can be composed of nitrogen-fixing cyanobacteria and cyanolichens, lichens associated with green algae, free-living green algae and diatoms, fungi, mosses, and liverworts . These diverse assemblages are of primary importance to vascular plant communities through their effects on soil water and fertility , , . Biological soil crusts (biocrusts hereafter), can increase rainwater infiltration, improve soil moisture retention, increase soil organic matter, and prevent soil erosion , , . Mechanistically, as biocrusts hydrate and desiccate, they can release large amounts of carbon and nitrogen into ecosystems . Collectively, these effects can directly facilitate plants. Biocrusts typically form a complex matrix with vascular plants at both small and large scales , . This soil/biocrust matrix forms a far more variable “playing field” for plant communities than soil without biocrusts because various plant species respond differently to one physical component or the other , . The effect of biocrusts corresponds with their composition, and they can have negative, neutral, or positive effects on vascular plants , – ; see meta-analysis and review by Havrilla et al. . This variation in effects depends on the composition of biocrusts, vascular plant functional groups, and the provenance of the vascular plants. However, in field conditions, many studies have found that the germination and survival of native plant species is higher, or not changed, in patches with biocrusts than in area without crusts . Their facilitative and ecosystem engineering roles have been of substantial interest because of the positive effects of biocrusts on soil water and nitrogen , – . Direct facilitation infers some degree of interdependence among species , , but simple pairwise direct facilitation can lead to facilitative “interaction chains” that can have complex, but not necessarily obligate, effects through communities. Interaction chains derived from linked direct pairwise interactions between species occur when “…one species directly alters the abundance of a second species, the change in the second species affects a third species, and so on” . These chains can also be characterized as “facilitation cascades” when a series of successive facilitation interactions occur among nested groups of species – . To our knowledge, all but one of currently reported facilitation cascades have trophic dimensions and have focused on intertidal marine systems (but see Losapio et al. ). As an exception, but still trophic in nature, Angelini and Silliman found that Quercus virginiana facilitated epiphytes which in turn dramatically increased invertebrate feeding guilds. Facilitation cascades or positive interaction chains between plants, are likely to occur in semi-arid systems where biocrusts act as “a basal habitat former” , analogous to Altieri et al. positive effects of cordgrass on mussels, for vascular plant species such as shrubs, which then may have positive effects on other species . Biocrusts can certainly act as an ‘organizing principle’ at multiple scales , and plant species in arid ecosystems can have strong facilitative effects on each other . If these interactions are linked, they can likely also support facilitation cascades among autotrophs. Facilitation cascades, to date, have only focused on links that include consumers or the formation of habitats for animals, but not cascades formed only among autotrophs. Understanding the potential of biocrust-based autotrophic cascades is important because facilitative cascades can amplify biodiversity , , and because biocrusts have been extensively destroyed by livestock and recreation , . Further, the consequences of this damage can be exacerbated by warming temperatures . We explored possible facilitation cascades in semi-arid shrub steppe in western Montana dominated by the shrub Artemisia tridentata Nutt. ssp. wyomingensis Beetle & Young (Wyoming big sagebrush) and bunchgrasses, and where biocrusts commonly coexist. As noted, biocrusts can have positive and negative effects, and A. tridentata and relevant bunchgrasses have been shown to have facilitative effects on other species – . Intense competition is also common among these species , , , , and thus interaction cascades may not be entirely facilitative, but instead may be “realized”, which include nested negative and positive interactions . We integrated two-year field experiments designed to test interactions among biocrusts and two co-dominant vascular species with small-scale and large-scale measurements of spatial patterns of plants, soil water, and soil nitrogen to infer interactions among other species and at a larger spatial scale. We also conducted greenhouse experiments to measure the nitrogen-fixing capacity of the biocrusts. We organized these experiments around the following three predictions. (1) Biocrusts facilitate vascular plants. (2) Artemisia facilitates other plants and biocrusts growing under their canopies. (3) Interaction cascades initiated by biocrusts correspond with increased soil nitrogen and moisture. In total, this work identifies direct and indirect effects of biocrusts in the context of change and other key species within the region.
Site We integrated field experiments and spatial patterns in semi-arid sagebrush steppe at Calf Creek Wildlife Management Area (CCWMA), in western Montana on the east side of the Bitterroot Valley, Montana (46.2805–114.0047; 1380 m elevation). Descriptions of the integrated experiments and spatial pattern measurements are listed in Supplementary Table . CCWMA is maintained as winter range for elk ( Cervus elaphus ), and mule deer ( Odocoileus hemionus ). White-tailed deer ( Odocoileus virginianus ) are also common at the site. Cattle are fenced out of CCWMA, but we have observed occasional bovine trespassers. Mean annual precipitation at the site is approximately 30 cm. Artemisia tridentata is dominant and occurs in a patchy pattern with areas composed of bunchgrasses and forbs. We focused on the spatial patterns of the native bunchgrasses which were the most common at the site: Festuca idahoensis Elmer (Idaho fescue), Koeleria macrantha (Ledeb.) Schult. (prairie junegrass), Pseudoroegneria spicata (Pursh) Á. Löve (bluebunch wheatgrass), Hesperostipa comata (Trin. & Rupr.) (needle-and-thread grass), Leymus cinereus (Scribn. & Merr.) (Great Basin wild rye), and Poa secunda (J. Pres) (Sandberg’s bluegrass). We then focused on F. idahoensis (hereafter Festuca ) in experiments testing the effects of Artemisia and biocrusts. The large-scale patches of both Artemisia and herbaceous communities were typically 10 to > 100 m in cross section when visually surveyed during experimental setup. We also observed relatively low abundances of the exotic invaders Centaurea stoebe and Bromus tectorum . Spatial patterns Based on other studies finding that biocrust cover increased with shrub cover at large scales, but decreased with shrub cover at small scales , we measured spatial patterns at the following two scales: at the large scale of patches containing many Artemisia shrubs vs. similar sized patches without any shrubs, and at the small scale of individual Artemisia understories vs. small open areas within Artemisia patches. Because our observations suggested that Artemisia might create interstitial spaces with low grass density, we sampled at the large scale to determine whether bunchgrass species or biocrusts were associated or disassociated with Artemisia patches. Because shrub canopies might shade biocrusts, we sampled at the small scale to determine whether bunchgrass species or biocrusts within patches were associated or disassociated with Artemisia individuals. At the large scale, 30 1 × 1 m quadrats were randomly located within haphazardly chosen Artemisia and grassland patches. In each quadrat, the number of each of the bunchgrass species listed above were counted and the percent cover of Artemisia and biocrust was estimated visually to the nearest 10%. We used the “basic distance” approach with 20 m transects run through Artemisia patches to randomly choose shrubs for spatial associations and sampled vegetation at random points chosen within each of the 5 m section blocks on each transect. We located a 30 × 30 cm quadrat under the north quadrant of each of 50 Artemisia shrubs, and a 30 × 30 quadrat in the nearest open space to the north of each of these sampled Artemisia individuals. In each quadrat, we visually estimated the percent cover of biocrust to the nearest 5% and counted all individuals of the grass species listed above. The species were separated into those with or without flower stalks. Finally, to estimate the effect of crust on juvenile Artemisia , we sampled five 20 × 100 m areas for juveniles (< 20 cm tall) and recorded whether they were in the open, under adult shrubs, or within a biocrust patch. Field experiments To experimentally test the effects of biocrusts on Artemisia , we established 20 30 × 30 cm plots in areas of undisturbed biocrust cover in June 1996. In half of these plots, the biocrusts were removed and in the other half biocrusts were left intact. A one-year-old Artemisia plant was transplanted into the center of each plot. These juvenile Artemisia had been grown from locally collected seed in a blend of silica sand and soils collected from the field study site in a greenhouse before transplanting. Each plot was caged with 12-gauge wire mesh to prevent trampling and herbivory by native ungulates. These transplants were left in place for two years, but after a year, our observations suggested that growth was very slow, perhaps too slow to provide detectable comparisons of biomass. Therefore, we measured photosynthetic rates of Artemisia towards the end of the growing season (mid-September) in 1996 and again in 1997 using a LI-COR 6200 portable infrared gas analysis system. We placed entire Artemisia seedling shoots into a custom-made 1.2 L cuvette and exposed them to full sunlight without cloud interference. After two years, the aboveground biomass of Artemisia transplants was harvested, dried at 60 °C, and weighed. To experimentally test the effects of biocrusts on Festuca , we established 20 30 × 30 cm plots in areas of undisturbed biocrust cover in June 1996. In half of these plots, the biocrusts were removed and in the other half they were left intact. A one-year-old Festuca plant was transplanted into the center of each plot. These had been grown from locally collected seed in a greenhouse in a blend of silica sand and soils collected from the field study site before transplanting. Each plot was caged with a 12-gauge wire mesh to prevent trampling and herbivory by native ungulates. These transplants were in situ for two years. To test the effect of Artemisia on Festuca , we randomly selected 60 mature Artemisia shrubs and 30 were removed and 30 left undisturbed. We also selected 60 open areas with biological soil crust, at least two meters away from the nearest Artemisia for shade treatments. Half of these 60 open areas were covered with a shade screen treatment using shade cloth that blocked approximately 30% of sunlight and half were left without shade screens. In each of these 120 plots, a single one-year old Festuca plant was transplanted. In mid-September 1997, we measured Festuca survivorship, the proportion of grazed plants, and harvested and weighed aboveground biomass. Any Festuca that showed clear signs of leaf herbivory was counted as grazed. We measured soil moisture gravimetrically across all treatments in both the Artemisia -effect and soil crust-effect experiments. Soil samples were collected in all plots in mid-September 1997 when there had been no precipitation for over a week, from a depth of 5–10 cm. Each sample was weighed, dried at 100 °C for 2 days and weighed again for gravimetric soil moisture. In mid-October 2022, again roughly a week after light precipitation, we sampled soil from 5 to 10 cm depth beneath Artemisia , beneath biocrusts, and in the open. Samples were weighed, dried at 100 °C for 2 days and weighed again for gravimetric soil moisture. Greenhouse experiments We conducted a greenhouse experiment to test the effects of biocrusts on the germination and establishment of Festuca . The greenhouse was on the University of Montana campus (46.8609, − 113.98401). In June 1996, soil was collected from the field site and was added to 50 × 25 × 10 cm rectangular trays. Biocrusts removed for the field experiment were established on half of these trays, and the other half were left without biocrusts. Fifty Festuca seeds were sown in each tray, spray-misted twice per day until seeds germinated, and were subsequently watered two times per week for the duration of the experiment. Germination was monitored, and after 120 days, all surviving Festuca plants were counted again as “established”, then harvested, dried at 60 °C, and weighed. Laboratory experiments We conducted laboratory experiments to determine the cyanobacterial composition of biocrusts and to estimate the nitrogen-fixing capacity of biocrusts over a range of temperatures and water saturation. Biocrusts were collected from the field site. Belnap stated that “cyanobacteria form the matrix of biological soil crusts”, thus we cultured and isolated cyanobacterial species on 1% agar plus medium D and the dominant genera were identified using phase contrast microscopy. For nitrogen-fixation, biocrusts were maintained in enclosed growth chambers under controlled temperature and light similar to those typical in the field. The source of photosynthetically active radiation (PAR) in the 400–700 nm range was from very high output cool-white florescent lamps. Sylvania black light fluorescent lamps (peak output 365 nm, range 280–360 nm) provided a source of UVA radiation, and Westinghouse FS—40 fluorescent lamps (peak output 310 nm, range 280–360 nm) were the source of UVB radiation. Biocrust samples used for the acetylene reduction assays were 2 × 2 cm. The cuvettes used for crusts were 12 ml, injected with 1 ml acetylene (C 2 H 2 ), and incubated for 1 h. Following incubation, gas samples were withdrawn and C 2 H was quantified using a Gow-Mac 69–750 gas chromatograph. Biocrust nitrogenase activity was measured over a range of temperatures from 0 to 45 °C, and a range of biocrust water saturation from about 10% to full saturation. Statistical analyses All statistics were performed in SigmaPlot 14.5 (Systat Software, Inc. 2020) and cross-validated in R 4.2.1 . Linear regressions were used to test the relationships between vascular plant cover in large-scale (> 10 m diameter) patches for both herbaceous plants and Artemisia . For the relationship between Artemisia juveniles and biocrust, we used a Chi-square test, and for comparisons of herbivory on Festuca under Artemisia and in the open we used a Fischer’s exact test. For all other pairwise comparisons, we used t-tests when data were normally distributed and Mann–Whitney Rank Sum tests when assumptions of normality were not met—these are noted in the results. For the four concomitant treatments in the field experiment testing the effects of Artemisia on Festuca , the data failed the Shapiro–Wilk normality test, so we used a Kruskal–Wallis one way ANOVA on ranks followed by Dunn’s Method pairwise multiple comparisons. Our goal for N-fixation was simply to determine the capacity of our biocrusts to provide nitrogen to the system, therefore we grouped rates by two ranges of temperatures and two ranges of water saturation and presented these rates in box plots. For comparisons of soil water among Artemisia , biocrusts and soil without biocrusts in 2022, data were not normally distributed so we used a Kruskal–Wallis one way ANOVA on ranks followed by Tukey tests for pairwise comparisons. For comparisons of ammonium and nitrate in soils we used one way ANOVA followed by Holm-Sidak pairwise multiple comparisons.
We integrated field experiments and spatial patterns in semi-arid sagebrush steppe at Calf Creek Wildlife Management Area (CCWMA), in western Montana on the east side of the Bitterroot Valley, Montana (46.2805–114.0047; 1380 m elevation). Descriptions of the integrated experiments and spatial pattern measurements are listed in Supplementary Table . CCWMA is maintained as winter range for elk ( Cervus elaphus ), and mule deer ( Odocoileus hemionus ). White-tailed deer ( Odocoileus virginianus ) are also common at the site. Cattle are fenced out of CCWMA, but we have observed occasional bovine trespassers. Mean annual precipitation at the site is approximately 30 cm. Artemisia tridentata is dominant and occurs in a patchy pattern with areas composed of bunchgrasses and forbs. We focused on the spatial patterns of the native bunchgrasses which were the most common at the site: Festuca idahoensis Elmer (Idaho fescue), Koeleria macrantha (Ledeb.) Schult. (prairie junegrass), Pseudoroegneria spicata (Pursh) Á. Löve (bluebunch wheatgrass), Hesperostipa comata (Trin. & Rupr.) (needle-and-thread grass), Leymus cinereus (Scribn. & Merr.) (Great Basin wild rye), and Poa secunda (J. Pres) (Sandberg’s bluegrass). We then focused on F. idahoensis (hereafter Festuca ) in experiments testing the effects of Artemisia and biocrusts. The large-scale patches of both Artemisia and herbaceous communities were typically 10 to > 100 m in cross section when visually surveyed during experimental setup. We also observed relatively low abundances of the exotic invaders Centaurea stoebe and Bromus tectorum .
Based on other studies finding that biocrust cover increased with shrub cover at large scales, but decreased with shrub cover at small scales , we measured spatial patterns at the following two scales: at the large scale of patches containing many Artemisia shrubs vs. similar sized patches without any shrubs, and at the small scale of individual Artemisia understories vs. small open areas within Artemisia patches. Because our observations suggested that Artemisia might create interstitial spaces with low grass density, we sampled at the large scale to determine whether bunchgrass species or biocrusts were associated or disassociated with Artemisia patches. Because shrub canopies might shade biocrusts, we sampled at the small scale to determine whether bunchgrass species or biocrusts within patches were associated or disassociated with Artemisia individuals. At the large scale, 30 1 × 1 m quadrats were randomly located within haphazardly chosen Artemisia and grassland patches. In each quadrat, the number of each of the bunchgrass species listed above were counted and the percent cover of Artemisia and biocrust was estimated visually to the nearest 10%. We used the “basic distance” approach with 20 m transects run through Artemisia patches to randomly choose shrubs for spatial associations and sampled vegetation at random points chosen within each of the 5 m section blocks on each transect. We located a 30 × 30 cm quadrat under the north quadrant of each of 50 Artemisia shrubs, and a 30 × 30 quadrat in the nearest open space to the north of each of these sampled Artemisia individuals. In each quadrat, we visually estimated the percent cover of biocrust to the nearest 5% and counted all individuals of the grass species listed above. The species were separated into those with or without flower stalks. Finally, to estimate the effect of crust on juvenile Artemisia , we sampled five 20 × 100 m areas for juveniles (< 20 cm tall) and recorded whether they were in the open, under adult shrubs, or within a biocrust patch.
To experimentally test the effects of biocrusts on Artemisia , we established 20 30 × 30 cm plots in areas of undisturbed biocrust cover in June 1996. In half of these plots, the biocrusts were removed and in the other half biocrusts were left intact. A one-year-old Artemisia plant was transplanted into the center of each plot. These juvenile Artemisia had been grown from locally collected seed in a blend of silica sand and soils collected from the field study site in a greenhouse before transplanting. Each plot was caged with 12-gauge wire mesh to prevent trampling and herbivory by native ungulates. These transplants were left in place for two years, but after a year, our observations suggested that growth was very slow, perhaps too slow to provide detectable comparisons of biomass. Therefore, we measured photosynthetic rates of Artemisia towards the end of the growing season (mid-September) in 1996 and again in 1997 using a LI-COR 6200 portable infrared gas analysis system. We placed entire Artemisia seedling shoots into a custom-made 1.2 L cuvette and exposed them to full sunlight without cloud interference. After two years, the aboveground biomass of Artemisia transplants was harvested, dried at 60 °C, and weighed. To experimentally test the effects of biocrusts on Festuca , we established 20 30 × 30 cm plots in areas of undisturbed biocrust cover in June 1996. In half of these plots, the biocrusts were removed and in the other half they were left intact. A one-year-old Festuca plant was transplanted into the center of each plot. These had been grown from locally collected seed in a greenhouse in a blend of silica sand and soils collected from the field study site before transplanting. Each plot was caged with a 12-gauge wire mesh to prevent trampling and herbivory by native ungulates. These transplants were in situ for two years. To test the effect of Artemisia on Festuca , we randomly selected 60 mature Artemisia shrubs and 30 were removed and 30 left undisturbed. We also selected 60 open areas with biological soil crust, at least two meters away from the nearest Artemisia for shade treatments. Half of these 60 open areas were covered with a shade screen treatment using shade cloth that blocked approximately 30% of sunlight and half were left without shade screens. In each of these 120 plots, a single one-year old Festuca plant was transplanted. In mid-September 1997, we measured Festuca survivorship, the proportion of grazed plants, and harvested and weighed aboveground biomass. Any Festuca that showed clear signs of leaf herbivory was counted as grazed. We measured soil moisture gravimetrically across all treatments in both the Artemisia -effect and soil crust-effect experiments. Soil samples were collected in all plots in mid-September 1997 when there had been no precipitation for over a week, from a depth of 5–10 cm. Each sample was weighed, dried at 100 °C for 2 days and weighed again for gravimetric soil moisture. In mid-October 2022, again roughly a week after light precipitation, we sampled soil from 5 to 10 cm depth beneath Artemisia , beneath biocrusts, and in the open. Samples were weighed, dried at 100 °C for 2 days and weighed again for gravimetric soil moisture.
We conducted a greenhouse experiment to test the effects of biocrusts on the germination and establishment of Festuca . The greenhouse was on the University of Montana campus (46.8609, − 113.98401). In June 1996, soil was collected from the field site and was added to 50 × 25 × 10 cm rectangular trays. Biocrusts removed for the field experiment were established on half of these trays, and the other half were left without biocrusts. Fifty Festuca seeds were sown in each tray, spray-misted twice per day until seeds germinated, and were subsequently watered two times per week for the duration of the experiment. Germination was monitored, and after 120 days, all surviving Festuca plants were counted again as “established”, then harvested, dried at 60 °C, and weighed.
We conducted laboratory experiments to determine the cyanobacterial composition of biocrusts and to estimate the nitrogen-fixing capacity of biocrusts over a range of temperatures and water saturation. Biocrusts were collected from the field site. Belnap stated that “cyanobacteria form the matrix of biological soil crusts”, thus we cultured and isolated cyanobacterial species on 1% agar plus medium D and the dominant genera were identified using phase contrast microscopy. For nitrogen-fixation, biocrusts were maintained in enclosed growth chambers under controlled temperature and light similar to those typical in the field. The source of photosynthetically active radiation (PAR) in the 400–700 nm range was from very high output cool-white florescent lamps. Sylvania black light fluorescent lamps (peak output 365 nm, range 280–360 nm) provided a source of UVA radiation, and Westinghouse FS—40 fluorescent lamps (peak output 310 nm, range 280–360 nm) were the source of UVB radiation. Biocrust samples used for the acetylene reduction assays were 2 × 2 cm. The cuvettes used for crusts were 12 ml, injected with 1 ml acetylene (C 2 H 2 ), and incubated for 1 h. Following incubation, gas samples were withdrawn and C 2 H was quantified using a Gow-Mac 69–750 gas chromatograph. Biocrust nitrogenase activity was measured over a range of temperatures from 0 to 45 °C, and a range of biocrust water saturation from about 10% to full saturation.
All statistics were performed in SigmaPlot 14.5 (Systat Software, Inc. 2020) and cross-validated in R 4.2.1 . Linear regressions were used to test the relationships between vascular plant cover in large-scale (> 10 m diameter) patches for both herbaceous plants and Artemisia . For the relationship between Artemisia juveniles and biocrust, we used a Chi-square test, and for comparisons of herbivory on Festuca under Artemisia and in the open we used a Fischer’s exact test. For all other pairwise comparisons, we used t-tests when data were normally distributed and Mann–Whitney Rank Sum tests when assumptions of normality were not met—these are noted in the results. For the four concomitant treatments in the field experiment testing the effects of Artemisia on Festuca , the data failed the Shapiro–Wilk normality test, so we used a Kruskal–Wallis one way ANOVA on ranks followed by Dunn’s Method pairwise multiple comparisons. Our goal for N-fixation was simply to determine the capacity of our biocrusts to provide nitrogen to the system, therefore we grouped rates by two ranges of temperatures and two ranges of water saturation and presented these rates in box plots. For comparisons of soil water among Artemisia , biocrusts and soil without biocrusts in 2022, data were not normally distributed so we used a Kruskal–Wallis one way ANOVA on ranks followed by Tukey tests for pairwise comparisons. For comparisons of ammonium and nitrate in soils we used one way ANOVA followed by Holm-Sidak pairwise multiple comparisons.
Spatial patterns Biocrust cover was 54 ± 4% in large patches of Artemisia vs. 34 ± 4% in large open patches without Artemisia (Mann–Whitney U = 68.5; P > 0.001; Fig. ). In each patch type, increasing vascular plant cover corresponded with decreasing biocrust cover (Fig. ). At this large scale,the six selected bunchgrass species were twice as common in open patches without Artemisia as in Artemisia patches, 19.1 ± 1.4 vs. 10.4 ± 1.9 individuals per m 2 (t = − 5.199, P < 0.001). Festuca , the target experimental species, occurred at similar densities in both patch types; 3.0 ± 0.9 in herbaceous patches vs. 1.6 ± 0.4 in Artemisia patches (Mann–Whitney U = 192.0; P = 0.836). In contrast to the large-scale patterns, at the scale of individual Artemisia shrubs, biocrusts were less abundant under Artemisia than in the open matrix surrounding Artemisia . Within Artemisia patches, and under individual Artemisia shrubs, biocrust cover was 39 ± 4% vs. 84 ± 3% in the inter-shrub open matrix (Mann–Whitney U = 204.0, P < 0.001). The richness of the six selected bunchgrass species under Artemisia was 0.9 ± 0.2 vs. 1.9 ± 0.1 per 0.25 m 2 in the open matrix (Mann–Whitney U = 602.50, P < 0.001). The total density of these species was 1.9 ± 0.3 per m 2 under Artemisia vs. 4.7 ± 0.4 in the open matrix (Mann–Whitney U = 531.0, P < 0.001). There were more reproductive Festuca in the open matrix than under Artemisia (0.5 ± 0.1 vs. 0.3 ± 0.1 per m 2 ; Mann–Whitney U = 1019.000; P = 0.050), but non-reproductive Festuca were four times more abundant under Artemisia (0.8 ± 0.2 vs. 0.2 ± 0.1 per m 2 ; Mann–Whitney U = 947.0; P = 0.004). We recorded 23 juvenile Artemisia in Artemisia patches, none of which were under mature Artemisia or in open soils without biocrust, and all of which were in biocrusts. Based on this observed number and the expected numbers derived from the 84% biocrust cover in the interstitial spaces among Artemisia shrubs, biocrusts correlated positively with the recruitment of Artemisia (Chi-square test, χ 2 = 4.41, P = 0.036). Field experiments The aboveground biomass of juvenile Artemisia planted in biocrusts did not differ from that of Artemisia planted where biocrusts had been removed (Fig. a; t = 0.386; df = 2,14; P = 0.706). However, these same Artemisia planted in biocrusts photosynthesized at higher rates in 1997 than those planted where biocrusts had been removed (Fig. c; t = 6.301; df = 2,11; P < 0.001), but this was not the case in 1996 (Fig. b). Festuca planted in biological soil crust was almost twice as large after two growing seasons as those growing in no biological soil crust treatments (Mann–Whitney U = 55.0; df = 2, 39; P < 0.001; Fig. ). Artemisia facilitated Festuca after two growing seasons, as Festuca biomass under Artemisia canopies was roughly twice as much as those in the open (Kruskal–Wallis ANOVA on Ranks, Dunn’s pairwise, Q = 2.815, P = 0.029; Fig. ). Festuca biomass under Artemisia canopies was also different than that of Festuca planted under shade placed in the open (Q = 3.494, P = 0.003). However, the biomass of Festuca planted under Artemisia canopies was not different than those planted where Artemisia had been removed (Q = 1.960, P = 0.300). Greenhouse experiment Biocrusts strongly facilitated Festuca in the greenhouse experiment. When grown in crusts, Festuca germination was roughly five times higher than without crusts (Fig. ; Mann–Whitney U = 55.000 , P < 0.001), and establishment was almost four times greater in the biocrust treatment (t = 4.538; df = 2,38; P < 0.001). Potential mechanisms Nitrogen Laboratory cultures of biological soil crust cyanobacterial species found four predominant genera of cyanobacteria in the biocrusts. Two of these genera, Scytonema and Nostoc , are strong nitrogen fixers, and Microcoleus and Phormidium are weak nitrogen fixers. Laboratory assays of biocrust nitrogen fixation suggest that under a broad range of natural field temperature and moisture conditions, sagebrush steppe biocrusts fix a potentially significant amount of nitrogen (Fig. ). In 2022, concentrations of ammonium did not differ among soils under biocrusts, under Artemisia or in the open (Fig. ; one way ANOVA, F = 1.365; df = 1,29; P = 0.271). Nitrate concentrations were higher under biocrusts (t = 2.917; P = 0.020) and Artemisia (t = 2.376; P = 0.048) than in the open (F = 4.754; df = 1,29; P = 0.016). There was no difference in nitrate in soils beneath biocrusts and Artemisia (t = 0.435; P = 0.667). Water In 1997, when soils were much wetter overall, biocrusts had no effect on soil moisture content in mid-September 1997, with the mean percent water content of soil under biocrusts at 11.8 ± 1.1% and soil without biocrusts at 11.7 ± 1.5% (t = 0.0408; df = 26.1, P = 0.968). In 2022, when soils were much drier overall, soil moisture was 4.0 ± 0.7% under biocrusts vs. 2.3 ± 1.3% in the open without biocrusts (Kruskal–Wallis one way ANOVA on Ranks, H = 11.821, df = 2, P = 0.003; Tukey test = 0.003). In 1997, soil moisture under Artemisia did not differ from that in the open (11.5 ± 0.5 vs. 11.0 ± 0.4; t = 0.713, df = 3,61, P = 0.479). In 2022, soil moisture was 2.4 ± 0.2% under Artemisia , which did not differ from that in the open without biocrusts (Tukey test, P = 0.712), but was lower than that under biocrusts (Tukey test, P = 0.033). Herbivory Of the surviving Festuca planted under Artemisia , 32% (9 of 28) were damaged by herbivores, vs. 83% (19 of 23) in the open matrix (Fischer’s exact test = 0.002; P < 0.001).
Biocrust cover was 54 ± 4% in large patches of Artemisia vs. 34 ± 4% in large open patches without Artemisia (Mann–Whitney U = 68.5; P > 0.001; Fig. ). In each patch type, increasing vascular plant cover corresponded with decreasing biocrust cover (Fig. ). At this large scale,the six selected bunchgrass species were twice as common in open patches without Artemisia as in Artemisia patches, 19.1 ± 1.4 vs. 10.4 ± 1.9 individuals per m 2 (t = − 5.199, P < 0.001). Festuca , the target experimental species, occurred at similar densities in both patch types; 3.0 ± 0.9 in herbaceous patches vs. 1.6 ± 0.4 in Artemisia patches (Mann–Whitney U = 192.0; P = 0.836). In contrast to the large-scale patterns, at the scale of individual Artemisia shrubs, biocrusts were less abundant under Artemisia than in the open matrix surrounding Artemisia . Within Artemisia patches, and under individual Artemisia shrubs, biocrust cover was 39 ± 4% vs. 84 ± 3% in the inter-shrub open matrix (Mann–Whitney U = 204.0, P < 0.001). The richness of the six selected bunchgrass species under Artemisia was 0.9 ± 0.2 vs. 1.9 ± 0.1 per 0.25 m 2 in the open matrix (Mann–Whitney U = 602.50, P < 0.001). The total density of these species was 1.9 ± 0.3 per m 2 under Artemisia vs. 4.7 ± 0.4 in the open matrix (Mann–Whitney U = 531.0, P < 0.001). There were more reproductive Festuca in the open matrix than under Artemisia (0.5 ± 0.1 vs. 0.3 ± 0.1 per m 2 ; Mann–Whitney U = 1019.000; P = 0.050), but non-reproductive Festuca were four times more abundant under Artemisia (0.8 ± 0.2 vs. 0.2 ± 0.1 per m 2 ; Mann–Whitney U = 947.0; P = 0.004). We recorded 23 juvenile Artemisia in Artemisia patches, none of which were under mature Artemisia or in open soils without biocrust, and all of which were in biocrusts. Based on this observed number and the expected numbers derived from the 84% biocrust cover in the interstitial spaces among Artemisia shrubs, biocrusts correlated positively with the recruitment of Artemisia (Chi-square test, χ 2 = 4.41, P = 0.036).
The aboveground biomass of juvenile Artemisia planted in biocrusts did not differ from that of Artemisia planted where biocrusts had been removed (Fig. a; t = 0.386; df = 2,14; P = 0.706). However, these same Artemisia planted in biocrusts photosynthesized at higher rates in 1997 than those planted where biocrusts had been removed (Fig. c; t = 6.301; df = 2,11; P < 0.001), but this was not the case in 1996 (Fig. b). Festuca planted in biological soil crust was almost twice as large after two growing seasons as those growing in no biological soil crust treatments (Mann–Whitney U = 55.0; df = 2, 39; P < 0.001; Fig. ). Artemisia facilitated Festuca after two growing seasons, as Festuca biomass under Artemisia canopies was roughly twice as much as those in the open (Kruskal–Wallis ANOVA on Ranks, Dunn’s pairwise, Q = 2.815, P = 0.029; Fig. ). Festuca biomass under Artemisia canopies was also different than that of Festuca planted under shade placed in the open (Q = 3.494, P = 0.003). However, the biomass of Festuca planted under Artemisia canopies was not different than those planted where Artemisia had been removed (Q = 1.960, P = 0.300).
Biocrusts strongly facilitated Festuca in the greenhouse experiment. When grown in crusts, Festuca germination was roughly five times higher than without crusts (Fig. ; Mann–Whitney U = 55.000 , P < 0.001), and establishment was almost four times greater in the biocrust treatment (t = 4.538; df = 2,38; P < 0.001).
Nitrogen Laboratory cultures of biological soil crust cyanobacterial species found four predominant genera of cyanobacteria in the biocrusts. Two of these genera, Scytonema and Nostoc , are strong nitrogen fixers, and Microcoleus and Phormidium are weak nitrogen fixers. Laboratory assays of biocrust nitrogen fixation suggest that under a broad range of natural field temperature and moisture conditions, sagebrush steppe biocrusts fix a potentially significant amount of nitrogen (Fig. ). In 2022, concentrations of ammonium did not differ among soils under biocrusts, under Artemisia or in the open (Fig. ; one way ANOVA, F = 1.365; df = 1,29; P = 0.271). Nitrate concentrations were higher under biocrusts (t = 2.917; P = 0.020) and Artemisia (t = 2.376; P = 0.048) than in the open (F = 4.754; df = 1,29; P = 0.016). There was no difference in nitrate in soils beneath biocrusts and Artemisia (t = 0.435; P = 0.667). Water In 1997, when soils were much wetter overall, biocrusts had no effect on soil moisture content in mid-September 1997, with the mean percent water content of soil under biocrusts at 11.8 ± 1.1% and soil without biocrusts at 11.7 ± 1.5% (t = 0.0408; df = 26.1, P = 0.968). In 2022, when soils were much drier overall, soil moisture was 4.0 ± 0.7% under biocrusts vs. 2.3 ± 1.3% in the open without biocrusts (Kruskal–Wallis one way ANOVA on Ranks, H = 11.821, df = 2, P = 0.003; Tukey test = 0.003). In 1997, soil moisture under Artemisia did not differ from that in the open (11.5 ± 0.5 vs. 11.0 ± 0.4; t = 0.713, df = 3,61, P = 0.479). In 2022, soil moisture was 2.4 ± 0.2% under Artemisia , which did not differ from that in the open without biocrusts (Tukey test, P = 0.712), but was lower than that under biocrusts (Tukey test, P = 0.033). Herbivory Of the surviving Festuca planted under Artemisia , 32% (9 of 28) were damaged by herbivores, vs. 83% (19 of 23) in the open matrix (Fischer’s exact test = 0.002; P < 0.001).
Laboratory cultures of biological soil crust cyanobacterial species found four predominant genera of cyanobacteria in the biocrusts. Two of these genera, Scytonema and Nostoc , are strong nitrogen fixers, and Microcoleus and Phormidium are weak nitrogen fixers. Laboratory assays of biocrust nitrogen fixation suggest that under a broad range of natural field temperature and moisture conditions, sagebrush steppe biocrusts fix a potentially significant amount of nitrogen (Fig. ). In 2022, concentrations of ammonium did not differ among soils under biocrusts, under Artemisia or in the open (Fig. ; one way ANOVA, F = 1.365; df = 1,29; P = 0.271). Nitrate concentrations were higher under biocrusts (t = 2.917; P = 0.020) and Artemisia (t = 2.376; P = 0.048) than in the open (F = 4.754; df = 1,29; P = 0.016). There was no difference in nitrate in soils beneath biocrusts and Artemisia (t = 0.435; P = 0.667).
In 1997, when soils were much wetter overall, biocrusts had no effect on soil moisture content in mid-September 1997, with the mean percent water content of soil under biocrusts at 11.8 ± 1.1% and soil without biocrusts at 11.7 ± 1.5% (t = 0.0408; df = 26.1, P = 0.968). In 2022, when soils were much drier overall, soil moisture was 4.0 ± 0.7% under biocrusts vs. 2.3 ± 1.3% in the open without biocrusts (Kruskal–Wallis one way ANOVA on Ranks, H = 11.821, df = 2, P = 0.003; Tukey test = 0.003). In 1997, soil moisture under Artemisia did not differ from that in the open (11.5 ± 0.5 vs. 11.0 ± 0.4; t = 0.713, df = 3,61, P = 0.479). In 2022, soil moisture was 2.4 ± 0.2% under Artemisia , which did not differ from that in the open without biocrusts (Tukey test, P = 0.712), but was lower than that under biocrusts (Tukey test, P = 0.033).
Of the surviving Festuca planted under Artemisia , 32% (9 of 28) were damaged by herbivores, vs. 83% (19 of 23) in the open matrix (Fischer’s exact test = 0.002; P < 0.001).
The most salient aspect of our results is that biocrusts facilitated a key vascular plant in this system, the shrub Artemisia tridentata. Through this facilitation, biocrusts appeared to amplify a cascade of extended effects ranging from facilitative effects on Festuca idahoensis to apparent competitive effects on other bunchgrasses, juvenile Artemisia , and biocrusts themselves; the latter three all based on spatial patterns (Fig. ). Importantly, the effect of Artemisia on biocrusts may depend on the scale at which patterns are measured. At larger scales, the negative relationship between Artemisia shrubs and grasses in general appeared to create interstitial spaces around shrubs that actually increased the abundance of biocrusts in large shrubby patches relative to large herbaceous patches, despite the negative effects of individual shrubs within large patches. This may be because bunchgrass density was roughly four times less in these interstitial spaces than in the large open patches without Artemisia . Soliveres and Eldridge described similar scale-based differences in positive and negative spatial associations among shrubs and biocrusts in Australia, with biocrust cover and diversity increasing with shrub cover at large scales, and negative effects at small scales. In addition to indirect facilitation through Artemisia , biocrusts also strongly directly facilitated Festuca , evidenced by higher germination and establishment in the greenhouse, and higher growth in the field experiment. Secondary components of cascades included facilitative effects of Artemisia on Festuca —increased growth of Festuca under Artemisia in field experiments and spatial associations between non-flowering Festuca and Artemisia shrubs in the field. Other cascading outcomes, based on spatial associations, included the apparent suppression of other bunchgrasses and juvenile Artemisia by adult Artemisia , with all suggesting that biocrusts may drive important cascades in intermountain shrub steppe. Components of these chains of interactions are consistent with facilitation cascades described in shallow intertidal zones and sea beds , . The latter described a scenario in which cordgrass establishes and then facilitates a number of other species including mussels, snails, and seaweeds. Much like Artemisia in our case, facilitated mussels increase the densities of other species such as amphipods and barnacles (also see Altieri et al. ). They suggested that cordgrass acted as a “primary facilitator”, perhaps like biocrusts, and mussels acted as a “secondary facilitator”, perhaps like Artemisia . An important difference in our system, however, was that biocrust-facilitated Artemisia appeared to have many negative, or competitive, effects on most bunchgrass species. Thus, our results might be more like results for “multiple, independent cascades” in mangrove-dominated systems that appeared to be initiated by “a single basal facilitator” . The spatial patterns of the large suite of species we measured also suggest a cascade more in line with that of Gribben et al. . They presented evidence for “a realized facilitation cascade” as a “function of nested negative and positive interactions” that vary as the density of key species varies, rather than a “collection of hierarchical positive interactions”. Although we do not explore this, these complex negative and positive interactions have the potential to maintain community diversity through intransitive interactions . Other facilitation cascades in the literature generally have more obligate relationships than the biocrust → Artemisia → Festuca cascade explored here. In part, this may be because none of our cascade components are consumers that can obviously have strong, obligate relationships with prey abundances. Each of our autotrophic components can certainly exist without the others, thus, the cascades suggested by our results are relatively more facultative and contextual than others reported in the literature. The variable negative and positive interactions in a realized facilitation cascade might be related to facilitation and competition being highly context dependent, and such context dependence appears to be very much the case for the effects of biocrusts in general . In a large-scale multi-factorial field experiment in intermountain prairie, vegetation very similar to ours, biocrusts inhibited the overall emergence of four different exotic seedlings in a wet year, but not a dry year, and had no effect on the emergence of four different native species . The effects of biocrusts in rosemary scrub in Florida vary among species, distance from shrubs, and time since fire . Experimental removal of biocrusts from around established Bouteloua gracilis bunchgrasses reduced their performance . These kinds of variable effects are highlighted in a meta-analytical review which found that biocrusts dominated by lichens, such as ours, inhibited overall plant performance—on average . In this review, biocrusts had more positive effects on plant species without mutualistic nitrogen-fixing bacteria, such as our Festuca and Artemisia than species with these symbionts. More importantly, they found that lichen-dominated biocrusts across studies suppressed germination, the opposite of what we found in the greenhouse experiment for germination and establishment, but improved growth, as we found in the field for Festuca , and corresponded with the higher photosynthetic rates of Artemisia in biocrusts in the field. The effects of Artemisia on native bunchgrasses are less studied than those of biocrusts, but they are also variable. Early studies indicated that Artemisia effects are generally, but conditionally, competitive , , , and in a detailed experimental study the roots of Artemisia outcompeted those of P. spicata for soil nitrogen . However, others have found that Poa secunda , a species apparently excluded from Artemisia understories in our study, was more common under Artemisia than in the open, but only at the driest sites . In this study, Elymus elymoides abundance was higher under Artemisia canopies than in open interspaces at both drier and wetter sites. In experimental manipulations, Poulos et al. found that Artemisia facilitated the native forb, Penstemon palmeri , but this generated complex interactions with other herbaceous species in the understory, another aspect of interaction cascades that we did not pursue. In eastern Oregon, a wide range of different functional groups of vascular plants are positively associated with Artemisia much as found for many other shrub species , . Furthermore, there is evidence that Artemisia can have positive allelopathic (i.e., biochemically mediated) effects on conspecifics and negative allelopathic effects on understory grasses . As a group, our six selected native bunchgrass species were negatively associated with Artemisia . All of this variability in shrub effects considered the facilitative effect of biocrusts on Artemisia clearly has the potential to affect a diverse variety of interaction cascades in addition to ours. In fact, although we focused solely upon autotrophs in our study, because less is known about cascades among autotrophs, Artemisia is a keystone species for consumers. Artemisia provides habitat for obligate consumers including the greater sage-grouse ( Centrocerus urophasianus ), sagebrush sparrow ( Artemisiospiza nevadensis ), sage thrasher ( Oreoscoptes montanus ), and pronghorn antelope ( Antilocapra americana ) . As such, the maintenance of larger scale community structure and diversity in sagebrush steppe may be affected by even more extended and trophic biocrust- Artemisia facilitation cascades. Our results indicating a positive effect of biocrusts on Artemisia are, to our knowledge, somewhat atypical in the literature; however, a broad survey across the Great Basin found that biological crusts correlated positively with greater growth of sagebrush canopies . Zhang and Belnap found that biocrusts improved the concentration of various nutrients in the leaf tissue of two Asian shrub species. And in a meta-analysis, averaged across all biocrust types, biocrusts facilitated the growth of non-nitrogen-fixing shrubs, such as Artemisia . Artemisia tridentata with biocrusts cover vast parts of western North America, and yet we know surprisingly little about the effects of this shrub species on biocrusts. Condon et al. found that Artemisia -dominated plant communities were associated with “rapidly reproducing biocrusts”. Navas Romeroa et al. found positive spatial associations among different shrub species and biocrusts at some sites, but neutral associations at other sites. Others have reported higher biocrust development under Artemisia ordosica canopies in China – . In New South Wales, Australia increasing shrub cover, at the large scale of “sites”, corresponded with increasing biocrust richness and cover, much like the large-scale patterns we report here . However, much as we found at small scales, biocrusts responded negatively to shrubs as compared to open patches . In our study, biocrust cover decreased as vascular plant cover increased in both large-scale Artemisia and large-scale herbaceous patches, consistent with the small-scale negative relationship between Artemisia individuals and biocrust cover. She et al. found that biocrust cover rapidly increased with the cover of shrubs, dominated by A. ordosica when shrub productivity was low. As shrub productivity increased, the cover of biocrust did not decrease, but stabilized. In contrast, and like our results, biocrusts were abundant at low herbaceous productivity but declined rapidly with increasing herbaceous productivity. This corresponded with a strong decline in light associated with increasing herbaceous productivity, but not with increasing shrub productivity. One overlooked potential component of our cascade is that Artemisia may have substantially altered the composition of biocrusts, which might have had consequences for vascular plants, but which we did not measure. In the field, we observed that biocrusts under mature Artemisia canopies appeared to contain much more moss, Tortula roralis , than biocrusts in the open. Moss dominated crusts might have weaker positive or stronger competitive effects on vascular plants than cyanobacteria and lichen-dominated biocrusts, which were more common in the open, and with which we conducted laboratory measurements. Biocrusts can improve the retention of soil water and increase concentrations of soil nitrogen , , . Our measurements were only taken as snapshots in time but suggest that both of these processes may have been important. When soils were wet, in 1997, there were no differences in percent moisture among Artemisia , biocrusts, and open soil with no biocrusts. But, in 2022, when soils were much drier, soil under biocrusts was wetter than either in the open or under Artemisia , suggesting that biocrusts may help retain water. Also in 2022, soil nitrate concentrations were almost three times higher under biocrust as in the open, but nitrate under Artemisia was as high as under biocrusts. Interestingly, the mechanism by which Artemisia facilitated Festuca also appeared to be substrate related, perhaps via nitrogen as suggested by our results; i.e., in treatments where Artemisia was removed the biomass of Festuca was not significantly different than where it was not removed, perhaps due to its legacy effects on soils. Furthermore, when shade was provided to open sites where shrubs had not affected soils, Festuca biomass did not increase. However, shrubs can affect other species via many other effects on substrate that we did not measure . Furthermore, because the species involved have different lifespans and recruitment requirements, these effects must have temporal components which are best considered in the context of priority effects but exploring priority effects are beyond the scope of this study. For simplicity, we treated biocrusts as a single entity, but biocrusts are exceedingly complex communities themselves, being dynamic combinations of nitrogen-fixing cyanobacteria and cyanolichens, other bacteria, green algae, fungi, mosses, and liverworts . Pietrasiak et al. compared functions among 10 different categories of biocrusts, each which certainly can vary in composition, and found that cyanolichen-dominated crusts fixed nitrogen at higher rates than all other crust types, photosynthesized at higher rates than most other crust types, and generated soil aggregate stability as high as any other crust type. The biocrusts in the open at our sites were dominated by lichens and combined with our identification of cyanobacteria indicate that we predominantly measured cyanolichen crusts. Importantly, different biocrusts, or different compositions of cyanolichen crusts, could potentially generate different and unique interaction chains and cascades, and this is an opportunity for deeper context specific research in arid and semi-arid systems that support biocrusts.
Supplementary Information.
|
The usefulness of immunohistochemistry for phosphohistone H3 as a prognostic factor in myxoid liposarcoma | b5907aff-af6f-4ea6-9821-a59ed948627d | 10036607 | Anatomy[mh] | Soft tissue sarcoma (STS) is a rare malignant neoplasm that accounts for approximately 1% of all malignancies. Larger, deep-seated, and histological high-grade tumors are associated with poor survival. In addition, specific prognostic factors for some subtypes of STS are documented in the literature , . For instance, our study described the role of several glycans in STS with myxoid substance and reported that chondroitin sulfate synthase 1 (CHSY1) expression was closely associated with their malignant potential . Since myxoid liposarcoma (MLS) is one of the most common subtypes of STS , – , the same study examined CHSY1 expression in MLS . The results showed that the frequency of CHSY1 in MLS was 25%, which was lower than that of other histologic types including myxofibrosarcoma, malignant peripheral nerve sheath tumor, and low-grade fibromyxoid sarcoma; however, the expression was limited to those showing round cell morphology indicative of poor patient prognosis. Although previous reports have shown prognostic factors for MLS including tumor size, surgical margin, histological grade, distant metastasis, and proportion of round cell component that has been described as a hypercellularity in the 2020 WHO classification, its pathological prognostic factors remain to be fully defined – . Therefore, we sought other biomarkers that predict the prognosis of these patients more accurately. Epigenetic changes in DNA or histone modification have recently been indicated to be related to the malignant transformation of cells. Histone is the core protein of nucleosome that is a fundamental structure of chromatin. The modification of histone tails can diversify the chromatin structure and promote or suppress the proliferation of cells. In particular, the phosphorylation of histone H3 is known to be related to mitosis , . Phosphohistone H3 (PHH3) is specifically expressed in the G2-M period of the cell cycle, in contrast to Ki-67 antigen which is expressed in every period except G0 , . Accordingly, there are several reports that argue the usefulness of PHH3 as a marker of mitotic count in brain tumor, melanoma, gastrointestinal stromal tumors (GIST), and others – . Contrariwise, there are very few reports for sarcoma patients. We focused on the assessment of PHH3 as a new index for the histological grading of MLS. Since the tendency of PHH3 among various subtypes of sarcoma is unknown, we focused our study on a single tumor type. In this study, we investigated the relationship between the number or proportion of PHH3-positive tumor cells and the prognosis in MLS, and we assessed the usefulness of immunohistochemistry for PHH3 as a potential biomarker in comparison with Ki-67 antigen and other prognostic factors.
Clinical characteristics of patients The median age of patients at operation was 53.0 years (range 31–85), of which 15 were males and ten were females. The median follow-up period was 8.2 years (range 1.5–22.3). The median diameter of tumors was 9.0 cm (range 4.4–23.0). The tumors eventually metastasized in five patients, including three patients who were already detected at first visit. The oncological outcomes were continuous disease free (CDF) in 17 patients, alive with disease (AWD) in one, dead of disease (DOD) in five, and dead of other disease (DOOD) in two (Table ). The outcomes of all patients with metastasis were DOD. Clinicopathological and immunohistochemical data for each individual patient are shown in Supplemental Table . Statistical analysis of PHH3-positive tumor cells in MLS In PHH3 immunohistochemistry, the number of total tumor cells and positive tumor cells on the hot spots were measured in 10 high power fields (HPF) (Fig. ). The positive rate of PHH3 was subsequently calculated for consideration as PHH3 index. The median number of PHH3-positive tumor cells was 6 (range 0–45) cells per 10 HPF, and the PHH3 index was 0.3 (range 0.0–10.7) % on immunostained specimens (Table ). The cut-off values of the highest sensitivity and specificity were determined by the receiver operating characteristic (ROC) curves. For example, the cut-off values of PHH3-positive tumor cells and the PHH3 index were 30 cells per 10 HPF and 2%, respectively (Fig. A, Table ). The differences in disease-specific survival rate (DSS) at cut-off values were calculated by the Kaplan–Meier method (Fig. A, B). In the univariate analysis using the Log-rank test, the number of PHH3-positive tumor cells and the PHH3 index demonstrated a statistically significant difference ( P < 0.001). In both the number of PHH3-positive tumor cells and the PHH3 index, the five-year DSS below and above the cut-off value was 90.4% and 33.3%, respectively (Table ). In the multivariate analysis using Cox proportional-hazard regression analyses, the PHH3 index was a significant factor ( P = 0.03) (Table ). Statistical analysis of Ki-67 in MLS Ki-67 was measured by the same method with PHH3 for comparison (Fig. ). The median number of Ki-67-positive tumor cells and the Ki-67 index were 54 (range 0–984) cells per 10 HPF and 2.2 (range 0.0–31.3) %, respectively (Table ). The cut-off value of the Ki-67 index was determined to be 19% based on the ROC curve (Fig. B). The value of Ki-67-positive tumor cells was difficult to determine due to its low area under the curve (AUC) value on the ROC curve. The differences in DSS at the cut-off value were calculated using the Kaplan–Meier method (Fig. C), and univariate analysis was performed using the Log-rank test. The Ki-67 index presented a statistically significant difference ( P = 0.01). The five-year DSS below and above the cut-off value was 86.5% and 50.0% in the Ki-67 index (Table ). In the multivariate analysis, the Ki-67 index was not a significant factor ( P = 0.10) (Table ). Correlation between the PHH3 and Ki-67 indices There was a significant and weak positive correlation between the PHH3 and Ki-67 indices ( P = 0.01; correlation coefficient = 0.496) (Fig. ). Other factors Other prognostic factors described in previous reports were investigated, including age, sex, depth, size, surgical margin, the French Fédération Nationale des Centres de Lutte Contre Le Cancer grading system (FNCLCC), necrosis, and proportion of round cell component (RC%). The following cut-off values were based on previously reported literature: age, 60 years; size, 10 cm; RC%, 5% , , , – . Age and sex demonstrated statistical significance in univariate analysis ( P ≤ 0.05). A grade of II or more under the FNCLCC system tended to be associated with the poor survival rate ( P ≤ 0.10) (Fig. D). RC% was a significant factor in the multivariate analysis ( P = 0.02) (Table ). Depth, size, surgical margin and necrosis did not exhibit statistical significance (Table ).
The median age of patients at operation was 53.0 years (range 31–85), of which 15 were males and ten were females. The median follow-up period was 8.2 years (range 1.5–22.3). The median diameter of tumors was 9.0 cm (range 4.4–23.0). The tumors eventually metastasized in five patients, including three patients who were already detected at first visit. The oncological outcomes were continuous disease free (CDF) in 17 patients, alive with disease (AWD) in one, dead of disease (DOD) in five, and dead of other disease (DOOD) in two (Table ). The outcomes of all patients with metastasis were DOD. Clinicopathological and immunohistochemical data for each individual patient are shown in Supplemental Table .
In PHH3 immunohistochemistry, the number of total tumor cells and positive tumor cells on the hot spots were measured in 10 high power fields (HPF) (Fig. ). The positive rate of PHH3 was subsequently calculated for consideration as PHH3 index. The median number of PHH3-positive tumor cells was 6 (range 0–45) cells per 10 HPF, and the PHH3 index was 0.3 (range 0.0–10.7) % on immunostained specimens (Table ). The cut-off values of the highest sensitivity and specificity were determined by the receiver operating characteristic (ROC) curves. For example, the cut-off values of PHH3-positive tumor cells and the PHH3 index were 30 cells per 10 HPF and 2%, respectively (Fig. A, Table ). The differences in disease-specific survival rate (DSS) at cut-off values were calculated by the Kaplan–Meier method (Fig. A, B). In the univariate analysis using the Log-rank test, the number of PHH3-positive tumor cells and the PHH3 index demonstrated a statistically significant difference ( P < 0.001). In both the number of PHH3-positive tumor cells and the PHH3 index, the five-year DSS below and above the cut-off value was 90.4% and 33.3%, respectively (Table ). In the multivariate analysis using Cox proportional-hazard regression analyses, the PHH3 index was a significant factor ( P = 0.03) (Table ).
Ki-67 was measured by the same method with PHH3 for comparison (Fig. ). The median number of Ki-67-positive tumor cells and the Ki-67 index were 54 (range 0–984) cells per 10 HPF and 2.2 (range 0.0–31.3) %, respectively (Table ). The cut-off value of the Ki-67 index was determined to be 19% based on the ROC curve (Fig. B). The value of Ki-67-positive tumor cells was difficult to determine due to its low area under the curve (AUC) value on the ROC curve. The differences in DSS at the cut-off value were calculated using the Kaplan–Meier method (Fig. C), and univariate analysis was performed using the Log-rank test. The Ki-67 index presented a statistically significant difference ( P = 0.01). The five-year DSS below and above the cut-off value was 86.5% and 50.0% in the Ki-67 index (Table ). In the multivariate analysis, the Ki-67 index was not a significant factor ( P = 0.10) (Table ).
There was a significant and weak positive correlation between the PHH3 and Ki-67 indices ( P = 0.01; correlation coefficient = 0.496) (Fig. ).
Other prognostic factors described in previous reports were investigated, including age, sex, depth, size, surgical margin, the French Fédération Nationale des Centres de Lutte Contre Le Cancer grading system (FNCLCC), necrosis, and proportion of round cell component (RC%). The following cut-off values were based on previously reported literature: age, 60 years; size, 10 cm; RC%, 5% , , , – . Age and sex demonstrated statistical significance in univariate analysis ( P ≤ 0.05). A grade of II or more under the FNCLCC system tended to be associated with the poor survival rate ( P ≤ 0.10) (Fig. D). RC% was a significant factor in the multivariate analysis ( P = 0.02) (Table ). Depth, size, surgical margin and necrosis did not exhibit statistical significance (Table ).
In the present study, we demonstrated that the number of PHH3-positive tumor cells and PHH3 index show a significant correlation with the prognosis of MLS patients. Hendzel et al. , reported that the phosphorylation of histone H3 at Ser10 represents a powerful marker for mitotic chromosome condensation in cell proliferation. It is known that PHH3 at Ser10 or Ser28 is specifically expressed during G2 to M phases of a cell cycle, and some authors reported its usefulness as a marker to detect mitotic forms . Fukushima et al. suggested that PHH3 may be a sensitive and useful marker for meningioma grading based on the mitotic figures in the WHO criteria. Alkhasawneh et al. presented that PHH3 is associated with inferior overall survival in GIST compared to Ki-67. The effect of immunohistochemistry of PHH3 has also been described in several reports on astrocytoma, melanoma, uterine smooth muscle tumors, pulmonary neuroendocrine carcinomas, and other malignancies – . In reference to these articles, we focused on PHH3 in MLS patients and investigated the immunohistochemistry of PHH3, in addition to its correlation with prognosis and usefulness as a marker of malignancy compared to other prognostic factors. In this study, we determined a cut-off value of 30 cells per 10 HPF in the number of PHH3-positive tumor cells and 2% in the PHH3 index. Statistically significant differences were found in both criteria by univariate analysis and in the PHH3 index by multivariate analysis, and these significant differences showed their usefulness as predictive factors of progression. Ki-67 is widely known as an indicator of cell proliferation. This protein exists in the nucleus and is expressed in all phases except G0 of the mitotic cycle . It is used as a predictive factor of tumor prognosis. For example, Pathmanathan et al. reported that the Ki-67 index was the most powerful and independent predictor of survival in node-negative patients with breast cancer. In this study, a Ki-67 index of over 19% was significantly related to prognosis in MLS patients by univariate analysis. Furthermore, a significant and weak positive correlation was detected between the PHH3 and Ki-67 indices. The weak correlation potentially caused by the result skewed by a few outliers, therefore, if there had been more cases, the correlation may have been higher. Previous studies have also reported a significant correlation between these indices, and the PHH3 index has been identified as a more sensitive predictor of survival , . Several prognostic factors of MLS were reported in previous articles. In addition, many authors have considered a range of factors related to prognosis, such as tumor size, histological grade, local recurrence, distant metastasis, age, and sex – , , . The proportion of round cell component, FNCLCC system, Broders’ Grading System, and American Joint Committee on Cancer system (AJCC system) have been adopted to assess histological grade; however, a large body of literature has validated that a statistically significant outcome was obtained with an RC% of greater than 5% , , , . In the current study, age and sex were significant factors in the univariate analysis, and RC% was also significant in the multivariate analysis; however, the FNCLCC system did not show statistical significance. According to the results of this study, we consider both the PHH3 and Ki-67 indices as useful prognostic factors in MLS. However, the PHH3 and Ki-67 indices present different cut-off values of 2% and 19%, respectively. The use of the PHH3 index is easier than other indices due to its small cut-off value, as the detection of a few positive tumor cells in a single HPF can lead to crossing over the threshold. The PHH3 index may be a more convenient figure than the other in that point. The main limitation of our study was the small number of patients. Although the long follow-up period and small number of drop-out cases were notable advantages, further studies should be conducted with a larger number of patients. Due to the difficulty of standardization of immunohistochemistry protocols and readout for PHH3 and Ki-67, the current study may have limited generalizability for other laboratories. Standardized methods such as automated immunostaining or artificial intelligence-assisted detection should be considered in the future. In summary, we examined the histological grade of myxoid liposarcoma, especially the immunohistochemistry of PHH3 and its association with prognosis in this study. The number of PHH3-positive tumor cells, PHH3 index, and Ki-67 index were statistically correlated with the prognosis of MLS. In conclusion, the immunohistochemistry of PHH3 may be associated with prognosis and could serve as a valid criterion of histological grade in MLS. Additional studies about other subtypes of STS are expected in the future.
General information All medical protocols in this study adhered to the Declaration of Helsinki. This study was approved by the Institutional Review Board of Shinshu University School of Medicine (protocol number: 608) with written informed consent obtained from each participant and/or their legal representative. Patient sample We evaluated 32 patients with MLS who were treated at our hospital from 1995 to 2014. Two patients with unknown progression status and 5 patients who underwent non-surgical treatment were excluded. After these exclusions, the remaining 25 patients were included in the study. FUS-DDIT3/EWSR1-DDIT3 fusion was confirmed by fluorescence in situ hybridization testing in 5 patients who were treated since 2010. Immunohistochemistry Formalin-fixed and paraffin-embedded surgical specimens were prepared in cross sections. Sections of specimens with the highest cell density and rich atypical cells were selected. For immunohistochemistry, mouse monoclonal anti-Ki-67 (clone MIB-1) and rabbit polyclonal anti-PHH3 (Cell Marque, cat. no. 369A-15) antibodies were purchased from Dako (Glostrup, Denmark) and Merck (Darmstadt, Germany), respectively. Antigen retrieval for Ki-67 antigen was carried out by heating tissue slides in 10 mM Tris–HCl buffer (pH 8.0) and 1 mM EDTA with a microwave oven for 30 min (for Ki-67) or a pressure cooker for 10 min (for PHH3). As secondary antibodies, a Histofine Simple Stain MAX-PO (M) Kit (Nichirei Biosciences, Tokyo, Japan) was used for Ki-67 antigen, and a Histofine Simple Stain MAX-PO (R) (Nichirei Biosciences) was used for PHH3 antigen. Peroxidase activity was visualized using a diaminobenzidine/H 2 0 2 solution. Counterstaining was conducted using a Carrazzi's Hematoxylin solution (× 2) prepared from hematoxylin (H9627; Sigma-Aldrich, MO, USA). As an on-slide control, tumor cells in the tissue slides were used for positive control. Negative control immunohistochemistry was performed by omitting the primary antibodies from the procedure, and no specific staining was observed. This procedure was determined to be adequate, because the positive and negative controls were stained precisely. Assessment of PHH3 and Ki-67 In both PHH3 and Ki-67 immunohistochemistry, a tumor cell was determined to be positive when more than 70% of the nuclear area of the tumor cell was immunostained dark brown (Supplemental Fig. ). According to previous reports , , , the number of total tumor cells and positive tumor cells on the hot spots of each specimen were manually measured in 10 HPF by a surgical oncologist (Ak.T.) and confirmed by a pathologist (J.N.). The positive rates of PHH3 and Ki-67 were subsequently calculated for consideration as PHH3 and Ki-67 indices. Both were measured by the same method for comparison. The FNCLCC grading, necrosis, and round cell component were evaluated on hematoxylin and eosin (H & E) specimens of the entire cross section by pathologists (T.U. and J.N.). Statistical analysis Prognostic factors of the examination criteria were statistically analyzed. DSS was also evaluated by defining DOD as the endpoint. The cut-off values regarding DOD were determined using an ROC curve for the number of PHH3 positive cells, PHH3 and Ki-67 indices. Other items deemed to be common were based on previous reports. Survival analyses were performed using the Kaplan–Meier method, and univariate analyses were performed using the Log-rank test. The multivariate analysis was performed using the Cox proportional-hazard regression analyses with the PHH3 index, Ki-67 index, and RC% as explanatory variables. Statistical significance was defined as a P -value of 0.05 or less. The software R version 4.0.3 was used for analyses.
All medical protocols in this study adhered to the Declaration of Helsinki. This study was approved by the Institutional Review Board of Shinshu University School of Medicine (protocol number: 608) with written informed consent obtained from each participant and/or their legal representative.
We evaluated 32 patients with MLS who were treated at our hospital from 1995 to 2014. Two patients with unknown progression status and 5 patients who underwent non-surgical treatment were excluded. After these exclusions, the remaining 25 patients were included in the study. FUS-DDIT3/EWSR1-DDIT3 fusion was confirmed by fluorescence in situ hybridization testing in 5 patients who were treated since 2010.
Formalin-fixed and paraffin-embedded surgical specimens were prepared in cross sections. Sections of specimens with the highest cell density and rich atypical cells were selected. For immunohistochemistry, mouse monoclonal anti-Ki-67 (clone MIB-1) and rabbit polyclonal anti-PHH3 (Cell Marque, cat. no. 369A-15) antibodies were purchased from Dako (Glostrup, Denmark) and Merck (Darmstadt, Germany), respectively. Antigen retrieval for Ki-67 antigen was carried out by heating tissue slides in 10 mM Tris–HCl buffer (pH 8.0) and 1 mM EDTA with a microwave oven for 30 min (for Ki-67) or a pressure cooker for 10 min (for PHH3). As secondary antibodies, a Histofine Simple Stain MAX-PO (M) Kit (Nichirei Biosciences, Tokyo, Japan) was used for Ki-67 antigen, and a Histofine Simple Stain MAX-PO (R) (Nichirei Biosciences) was used for PHH3 antigen. Peroxidase activity was visualized using a diaminobenzidine/H 2 0 2 solution. Counterstaining was conducted using a Carrazzi's Hematoxylin solution (× 2) prepared from hematoxylin (H9627; Sigma-Aldrich, MO, USA). As an on-slide control, tumor cells in the tissue slides were used for positive control. Negative control immunohistochemistry was performed by omitting the primary antibodies from the procedure, and no specific staining was observed. This procedure was determined to be adequate, because the positive and negative controls were stained precisely.
In both PHH3 and Ki-67 immunohistochemistry, a tumor cell was determined to be positive when more than 70% of the nuclear area of the tumor cell was immunostained dark brown (Supplemental Fig. ). According to previous reports , , , the number of total tumor cells and positive tumor cells on the hot spots of each specimen were manually measured in 10 HPF by a surgical oncologist (Ak.T.) and confirmed by a pathologist (J.N.). The positive rates of PHH3 and Ki-67 were subsequently calculated for consideration as PHH3 and Ki-67 indices. Both were measured by the same method for comparison. The FNCLCC grading, necrosis, and round cell component were evaluated on hematoxylin and eosin (H & E) specimens of the entire cross section by pathologists (T.U. and J.N.).
Prognostic factors of the examination criteria were statistically analyzed. DSS was also evaluated by defining DOD as the endpoint. The cut-off values regarding DOD were determined using an ROC curve for the number of PHH3 positive cells, PHH3 and Ki-67 indices. Other items deemed to be common were based on previous reports. Survival analyses were performed using the Kaplan–Meier method, and univariate analyses were performed using the Log-rank test. The multivariate analysis was performed using the Cox proportional-hazard regression analyses with the PHH3 index, Ki-67 index, and RC% as explanatory variables. Statistical significance was defined as a P -value of 0.05 or less. The software R version 4.0.3 was used for analyses.
Supplementary Information 1. Supplementary Table 2.
|
Indian consensus statements on irritable bowel syndrome in adults: A guideline by the Indian Neurogastroenterology and Motility Association and jointly supported by the Indian Society of Gastroenterology | 0620ee6f-1adf-4ec7-a744-6d6b6b8826da | 10036984 | Internal Medicine[mh] | Irritable bowel syndrome (IBS) is a common condition, which leads to significant morbidity, work absenteeism, loss of productivity, and economic burden to the society and impacts the quality of life of patients . There have been several recent developments in the understanding of pathophysiology, diagnosis and treatment of IBS in India . Also, the availability of drugs in India is quite different from the rest of the world . The practicing condition of physicians in India is variable too. Thus, the Indian Neurogastroenterology and Motility Association (INMA), earlier called Indian Motility and Functional Disease Association, was keen on conducting an Indian IBS consensus with the following aims: (a) to provide guidelines to physicians for imparting standard care to patients suffering from IBS; (b) to share knowledge about IBS among doctors and healthcare professionals; and (c) contribute to the advancement of research on IBS. This consensus was also jointly supported by the Indian Society of Gastroenterology.
Eighteen consensus team members were chosen from different (eastern, western, northern, and southern) regions of India based on their interest and publications on functional gastrointestinal (GI) disorders, currently called disorders of gut-brain interaction (DGBI), especially IBS. Initially, 37 statements were composed. These statements were emailed to the team members and they were requested to provide their input. The received responses were collated and summarized and uploaded as a document on Google Docs and the team members were invited to edit the statements. Owing to the Coronavirus disease-19 (COVID-19) pandemic, lockdowns, and travel restrictions, face-to-face meetings could not be held for some of the consensus meetings. A Google meet was organized on January 12, 2021, to discuss the statements with the team members; some of the statements were deleted or modified at this stage, resulting in 28 statements in all including diagnostic criteria and epidemiology, etiopathogenesis and comorbidities, investigations, lifestyle modifications and treatments. A modified Delphi method was used in the consensus process. The statements were posted on Survey Monkey (SurveyMonkey Enterprise, San Mateo, CA, USA) for the first round of voting in February 2021. The second and final voting rounds were delayed due to the second wave of the COVID-19 pandemic and could not be held until the end of May 2021. Subsequently, the results of the consensus were presented to the members of the Indian IBS Consensus team on August 8, 2021, via Google meet. All members of the consensus team met in Lucknow, India on 7th of May 2022 and finalized the manuscript during the 5th Annual Congress of the INMA ( www.gimotilityindia.in ), earlier called Indian Motility and Functional Disease Association. A statement was regarded as accepted when the sum of the “completely accepted” and “accepted with minor reservation” votes was 80% or higher. The consensus was achieved on all 28 statements. The responses were collated and the summaries were examined for consensus. As shown in Table , the grade of the evidence and the level of agreement were established on the scheme of the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) Working Group . The result of the consensus was presented to the participating members of the INMA and the other delegates during the 5th Annual Congress of the Association ( www.gimotilityindia.in ) in Lucknow, India, on 7th of May 2022. The levels of evidence were reviewed for which preference was given to studies from India.
Statement 1: IBS is a common condition in clinical practice and the Indian community Voting summary: Accepted completely: (100%). Level of evidence: II-2 Grade of recommendation: B Several epidemiological studies showed IBS to be common in the Indian population . An early house-to-house survey in an urban population of Mumbai (2549 presumably healthy adults) primarily evaluating the prevalence of dyspepsia showed a 7.2% prevalence of IBS . Table summarizes the community-based studies on the prevalence of IBS in the Indian population. It shows the prevalence of IBS in the Indian community varies from 0.4% to 4.2% . The reason for the very low prevalence of IBS in the recent global study might be related to the design of the study . This study included only one member per family during the household survey even though this disorder shows clustering within the family. Hence, it might have underestimated the disease burden. Another reason could be related to the fact that overlap disorders, which is the commonest type of functional gastrointestinal disorder (FGID), might result in under-estimation of the disease burden of pure disorders . This study also showed that household survey generally records a lower prevalence than internet survey possibly due to recruitment bias in the latter method. It is important to note that in India, a household survey was conducted in contrast to the Western countries. Rome IV criteria underdiagnose IBS more than the Rome III criteria . Despite these methodological issues, the lower prevalence of IBS in India is quite noteworthy. It may be related to dietary, socio-cultural, gut microbiota and hygiene hypothesis-related factors. Statement 2: IBS is as prevalent in the male as the female population in India Voting summary: Accepted completely: (88.87%), accepted with minor reservation: (11.11%). Level of evidence: II-2 Grade of recommendation: B IBS is more prevalent among females than males in the West . Many Indian clinic studies showed that more than two-thirds of IBS patients in Indian studies are male . A possibility of referral bias due to male patients more often seeking healthcare facilities, particularly in the advanced centers from where studies are published has been considered . However, even in the community studies, the female preponderance was either lacking or the ratio of male to female was quite close (Table ). Hence, it may be concluded that male subjects also suffer from IBS quite commonly. Statement 3: IBS patients in India often have other overlapping FGIDs Voting summary: Accepted completely: (100%). Level of evidence: II-2 Grade of recommendation: B Earlier, it was thought that FGIDs (currently called DGBI) are isolated disorders without any overlap. However, several studies from Asia, including India, showed that the overlaps between various FGIDs are common rather than exceptions . In an epidemiological study from northern India, of 2774 subjects 413 (14.9%) had dyspepsia alone, 75 (2.7%) IBS alone and 115 (4.1%) had dyspepsia-IBS overlap, respectively by Rome III criteria . Another study by Rome IV criteria in northern India showed functional dyspepsia-IBS overlap in 4.4% of the 1309 subjects . In a recent multicentric study by the Rome Foundation, patients with multiple FGIDs were found to have more psychological comorbidity, healthcare utilization and IBS severity . It must be emphasized that owing to busy work schedules, the overlaps may often be overlooked by physicians, particularly in India and the adjoining nations, as they may primarily focus on predominant symptoms instead of directing adequate attention to recognize the overlaps in their clinical practice . To address these lacunae, it was felt imperative to develop, translate and validate diagnostic questionnaires into regional languages, which helped in the diagnosis of otherwise missed FGID overlaps . Table presents the IBS-FGID overlap in studies conducted entirely or partly in India . Statement 4: Etiopathogenesis of IBS is multi-dimensional including gut-specific mechanisms, altered gut-brain interaction, food intolerance, psychosocial and genetic factors Voting summary: Accepted completely: (100%). Level of evidence: II-2 Grade of recommendation: B The multifactorial pathogenesis of IBS has been reviewed recently . These factors include (i) peripheral factors such as abnormal GI motility, GI inflammation and altered gut permeability, luminal microenvironment alteration including gut microbiota dysbiosis and small intestinal bacterial overgrowth (SIBO), host-microbe interaction, bile acid malabsorption, pathogenic infection, dietary factors and neurohumoral dysregulation including altered serotonergic transmission and visceral hypersensitivity and (ii) central factors (psychological stress, cognitive dysfunction, abnormal emotional arousal system response, and sleep dysfunction). Genetic factors may underlie peripheral and central pathophysiological mechanisms. Statement 5: Patients with IBS, particularly those with diarrhea-predominant IBS (IBS-D), are more likely to have SIBO and gut dysbiosis Voting summary: Accepted completely: (72.22%), accepted with minor reservation: (27.78%). Level of evidence: II-1 Grade of recommendation: A A recent meta-analysis, which included at least four Indian studies, showed that 36.7% of IBS patients had a positive test for SIBO . Patients with IBS were 2.6 and 8.3 times more likely to have a positive test for SIBO as compared with healthy controls using glucose hydrogen breath test (GHBT) and jejunal aspirate culture, respectively . Patients with IBS-D were more likely to have positive GHBT as compared with the other subtypes . Somewhat similar results have been found in another meta-analysis . Statement 6: Excessive methane production slows gut transit and is associated with constipation-predominant IBS (IBS-C) Voting summary: Accepted completely: (77.78%), accepted with minor reservation: (22.22%). Level of evidence: I Grade of recommendation: B A meta-analysis showed that excess breath methane during lactulose or glucose hydrogen breath test was associated with chronic constipation . A few Indian studies showed patients with chronic constipation (functional constipation and IBS-C) had a greater amount of breath methane than non-constipating IBS . On treatment with rifaximin, constipation improved, breath methane reduced and colon transit improved as compared to treatment with placebo . Two other studies from the USA, one prospective and the other retrospective, also supported the findings of this study . More studies are needed on this issue. Statement 7: Gastrointestinal infection with varied pathogens may result in post-infection IBS (PI-IBS) Voting summary: Accepted completely: (100%). Level of evidence: II-1 Grade of recommendation: A Of the multiple factors involved in the pathogenesis of IBS (Table ) [21,22,27-53], development of the condition following acute infectious gastroenteritis is perhaps the strongest proof of micro-organic basis of a subset of these patients. Multiple studies from all over the World showed that following GI infection with bacterial, protozoal and viral pathogens may be followed by PI-IBS . Though infection and infestations are common in developing countries, including India, the studies on the development of PI-IBS from India are scanty. Of the four available studies on this issue from the Indian authors, one was entirely performed in a population in Bangladesh , a country with similar socio-cultural factors and environment (Table ) . Two of the recent studies were done among patients with COVID-19 . All the studies showed that the development of PI-IBS was common following acute GI infection and COVID-19. More studies from India are needed on this issue. Statement 8: A diagnosis of PI-IBS by Rome criteria does not exclude underlying post-infectious malabsorption syndrome (tropical sprue) Voting summary: Accepted completely: (66.67%), accepted with minor reservation: (22.22%), rejected with major reservation: (11.11%). Level of evidence: II-2 Grade of recommendation: B Following an attack of acute infectious gastroenteritis, though 90% of patients recover completely, the remaining may continue to have chronic GI symptoms such as loose motion, abdominal pain or discomfort, bloating, etc., beyond six months fulfilling the Rome criteria for IBS . Earlier, it was reported that the patients continuing to have frequent and loose stools, on investigations, may be found to have mucosal malabsorption such as abnormal d -xylose test (denoting carbohydrate malabsorption), fecal fat (denoting fat malabsorption) and abnormal mucosal histology ; these patients improve following treatment with antibiotics and vitamin supplementation (folic acid, vitamin B 12 ). The latter condition is known as post-infectious malabsorption syndrome, popularly called tropical sprue. There is considerable overlap in the clinical presentations of the patients with PI-IBS and tropical sprue . Since hardly any study published earlier on PI-IBS investigated patients for malabsorption syndrome, the exact frequency of tropical sprue in patients with PI-IBS was not known. However, in a recent study, about 10% of patients with PI-IBS undergoing work-up were found to have tropical sprue despite fulfilling Rome criteria for IBS . In another recently published study among patients with post-COVID-19 IBS, mucosal malabsorption documented by abnormal d -xylose test alone was present in 30% and by two abnormal tests (abnormal d -xylose test and low serum vitamin B 12 ) in 4% of patients . A study from the USA also showed the occurrence of both PI-IBS and tropical sprue following acute gastroenteritis among army personnel posted in the Iraq war suggesting a common link between the two conditions . Hence, it is important to investigate patients with PI-IBS for tropical sprue. This has also been recommended by the Rome PI-IBS committee . More studies are needed on this issue. Statement 9: COVID-19 may lead to post-COVID-19 IBS Voting summary: Accepted completely: (27.78%), accepted with minor reservation: (61.11%), accepted with major reservation: (5.56%), rejected with major reservation: (5.56%). Level of evidence: II-2 Grade of recommendatio n: B Based on the shared mechanistic shreds of evidence, a possibility of the development of PI-IBS following COVID-19 has been suggested . These mechanistic shreds of evidence include the presence of angiotensin-converting enzyme-2 (ACE-2) receptors in the GI tract resulting in infection of the GI tract by the virus, the presence of severe acute respiratory syndrome coronavirus-2 (SARS-CoV-2) ribonucleic acid (RNA) in the stool of these patients, presence of GI symptoms, including diarrhea, in about 10% to 20% patients with COVID-19, raised fecal calprotectin and mucosal serotonin, macroscopic and histological evidence of GI mucosal injury, increased mucosal permeability, gut microbiota dysbiosis, involvement of the nervous system including the enteric nervous system by the virus, and increased stress due to the pandemic . In the world’s first case–control study on post-COVID-19 FGID reported from India and Bangladesh, at six months following COVID-19, IBS, dyspepsia and their overlap developed in 5.3%, 2.1%, and 1.8% patients, respectively ; these figures were significantly greater than in the control population . The presence of anosmia, ageusia, chronic bowel dysfunction, dyspeptic symptoms at one and three months and psychological comorbidity were predictors for the development of post-COVID-19 FGID (now called DGBI) . In the second study on this from Georgia, USA, of the 164 of 1114 subjects with COVID-19 who participated in follow-up study, 108 (66%) fulfilled the criteria for at least one DGBI . Such a high frequency of development of post-COVID-19 DGBI in that study might be related to recruitment bias as only 164 of 1114 subjects participated in the follow-up study. Two other studies published recently, one from India and the other multicentric (GI-COVID-19 study), showed the occurrence of post-COVID-19 FGID (now called DGBI) after the occurrence of COVID-19 . In the Indian study, the frequency of post-COVID-19 IBS during six-month follow-up was 7% . In the other multicentric study, 3.2% of hospitalized patients with COVID-19 developed IBS by Rome IV criteria during 12-month follow-up . Statement 10: Psychological and somatoform comorbidities are common in IBS Voting summary: Accepted completely: (94.44%), accepted with minor reservation: (5.56%). Level of evidence: II-2 Grade of recommendation: B Psychological comorbidities, including anxiety and depression, are common in IBS patients both globally and India . As evidenced by the voting summary, the importance of psychological factors is apparent from a cent percent agreement among experts; this is quiet close to the agreement in the second Asian IBS consensus . In a recent meta-analysis on seven case–control studies including 590 patients with IBS and 1520 controls from India, the pooled odds ratios of anxiety and depression were 8.060 (95% confidence interval [CI] 4.007–16.213) and 7.049 (95% CI 3.281–15.147) compared to controls by random effect models, respectively . There was significant heterogeneity in the included studies . Moreover, most studies were from tertiary urban centers, posing the possibility of recruitment bias overestimating the frequency. Statement 11: Dietary FODMAPs (fermentable oligo-, di-, monosaccharides and polyols) may contribute to symptoms of IBS Voting summary: Accepted completely: (88.89%), accepted with minor reservation: (11.11%). Level of evidence: I Grade of recommendation: B Mechanistically, high-FODMAP foods cause symptom exacerbation in IBS patients . A randomized trial from northern India showed that a low-FODMAP diet mitigated GI symptoms (frequency of abdominal pain, abdominal distention, bowel habit satisfaction) and improved quality of life and resulted in a significant reduction in the need for medication in IBS-D patients . Milk, which contains lactose, is one of the high-FODMAP foods. The frequency and degree of lactose malabsorption are higher in southern than in northern Indian healthy populations . Though the frequency of lactose malabsorption was comparable among patients with IBS as that among healthy controls, the former group exhibited more symptoms than the latter (47.4% compared with 15.6%; p = 0.001) . Extrapolating these results, it may be assumed that dietary FODMAPs may contribute to symptom development in more vulnerable groups such as IBS patients. The patients with IBS also had malabsorption of fructose more often by fructose hydrogen breath test, which is another component of high-FODMAP foods, than controls ([14.4% vs. 2.4%]; p = 0.04). Patients with IBS-D more often had fructose malabsorption than those with other subtypes of IBS (91% vs. 56%) . Since dietary practices in different regions of India is quite heterogeneous (prevalence of vegetarianism/non-vegetarianism, preference for wheat vs. rice as staple cereal, volumes of milk/milk product consumption), the contribution of dietary FODMAPs as a symptom driver in patients with IBS may be regionally stratified . Owing to regional food preferences and differences in vegetarianism, the northern Indian diet is highest in FODMAP content, followed by central and western Indian diets. The northeastern Indian diet is the lowest in FODMAP content, followed by the southern and eastern Indian diet, which is low in FODMAP . Thus, IBS patients in different regions in India may have a varying predisposition to symptom development due to dietary FODMAPs . The World Gastroenterology Organization provides guidelines to provide clinical practice recommendations on a low-FODMAP diet . Statement 12: Diagnosis of IBS is primarily clinical and based on well-defined symptom-based criteria Voting summary: Accepted completely: (83.33%), accepted with minor reservation: (16.67%). Level of evidence: II-2 Grade of recommendation: A The diagnosis and classification of IBS are based on symptoms, with the need for investigation only when otherwise indicated . The symptoms in patients with IBS may also overlap with those of other FGIDs (overlap disorders) and psychosocial issues, which pose a challenge to correctly diagnose and treat the condition. Diagnosis and management of these overlap disorders have been recently addressed in an Asia–Pacific Consensus . There is no specific biomarker for the diagnosis of IBS to date. IBS is diagnosed by clinical criteria by a constellation of symptoms without an organic disease explaining these symptoms . Several criteria have been formulated for the diagnosis of IBS, such as Manning, Rome (I, II, III, IV), and Asian criteria (Table ) . A multicentric Indian IBS study comparing various criteria found that Manning criteria are more sensitive for IBS diagnosis as compared to Rome I, II, and III criteria in the Indian population . The Asian criteria proposed by the Asian Neurogastroenterology and Motility Association, though were better than the Rome I, II, and III criteria, performed worse than the Manning criteria . Several studies did show that fulfilling Rome criteria were associated with a positive diagnosis of IBS without the presence of obvious organic diseases in most of the patients . A systematic review reports that among patients meeting symptom-based criteria for IBS, the probability of organic disease is less than 1% . A house-to-house survey using an enhanced Asian Rome III questionnaire, endoscopy tests and molecular genotyping techniques in the Bangladeshi population, inferred that most people suffered from functional dyspepsia (a common FGID) ; of them, 114 of 547 (20.8%) undergoing upper GI endoscopy had organic lesions that could explain their dyspeptic symptoms . Since the epidemiological profile of Indian patients with FGIDs, including IBS, is somewhat similar to the Bangladeshi population , it sounds reasonable to conclude that fulfilling symptom-based Rome criteria may be associated with an organic diagnosis in a similar frequency. Statement 13: Rome III criteria may be preferred over Rome IV to diagnose IBS in India due to its higher sensitivity Voting summary: Accepted completely: (72.22%), accepted with minor reservation: (27.78%). Level of evidence: II-2 Grade of recommendation: B In a multinational study, the prevalence of IBS was lower as per Rome IV criteria than the Rome III criteria all over the World . The frequency of IBS in the internet survey countries by Rome III and IV criteria was 10.1% vs. 4.1% and in the household survey countries was 3.5% vs. 1.5%, respectively . It is important to mention that in India, a household survey was done on the urban and rural populations in the southern and northern Indian populations. In a cross-sectional study in a primary care setting in Malaysia, the frequency of IBS was reduced from 4% by Rome III criteria to 0.8% by Rome IV criteria . Since in the Rome IV criteria for diagnosis of IBS, the frequency threshold of symptoms is based on the studies from the Western population and the presence of abdominal pain has been made mandatory, Rome IV criteria have become quite insensitive worldwide, more so in India. Abdominal pain is less frequent and severe in the Indian population ; abdominal bloating and discomfort, which are not included as IBS-defining symptoms according to Rome IV criteria, are frequent in Indian patients with IBS. The natural history and health impact of Rome IV criteria-defined IBS is more severe in comparison to that diagnosed by Rome III criteria . The severity of IBS can be assessed by subjective quantification of the symptoms, their frequency and interference with the life in general (Fig. ) . This is called IBS symptom severity scale (IBS-SSS) (Fig. ) . Rome III criteria may be preferred over Rome IV to diagnose IBS in India due to its higher sensitivity. However, Manning and Asian criteria were found to be even superior to Rome III criteria . A recent Indian study did show that IBS was less often diagnosed when Rome IV criteria rather than Rome III criteria were used (6.2% vs. 9.5%, respectively) . It appears that Rome IV criteria led to an internal shift in various diagnostic categories of FGID as there was a higher frequency of diagnosis of functional diarrhea and functional constipation than IBS on application of Rome IV criteria . Statement 14: Rome’s subtyping of IBS into constipation and diarrhea-predominant conditions frequently leads to a large proportion of patients remaining unclassified in India Voting summary: Accepted completely: (77.78%), accepted with minor reservation: (22.22%). Level of evidence: II-2 Grade of recommendation: B Application of Rome criteria concerning stool frequency and stool type to stratify IBS patients into IBS-C and IBS-D subtypes often leads to a large proportion of patients being unclassified in India . As reported by a comparison of diagnostic criteria, up to 77.6% of patients could not be classified when the Rome criteria of stool frequency for IBS-C (< 3 stools/week) were adhered to, and 15.7% of patients remained unclassified when Rome criteria of stool form for IBS-C (Bristol stool form 1, 2) were used . Stool form-based feco-graphical analysis led to 16/51 (31.4%) remaining unclassified as either IBS-C/IBS-D . Baseline clinical characteristics of “IBS only” patients in several studies show that a significant proportion of them remain unclassified: 62/75 (82.7%) , 7/47 (14%) , 17/70 (25%) , 16/160 (10%) . The proportion of patients remaining unclassified was also high in IBS overlapping with other FGIDs: IBS-dyspepsia overlaps 72/115 (62.6%) . Comparatively, the proportion remaining unclassified in IBS patients was more by Rome III (9/124 [7.3%]) than by Rome IV (1/81 [1.2%]) criteria . Statement 15: Bristol stool form scale (BSFS) along with other symptoms such as straining, incomplete evacuation, urgency and patient-reported bowel pattern should be assessed to evaluate bowel movement Voting summary: Accepted completely: (94.44%), accepted with minor reservation: (5.56%). Level of evidence: II-2 Grade of recommendation: B In Asian consensus, Bristol stool forms 1–3 have been defined as constipation and 5–7 as diarrhea. Sub-classification of IBS, according to the bowel pattern, could be based on a modification of the Rome III criteria . The BSFS demonstrated substantial validity and reliability . The Asian consensus on IBS suggested that in addition to Bristol types 1 and 2 stool, type 3 stool should also be considered as constipation in Asia . In a multicentre study from India, improvement in subtyping IBS using BSFS as suggested in the Asian consensus has been reported ; in this study, applying stool types 3 (as hard stool) and 5 (as a soft stool) as abnormal stool forms allowed more patients to be subtyped compared to the use of Rome subtyping system that considers type 3 to type 5 stools as normal stools . Other constipation-associated symptoms such as straining, feeling of incomplete evacuation, patients’ perception and infrequent bowel movements (less than three bowel movements per week) should also be considered while sub-classifying IBS. Statement 16. A: In the absence of alarm features, a few baseline investigations are suggested for patients with suspected IBS. Statement 16. B: The presence of alarm features necessitates diagnostic tests to rule out organic disease. Voting summary: Accepted completely: (100%) Level of evidence: II-2 Grade of recommendation: B Though the yield of investigating patients to exclude organic disease after a criteria-based diagnosis of IBS in the West has been low, due to a lack of high-quality evidence from India and frequent occurrence of GI infection and infestation, a few routine investigations to exclude organic disorders and infection and infestations may be recommended at present (Fig. ). In the presence of alarm features, which include the onset of symptoms at an age older than 45 years, presence of anemia, blood in the stool, unintended weight loss, nocturnal symptoms, fever, abdominal mass and family history of colorectal cancer, thorough investigations to rule out organic disease are essential. However, it is well recognized that alarm features per se have a low yield of organic disease . However, by consensus, guidelines have adopted these symptoms to focus and fast-track tests for such patients, in systems where healthcare has long waiting times. There is one large cross-section study on this issue . Several studies from other parts of the World including the one mentioned above have looked at the only prevalence of organic disease in those labeled as IBS . A meta-analysis of 28,630 patients with possible IBS who underwent colonoscopy showed a pooled prevalence of colorectal cancer, inflammatory bowel disease and microscopic colitis to be 0.78%, 4.48%, and 2.35%, respectively . The second Asian consensus on IBS also suggested that the shreds of evidence supporting the statement on the predictive value of alarm symptoms to exclude organic disorders are somewhat low . Most studies from India on predicting the utility of alarm symptoms to suggest organic diseases are not undertaken among patients with IBS but have been done on another common FGID, namely functional dyspepsia. In a study from southern India on 900 patients with dyspepsia, upper GI endoscopy revealed benign organic lesions in 38% and malignancy in 5.5% . The authors concluded that the optimal age to begin screening for malignancy in dyspepsia patients in India seems to be 46.5 years . In another recent study from northern India on 294 patients with dyspepsia, the authors found that even among patients younger than 60 years with alarm features, a considerable number of patients had organic lesions (65%) including malignancy (16%) on upper GI endoscopy . These authors chose a cut-off value of 60 years for age as the American College of Gastroenterology (ACG) and Canadian Association of Gastroenterology (CAG) guidelines did not recommend upper GI endoscopy to investigate alarm features for dyspepsia patients under the age of 60 to exclude upper GI malignancy . The cut-off value of some of the alarm features such as age may differ in different populations. There is no Indian study on this issue concerning IBS. As the Asian consensus on colorectal cancer screening suggested that screening for colorectal cancer should start at the age of 50 in Asia , IBS patients in Asia including India may undergo a colonoscopic examination if they are 50 years or older, particularly in the presence of alarm symptoms. However, more studies are needed from India on this issue before a firm recommendation can be made. Statement 17: The multi-dimensional clinical profile of IBS needs to be incorporated into clinical practice as proposed in the Rome IV algorithm Voting summary: Accepted completely: (77.78%), accepted with minor reservation: (22.22%). Level of evidence: II-2 Grade of recommendation: B IBS is not a single disease but a syndrome with a constellation of multiple GI and non-GI symptoms, which might result from several pathophysiological mechanisms in variable combinations either in the gut or in the brain . There are growing shreds of evidence to suggest a greater role of gut abnormalities than central pathophysiological mechanisms in the genesis of IBS symptoms . Accordingly, the experts in the Rome IV Committee recommended underplaying the term “functional” to denote these disorders and re-name these conditions as the “Disorders of gut-brain interaction (DGBI)” . It is important to note that in this new terminology that replaces FGID to denote these conditions, “gut” has been kept over “brain” to recognize the greater importance of gut-level pathophysiological mechanisms in the genesis of symptoms . The Asian experts called these pathophysiological mechanisms micro-organic factors. Micro-organic factors such as slow colon transit and fecal evacuation disorder may be hidden behind the diagnosis of a patient with IBS-C. Similarly, IBS-D patients may have dietary intolerance, including that of lactose and fructose, bile acid malabsorption, non-celiac wheat sensitivity, SIBO and GI infection . A recent study showed that multi-modality care targeting various pathophysiological mechanisms of IBS is superior to gastroenterologist-provided standard care . Hence, unrevealing these pathophysiological mechanisms in each patient through thorough history taking, physical examination and investigation is a step towards personalized care of the patients with IBS. Multi-dimensional clinical profile (MDCP) is a step toward understanding the multiple factors contributing to symptom generation in patients with IBS (Table ) . This should be followed in clinical practice. MDCP assessment necessitates evaluation for the severity of IBS. Of different methods of severity assessment, IBS-SSS has been commonly used (Fig. ) . However, it has not yet been validated in Indian patients with IBS. Statement 18: Patients with refractory IBS symptoms need further pathophysiological evaluation Voting summary: Accepted completely: (77.78%), accepted with minor reservation: (22.22%). Level of evidence: II-1 Grade of recommendation: A IBS is a syndrome in which the symptoms are contributed by multiple pathophysiological mechanisms . Some of these pathophysiological factors differ in different subtypes of IBS. For example, in patients with IBS-C, the underlying physiological abnormalities may include, slow colon transit, fecal evacuation disorder or a variable combination of these two factors. Slow colon transit may result from excess methane production in the gut due to methanogen overgrowth. In contrast, the patients with non-C-IBS may have lactose and fructose malabsorption, bile acid malabsorption, non-celiac wheat and FODMAP intolerance, gut microbiota dysbiosis including SIBO, immune activation, and post-infection including post-COVID-19 etiology. These are, however, not watertight compartments; IBS-C patients may also have some of the pathophysiological factors listed under IBS-D and vice versa. Table summarizes the major Indian studies substantiating these factors contributing to IBS . Whereas the Asian workers suggested that IBS is a micro-organic condition to highlight the importance of these factors in the pathogenesis and management of the condition , the Rome IV experts brought in MDCP to give due importance to these factors (Table ) . Statement 19: Counseling, reassurance and lifestyle modification are important in the management of IBS Voting summary: Accepted completely: (88.89%), accepted with minor reservation: (11.11%). Level of evidence: II-2 Grade of recommendation: B The urban lifestyle associated with fad diets (junk, fast food, fatty food, tea, coffee, aerated drinks, etc.), substance abuse (smoking, chewing tobacco), low levels of physical activity, and psychological stress (depression, anxiety, insomnia) has been associated with higher prevalence of IBS . To inculcate lifestyle balance, remedial measures such as frequent counseling sessions and promoting physical activity (e.g. yoga, meditation) may help. Several randomized and non-randomized studies showed physical exercise, including yoga, is useful in the management of IBS . A few studies compared yoga with either anti-anxiety medications or a low-FODMAP diet and found that yoga was either as good as anti-anxiety medications or was better than a low-FODMAP diet . Sleep disorders, common in IBS patients, should be recognized and appropriately treated with pharmacotherapy . Melatonin has been found useful to treat sleep dysfunction . Drugs with addictive potential such as benzodiazepines should be discouraged in a chronic condition like IBS. Several randomized controlled trials showed the benefit of a low-FODMAP diet in the management of IBS . However, the studies on low-FODMAP diet in India are scanty and are limited to one study on IBS and another on functional dyspepsia . Though a recent review attempted to educate physicians about low-FODMAP diet in IBS , it may be somewhat premature in the bsence of more studies on efficacy and challenges of such treatment in India considering the high frequency of vegetarianism in the population . An Australian study showed that multi-modality care involving a psychologist and dietician in addition to a gastroenterologist is superior to the standard gastroenterologist-provided treatment of IBS . To ensure adequate compliance with lifestyle measures and dietary practices, patient education and counseling are the keys to success. Statement 20: The initial treatment of IBS is primarily symptom based Voting summary: Accepted completely: (94.44%), accepted with minor reservation: (5.56%). Level of Evidence: I Grade of Recommendation: A Since IBS is a syndrome diagnosed by symptom-based criteria, the goal of treatment of IBS is also to relieve patients’ symptoms and improve the quality of life (QOL). Currently, even the outcome of treatment of IBS also revolves around patient-reported outcome measures. Establishing a good doctor-patient relationship is an important art in the management of these patients. All bothersome symptoms should be targeted while treating these patients rather than only the predominant symptoms as suggested earlier taking into account specific IBS subtypes, symptom severity and contributing factors including psychosocial and dietary issues . IBS patients with low symptom load do find significant improvement in their symptoms by use of antispasmodics , bulking agents, antidiarrheals , pro-motility agents including laxatives, and low-FODMAP diet . An outline of the management of IBS is summarized in Fig. . Tables and list the drugs and their dosages used in different subtypes of IBS and some of the Indian studies. Statement 21: Antispasmodics are the first-line treatment of abdominal pain in patients with IBS and non-responsive patients may benefit from visceral neuromodulators Voting summary: Accepted completely: (94.44%), accepted with minor reservation: (5.56%). Level of evidence: I Grade of recommendation: A Several randomized controlled trials from Asia and meta-analyses showed antispasmodic drugs to be effective in the treatment of IBS, particularly when abdominal pain is a major symptom . These drugs also benefit exaggerated gastrocolic reflexes. The number needed to treat and harm with these drugs are quite good (5 and 17.5, respectively) . However, there is hardly any good-quality study from India on antispasmodic drugs in the management of IBS. In a randomized controlled trial on 200 patients with Rome III criteria-diagnosed IBS, the authors found that drotaverine was superior to mebeverine in alleviating pain and stools-related symptoms . It is important to mention that though some of the antispasmodic drugs such as anticholinergics are cheap, these are associated with significant adverse effects such as dry mouth and airway, visual blurring, palpitation, urinary dysfunction, constipation, and precipitation of angle-closure glaucoma. Hence, these drugs must be used with caution. Table summarizes some of the studies from India on different antispasmodic drugs in the management of IBS . Various psychoactive pharmacotherapy (currently called visceral neuromodulators) are useful to relieve abdominal pain in IBS patients working both through central and peripheral mechanisms . These drugs may even work in the absence of significant psychological comorbidity . Statement 22: Laxatives and antidiarrheals are the first-line treatment for IBS-C and IBS-D, respectively Voting summary: Accepted completely: (94.44%), accepted with minor reservation: (5.56%). Level of evidence: I Grade of recommendation: A Bulking agents, including fibers, are quite popular to treat patients with IBS-C. However, there are inadequate high-quality studies to evaluate their efficacy. The fibers could be either water-soluble (e.g., psyllium, ispaghula, calcium polycarbophil, methylcellulose) or insoluble (e.g., corn, wheat bran). In an Indian dose-finding study, the authors found that the optimum dose of ispaghula husk, which is the husk of Plantago ovata seeds, was 20 g per day and it improved constipation and abdominal pain in IBS patients . In a study from India on 20 patients, the authors concluded that the easing of bowel dissatisfaction appeared to be a major reason for the therapeutic success of ispaghula in IBS . However, the fibers, particularly the insoluble ones, may aggravate abdominal bloating and flatulence . The other laxatives that may cause bloating and flatulence include lactulose and to some extent lactitol. Milk of magnesia and polyethylene glycol may cause minimum or no bloating. The stimulant drugs such as senna and bisacodyl are quite effective purgatives and do not cause abdominal bloating; these may cause abdominal cramps . Loperamide and diphenoxylate are quite popular in the treatment of IBS-D. Several studies proved the efficacy of loperamide in the management of diarrhea but not abdominal pain or distension . Other medicines that may be useful in patients with IBS-D include ramosetron and visceral neuromodulators with anticholinergic activity such as amitriptyline. Table lists the drugs useful in treatment of different subtypes of IBS. Statement 23: Dietary FODMAP restriction is useful in a proportion of IBS patients Voting summary: Accepted completely: (94.44%), accepted with minor reservation: (5.56%), accepted with major reservation: (5.56%). Level of evidence: I Grade of recommendation: A Dietary intolerance contributes to symptoms in patients with IBS . A low-FODMAP diet in the management of IBS was first popularized in Australia. Subsequently, several randomized controlled trials and their meta-analysis proved the efficacy of a low-FODMAP diet in the management of IBS . A low-FODMAP diet is particularly useful for flatulence, bloating, abdominal pain, distension and diarrhea . However, studies on low-FODMAP diet in Indian patients with IBS are scanty. A randomized controlled trial on 166 patients showed that a strict low-FODMAP diet for the short-term and a modified low-FODMAP diet for the long-term may lead to symptom improvement in IBS-D patients . Another randomized controlled study from the same group showed the efficacy of a low-FODMAP diet in another common form of FGID called functional dyspepsia . However, considering the regional difference in dietary practices in India, there is a need for more studies to evaluate the efficacy, difficulty and patient acceptability of a low-FODMAP diet in different parts of India. The dietary variations, which may contribute to the challenges in the implementation of low-FODMAP diet in different regions of the country, may include the frequency of vegetarianism and lactose malabsorption, intake of milk in the diet, and the types of cereals commonly consumed (wheat vs. rice) . In a recently published analysis of available data, it has been shown that implementing a low-FODMAP diet in northeastern India may be most easy compared to northern India, where it may be most difficult. Implementing a low-FODMAP diet in southern and eastern India may be easy, moderately easy in western India and not easy in central India . Statement 24: Probiotics may be helpful but more studies are needed Voting summary: Accepted completely: (72.22%), accepted with minor reservation: (16.67%), accepted with major reservation: (11.11%). Level of evidence: II-2 Grade of recommendation: B Since gut microbiota dysbiosis and SIBO are associated with IBS, drugs manipulating gut microbiota such as antibiotics (rifaximin) and probiotics have been evaluated in its treatment. Another form of manipulation of gut microbiota using fecal transplantation is not recommended to treat IBS at present, but this therapy is only undertaken in a research setting. In an earlier consensus, the review of pieces of evidence suggested that Lactobacillus strains ameliorated flatulence significantly and abdominal or global symptom scores were improved by Bifidobacterium and Escherichia and Streptococcus strains both provided persistent symptom reduction . However, the current evidences are not enough to achieve a strong consensus on the use of specific probiotics as a routine practice and its dose as the therapy of choice. More studies are needed on this issue from India. Statement 25: Non-constipated and methane-producing IBS-C patients benefit from rifaximin Voting summary: Accepted completely: (55.56%), accepted with minor reservation: (33.33%), accepted with major reservation: (5.56%), reject with major reservation: (5.56%). Level of evidence: I Grade of recommendation: A Several studies and meta-analyses showed that high-breath methane on lactulose hydrogen breath test was associated with constipation, due to slow colon transit . One retrospective and one prospective study from the USA showed that reduction in breath methane with antibiotic treatment improved constipation and a combination of neomycin and rifaximin was superior to rifaximin alone . A randomized controlled trial showed rifaximin was superior to a placebo in patients with slow transit constipation with excess breath methane on lactulose hydrogen breath test in association with the reduction in the breath methane and acceleration of colon transit . In the previous Indian consensus on chronic constipation, the efficacy of rifaximin treatment for high-breath methane-producing constipation has been accepted . Rifaximin has been proven to be effective in the treatment of non-C-IBS in a large clinical trial, which led to its approval by the Food and Drug Administration in the USA . Statement 26: Pathophysiology-directed therapy is important in the management of refractory IBS Voting summary: Accepted completely: (83.33%), accepted with minor reservation: (11.11%), accepted with major reservation: (5.56%) Level of evidence: I Grade of recommendation: B There are controversies in the definition of refractory IBS . Refractory IBS patients either fail to respond appropriately to standard pharmacological treatment or show no improvement with conventional pharmacological interventions and continue to have severe symptoms. In the current consensus, if the symptoms persist or increase in severity even during 12 weeks of follow-up, the condition may be considered clinically refractory, albeit empirically. These patients should be referred to a center with facilities for further specialized investigations and multi-modality care not only by the gastroenterologist alone but by a GI disorder experienced psychologist/psychiatrist and dietician. These patients should be assessed clinically for symptom duration, symptom severity (IBS-SSS) (Figs. and ), symptom type, dietary practices and psychological issues. The stubbornness of refractory IBS may be explained by multi-dimensional pathophysiological mechanisms, including altered gut motility, intestinal barrier dysfunction, gut microbiota dysbiosis, including SIBO, gut immune dysfunction, visceral hypersensitivity, bile acid malabsorption, FODMAP sensitivity, psychological factors and altered gut-brain interactions . Investigations directed to these pathophysiological mechanisms, based on the local availability, should be undertaken (Figs. and ). Both pharmacological and non-pharmacological treatment directed to each of the pathophysiological mechanisms, even if needed combination treatment is recommended to treat patients with refractory IBS as reviewed previously . Pharmacological (Table ) and non-pharmacological interventions (e.g. cognitive behavioral therapy, gut-directed hypnotherapy, mindfulness-based stress reduction therapy [MBSRT]) directed against abnormal gut-brain interaction using visceral neuromodulators and various forms of psychotherapies have been found useful in the treatment of refractory IBS (Fig. ). A multi-modality care involving a psychologist and dietician in addition to a gastroenterologist is expected to be superior to the standard gastroenterologist-provided treatment of refractory IBS . Statement 27: Psychological interventions are useful in those with psychiatric comorbidities or refractory IBS Voting summary: Accepted completely: (83.33%), accepted with minor reservation: (11.11%), rejected with minor reservation: (5.56%) Level of evidence: II-1 Grade of recommendation: B Psychological comorbidity is common among IBS patients, particularly those presenting to tertiary care facilities . Patients with severe symptoms, non-responding patients and overlap disorders more often have psychological comorbidity . Psychological interventions either with pharmacotherapy or cognitive behavior therapy, gut-directed hypnotherapy, yoga and MBSRT are useful options to treat these patients in addition to standard treatment . It is important to mention here that while using different visceral neuromodulator/ psychoactive compounds, one should use those with prominent anticholinergic side effects in IBS-D patients and serotonin reuptake inhibitors (SSRIs) in IBS-C patients for additional benefit towards amelioration of GI symptoms (Table ). Published experience in non-pharmacological psychological intervention on IBS in India is limited. These interventions are also not available widely. In a non-randomized open-label controlled study on MBSRT on 47 IBS patients (30 on MBSRT and 17 pharmacological treatment alone), MBSRT led to a greater improvement in the quality of life and mindfulness components and reduced IBS symptoms as compared with the control group . One Indian study also showed the efficacy of yoga in the treatment of IBS . Statement 28: Overlapping functional GI disorders may require combination treatment Voting summary: Accepted completely: (94.44%), accepted with minor reservation: (5.56%), Level of evidence: II-2 Grade of recommendation: B IBS may occur concomitantly with other FGIDs, especially, IBS-functional dyspepsia overlap and IBS-GERD overlap . Table presents the IBS-FGID overlap in studies conducted entirely or partly in India. A recent Asia Pacific Consensus reviewed the clinical spectrum, pathophysiology and the management of overlapping FGIDs in details. The authors in this consensus did suggest either pathophysiology-guided or symptom-directed combination treatment to manage these patients. However, it is important to remember for possible drug interaction while combining different drugs in the management of IBS.
|
Diversity and Quantitative Detection of Clade I Type | 87bcb68e-102e-4eab-a01d-21b68b1df1b8 | 10037096 | Microbiology[mh] | Water sampling Water samples were collected from the ASTS location using Niskin bottles on a CTD rosette from the surface (5 m), deep chlorophyll maxima (DCM) (~43–50 m), core OMZ depths (250 and 500 m), and 1,000 m during three seasons i.e. the Spring intermonsoon (SIM), Fall intermonsoon (FIM), and Northeast monsoon (NEM). Water was processed through sterivex filters (0.22 μm; Millipore), filled with 1.7 mL of storage buffer (50 Mm, Tris pH 8.3, 40 mM EDTA, and 0.75 M sucrose) and stored at –80°C until DNA extraction. Measurements of physicochemical parameters Physicochemical parameters (depth, temperature, salinity, and pH) in every water sample were measured using different sensors mounted onto the CTD rosette. The standard Winkler titration method ( ) was used to measure dissolved oxygen (DO). Nutrient (ammonia, nitrate, nitrite, phosphate, and silicate) concentrations were assessed according to previously reported methods ( ). DNA extraction and PCR amplification Nucleic acid extraction from water samples was performed according to standard methods ( ). The amplification of DNA samples was conducted using the primer set nosZ1F (5′-WCSYTGTTCMTCGACAGCCAG-3′) and nosZ1R (5′-ATGTCGATCARCTGVKCRTTYTC-3′) ( ). The primers nosZ1F and nosZ1R targeted Clade I genes; Clade II genes were not amplified separately. PCR was performed using a Thermocycler machine (Applied Biosystem) following temperature conditions of 95°C for 5 min, for initial denaturation, 30 cycles at 95°C for 30 s, annealing at 60°C for 45 s, and a final extension step at 72°C for 90 s. A negative control (PCR mix and primers) was used in each PCR reaction and amplification was confirmed by agarose gel (1%) electrophoresis. Cloning and sequencing NosZ gene amplicons were purified, cloned into the PGEM-T Easy Vector (Promega), transformed inside high efficiency JM109 cells, and grown overnight at 37°C on LB/X-gal/IPTG plates. A minimum of 40 clones were selected per plate for colony PCR. Temperature conditions for colony PCR were as follows: an initial denaturation step at 94°C for 10 min, followed by 30 cycles at 94°C for 1 min, annealing at 55°C for 1 min with an elongation step at 72°C for 1 min, and a final extension at 72°C for 10 mins. PCR products were purified, measured, and sequenced using 15–50 ng of amplicons, adding 10 pmol each of the nosZ1F and nosZ1R primers in an ABI 3130 Genetic Analyzer (Applied Biosystems). Temperature conditions for sequencing were an initial denaturation at 96°C for 1 min, 30 cycles at 96°C for 10 s, annealing at 60°C for 45 s, elongation at 60°C for 4 mins, and a final extension at 60°C for 1 min. Sequence analysis NosZ gene sequences were assembled using DNA Baser sequence assembly software version 2. VecScreen was used to eliminate vector contamination. Non-chimeric consensus sequences without vector and primer residues were submitted to the National Center for Biotechnology Information (NCBI) database to obtain accession numbers and were used in additional analyses. Sequences were classified using 1,000 pseudo-bootstrap replications at a bootstrap value of 80% (standard error of only 1.3%). Alignments were trimmed using Gblocks ( ). Clone sequences were compared with the NCBI database and assigned to a phylum if their identity was more than 95%. OTU assessment and phylogenetic tree Sequences ( nosZ gene) were assigned to operational taxonomic units (OTUs) by the average neighbor rule ( ) using MOTHUR. Sequences obtained from all seasons from depths of 250 and 500 m were grouped as nosZ250m and nosZ500m, respectively, and all nosZ sequences obtained from 250 and 500 m were combined as CnosZ. OTUs were obtained at a sequence similarity of 95%, and representative sequences for each OTU of nosZ250m, nosZ500m, CnosZ, and top-hit nucleotide sequences from cultivated known strains from the NCBI database were used to build the phylogenetic tree (MEGA 6.0) with 1,000 replicate bootstrap analyses. qPCR assay Quantification of the nosZ gene was performed using the ABI 7500 Real-Time PCR system (Applied Biosystems). A plasmid carrying the nosZ gene fragment was cloned using the pCR4-TOPO vector and confirmed by sequencing. Ten-fold serial dilutions of a known copy number of plasmid DNA were subjected to a qPCR assay in triplicate to generate a standard curve and calculate the qPCR efficiency of the nosZ gene. A standard curve was generated by plotting threshold cycle values versus log10 of gene copy numbers. The slope, y-intercept, and coefficient of determination (r2) were assessed. The efficiency of amplification (E) was calculated using the following equation: E=–1+10 (–1/slope) . The abundance of the nosZ gene was quantified in triplicate. Each reaction contained a mixture of DNA (4 μL), the primer pair nosZ1F/nosZ1R (0.5 μL), and 5× qARTA Green qPCR Mix (12.5 μL). PCR cycles were performed according to the standard protocol ( ). The copy number of the target gene was calculated directly against the standard curve. The negative control had higher Ct values (~9 cycles) than the most diluted plasmid containing the target gene. Additionally, qPCR products were cloned and sequenced to confirm the identity of the gene. A post-amplification melting curve analysis showed that there was no target gene contamination in the reagents. Statistical analysis Spearman’s correlation coefficient was used to evaluate the relationships between physicochemical parameters (DO, TOC, NO 2 – NO 3 – , and NH 4 + ) and gene copy numbers for every individual season. One-way analysis of variance (ANOVA) was performed for each physicochemical parameter, OTU, and copy number. Paired differences between depths within each season were tested using Tukey’s post hoc tests. Principal component analysis (PCA) with varimax rotations for the above-described physicochemical parameters was performed to reduce the number of inter-correlated variables. Multiple regression analyses were conducted to investigate the relationships between OTUs, copy numbers (dependent variables), and principal component scores (predictor variables) obtained from PCA. Statistical analyses were conducted using IBM © SPSS 23.0. Each result was shown as the mean±standard deviation (SD). Rarefaction curve, diversity indices (Shannon’s and Simpson’s), and richness estimators (Chao 1 and ACE) were evaluated using MOTHUR. Non-parametric richness estimators were used to extrapolate the total richness of clone libraries from the observed number of OTUs. Diversity and richness estimators were calculated for individual clone libraries. NCBI Accession numbers NCBI accession numbers for the 171 nosZ gene sequences obtained in the present study are KX784867 to KX784885, KX911214 to KX911243, KY065372 to KY065445, and KY100043 to KY100090.
Water samples were collected from the ASTS location using Niskin bottles on a CTD rosette from the surface (5 m), deep chlorophyll maxima (DCM) (~43–50 m), core OMZ depths (250 and 500 m), and 1,000 m during three seasons i.e. the Spring intermonsoon (SIM), Fall intermonsoon (FIM), and Northeast monsoon (NEM). Water was processed through sterivex filters (0.22 μm; Millipore), filled with 1.7 mL of storage buffer (50 Mm, Tris pH 8.3, 40 mM EDTA, and 0.75 M sucrose) and stored at –80°C until DNA extraction.
Physicochemical parameters (depth, temperature, salinity, and pH) in every water sample were measured using different sensors mounted onto the CTD rosette. The standard Winkler titration method ( ) was used to measure dissolved oxygen (DO). Nutrient (ammonia, nitrate, nitrite, phosphate, and silicate) concentrations were assessed according to previously reported methods ( ).
Nucleic acid extraction from water samples was performed according to standard methods ( ). The amplification of DNA samples was conducted using the primer set nosZ1F (5′-WCSYTGTTCMTCGACAGCCAG-3′) and nosZ1R (5′-ATGTCGATCARCTGVKCRTTYTC-3′) ( ). The primers nosZ1F and nosZ1R targeted Clade I genes; Clade II genes were not amplified separately. PCR was performed using a Thermocycler machine (Applied Biosystem) following temperature conditions of 95°C for 5 min, for initial denaturation, 30 cycles at 95°C for 30 s, annealing at 60°C for 45 s, and a final extension step at 72°C for 90 s. A negative control (PCR mix and primers) was used in each PCR reaction and amplification was confirmed by agarose gel (1%) electrophoresis.
NosZ gene amplicons were purified, cloned into the PGEM-T Easy Vector (Promega), transformed inside high efficiency JM109 cells, and grown overnight at 37°C on LB/X-gal/IPTG plates. A minimum of 40 clones were selected per plate for colony PCR. Temperature conditions for colony PCR were as follows: an initial denaturation step at 94°C for 10 min, followed by 30 cycles at 94°C for 1 min, annealing at 55°C for 1 min with an elongation step at 72°C for 1 min, and a final extension at 72°C for 10 mins. PCR products were purified, measured, and sequenced using 15–50 ng of amplicons, adding 10 pmol each of the nosZ1F and nosZ1R primers in an ABI 3130 Genetic Analyzer (Applied Biosystems). Temperature conditions for sequencing were an initial denaturation at 96°C for 1 min, 30 cycles at 96°C for 10 s, annealing at 60°C for 45 s, elongation at 60°C for 4 mins, and a final extension at 60°C for 1 min.
NosZ gene sequences were assembled using DNA Baser sequence assembly software version 2. VecScreen was used to eliminate vector contamination. Non-chimeric consensus sequences without vector and primer residues were submitted to the National Center for Biotechnology Information (NCBI) database to obtain accession numbers and were used in additional analyses. Sequences were classified using 1,000 pseudo-bootstrap replications at a bootstrap value of 80% (standard error of only 1.3%). Alignments were trimmed using Gblocks ( ). Clone sequences were compared with the NCBI database and assigned to a phylum if their identity was more than 95%.
Sequences ( nosZ gene) were assigned to operational taxonomic units (OTUs) by the average neighbor rule ( ) using MOTHUR. Sequences obtained from all seasons from depths of 250 and 500 m were grouped as nosZ250m and nosZ500m, respectively, and all nosZ sequences obtained from 250 and 500 m were combined as CnosZ. OTUs were obtained at a sequence similarity of 95%, and representative sequences for each OTU of nosZ250m, nosZ500m, CnosZ, and top-hit nucleotide sequences from cultivated known strains from the NCBI database were used to build the phylogenetic tree (MEGA 6.0) with 1,000 replicate bootstrap analyses.
Quantification of the nosZ gene was performed using the ABI 7500 Real-Time PCR system (Applied Biosystems). A plasmid carrying the nosZ gene fragment was cloned using the pCR4-TOPO vector and confirmed by sequencing. Ten-fold serial dilutions of a known copy number of plasmid DNA were subjected to a qPCR assay in triplicate to generate a standard curve and calculate the qPCR efficiency of the nosZ gene. A standard curve was generated by plotting threshold cycle values versus log10 of gene copy numbers. The slope, y-intercept, and coefficient of determination (r2) were assessed. The efficiency of amplification (E) was calculated using the following equation: E=–1+10 (–1/slope) . The abundance of the nosZ gene was quantified in triplicate. Each reaction contained a mixture of DNA (4 μL), the primer pair nosZ1F/nosZ1R (0.5 μL), and 5× qARTA Green qPCR Mix (12.5 μL). PCR cycles were performed according to the standard protocol ( ). The copy number of the target gene was calculated directly against the standard curve. The negative control had higher Ct values (~9 cycles) than the most diluted plasmid containing the target gene. Additionally, qPCR products were cloned and sequenced to confirm the identity of the gene. A post-amplification melting curve analysis showed that there was no target gene contamination in the reagents.
Spearman’s correlation coefficient was used to evaluate the relationships between physicochemical parameters (DO, TOC, NO 2 – NO 3 – , and NH 4 + ) and gene copy numbers for every individual season. One-way analysis of variance (ANOVA) was performed for each physicochemical parameter, OTU, and copy number. Paired differences between depths within each season were tested using Tukey’s post hoc tests. Principal component analysis (PCA) with varimax rotations for the above-described physicochemical parameters was performed to reduce the number of inter-correlated variables. Multiple regression analyses were conducted to investigate the relationships between OTUs, copy numbers (dependent variables), and principal component scores (predictor variables) obtained from PCA. Statistical analyses were conducted using IBM © SPSS 23.0. Each result was shown as the mean±standard deviation (SD). Rarefaction curve, diversity indices (Shannon’s and Simpson’s), and richness estimators (Chao 1 and ACE) were evaluated using MOTHUR. Non-parametric richness estimators were used to extrapolate the total richness of clone libraries from the observed number of OTUs. Diversity and richness estimators were calculated for individual clone libraries.
NCBI accession numbers for the 171 nosZ gene sequences obtained in the present study are KX784867 to KX784885, KX911214 to KX911243, KY065372 to KY065445, and KY100043 to KY100090.
Hydrography Temperature, salinity, pH, and total organic carbon (TOC) were consistent, while the concentrations of DO and nutrients varied at core OMZ depths ( ). The average DO concentration decreased from 15.73 μmol L –1 at 250 m to 5.85 μmol L –1 at 500 m. The concentration of nitrate was higher at core OMZ depths during all three seasons and ranged between 15.33 to 32.79 μmol L –1 . The highest concentrations of nitrite (2.57 μmol L –1 ) and ammonia (NH 4 + 1.01 μmol L –1 ) were noted at 250 and 500 m during the NEM and SIM, respectively. The concentrations of DO, NO 2 – NO 3 – , and NH 4 + during all three seasons are shown in . Phylogenetic analyses Evolutionary differences between OTUs and representative sequences were assessed by phylogenetic analyses. PCR amplification of the nosZ gene (259-bp amplification product) was only positive for samples collected at 250 and 500 m in all seasons. Among the 210 clones sequenced, 171 non-chimeric nosZ gene sequences were obtained and the following clone libraries were built: SIM-250, SIM-500, FIM-250, FIM-500, NEM-250, and NEM-500. Phylogeny of nosZ denitrifiers The sequences of nosZ clones showed 80–93% identity with one another and 77–88% similarity to sequences in GeneBank. All OTUs obtained in the present study matched the sequences of the phylum Proteobacteria in the NCBI database. Taxonomic analyses reveals that most nosZ OTUs were associated with Pseudomonas , Rhodopseudomonas , Bradyrhizobium , and Alphaproteobacteria . A small percentage of bacteria was affiliated with Azospirillum , Achromobacter , Cupriavidus , Nisaea , Thalassobaculum , Sinorhizobium , Herbaspirillum , Burkholderia , and Alcaligenes . The phylogenetic analysis showed that OTUs in the present study closely matched the environmental sequences obtained from the suboxic zone of the AS, deep-sea waters of the Mediterranean Sea, the crop soil and wetland sediment of Mexico, terrestrial subsurface sediments, a marine aquaculture biofilter, and paddy soil. Among the 87 clones sequenced from nosZ250m, 17 OTUs were obtained ( ), the majority of which were affiliated with the genera Rhodopseudomonas (18%, 3 OTUs) and Pseudomonas (18%, 3 OTUs). Azospirillum (2 OTUs), Bradyrhizobium (2 OTUs), Achromobacter (2 OTUs), and Cupriavidus (2 OTUs) each contributed 12% of the total number of OTUs. Bacterial sequences belonging to the genera Nisaea (1 OTU), Thalassobaculum (1 OTU), and class Alphaproteobacteria (1 OTU) formed OTUs with a smaller number of clones. Thirteen OTUs were generated from nosZ500m (84 sequences) ( ), the majority of which belonged to the genera Bradyrhizobium (15%, 2 OTUs), Pseudomonas (15%, 2 OTUs), and Alphaproteobacteria (15%, 2 OTUs). Thalassobaculum (1 OTU), Rhodopseudomonas (1 OTU), Sinorhizobium (1 OTU), Azospirillum (1 OTU), Herbaspirillum (1 OTU), Burkholderia (1 OTU), and Alcaligenes (1 OTU) were the other minor OTUs present. Among the 171 clones sequenced from core OMZ depths, 27 OTUs were generated, which were affiliated with the classes Alpha- , Beta- , and Gammaproteobacteria ( ). The majority of OTUs were affiliated with the genera Bradyrhizobium (5 OTUs, 18.5%, 32 sequences), Azospirillum (5 OTUs, 18.5%, 32 sequences), and Pseudomonas (4 OTUs, 14.8%, 26 sequences). Three OTUs (11.1%) each were associated with Rhodopseudomonas (19 sequences), Sinorhizobium (19 sequences), and Achromobacter (19 sequences). One OTU (6 sequences) each was affiliated with the genera Thalassobaculum , Nisaea , and Burkholderia and the class Alphaproteobacteria ( ). Season- and depth-wise distribution of nosZ denitrifiers The seasonal distribution patterns of nosZ gene sequences at the class level from 250 and 500 m showed that the maximum number of sequences was affiliated with the class Alphaproteobacteria . At 250 m, Alphaproteobacteria , Oscillatoriophycideae , Bacilli , and Actinobacteria were detected during all three seasons. Acidobacteria was found during the SIM and FIM, Chloroflexia only during the SIM, and Clostridia and Cytophagia only during the NEM. The maximum number of classes was detected in the SIM (7), followed by the FIM (5) and NEM (5) ( ). At 500 m, Alphaproteobacteria was predominant, followed by Bacilli , Actinobacteria , and Oscillatoriophycideae . Clostridia and Mollicutes were the other classes present at small percentages. No significant variations among classes were observed between seasons, except for Clostridia and Mollicutes , which were only found during the NEM ( ). The number of classes detected at 250 m (8) was higher than that detected at 500 m (6). Abundance of the nosZ gene Standard curves for the nosZ gene were plotted and qPCR efficiency (81.93%) was calculated. The efficiency reading was used as a reference to calculate the concentrations of the gene in environmental DNA samples. At surface and DCM depths, the copy numbers of the nosZ gene were negligible, ranging between 0.05 and 0.09×10 6 copies L –1 and between 0.07 and 0.09×10 6 copies L –1 , respectively. At 250 m, the abundance of nosZ was the highest during the SIM (1.32×10 6 copies L –1 ), followed by the NEM (0.63×10 6 copies L –1 ) and FIM (0.49×10 6 copies L –1 ). At 500 m, the abundance of the nosZ gene was the highest during the SIM (1.50×10 6 copies L –1 ), followed by the NEM (0.63×10 6 copies L –1 ) and FIM (0.41×10 6 copies L –1 ). Overall, the highest copy number of the nosZ gene was detected at 500 m during the SIM (1.50×10 6 copies L –1 ). The abundance of the nosZ gene at 1,000 m varied from 0.05 to 0.08×10 6 copies L –1 . shows the seasonal abundance and distribution of nosZ genes. Effects of environmental parameters on OTUs and copy numbers During the SIM, DO (r=–0.701; P <0.01), nitrite (r=–0.985; P <0.001), and TOC (r=–0.593; P <0.05) negatively correlated with nosZ gene copy numbers, whereas ammonia (r=0.682; P <0.01) showed a positive correlation. Environmental parameters did not correlate with the abundance of the nosZ gene during the FIM, except for nitrite (r=–0.576; P <0.05), which showed a negative correlation. DO (r=–0.747; P <0.01) negatively correlated with the distribution of the nosZ gene during the NEM, while this correlation was positive for nitrite (r=0.827; P <0.001) ( ). A one-way ANOVA confirmed that physicochemical parameters, OTUs, and copy numbers significantly varied between depths ( ). Tukey’s post-hoc tests showed significant differences in DO, TOC, NO 3 – , and copy numbers between 250 and 500 m ( ) during all three seasons. PCA revealed three principal components with eigenvalues >1, which explained 92.39% of all variations in physicochemical parameters. The first component (PC1) explained a total variance of 42.20% and reflected a strong gradient caused by DO (0.943) and TOC (0.916). The second component (PC2) accounted for 27.06% of variance and was reflected by NH 4 + (0.958). The third component (PC3) explained 23.13% of variance and included NO 2 – (0.947) ( ). Multiple regression analyses showed that OTUs and copy numbers were influenced by DO, TOC, NO 2 – , and NO 3 – , while copy numbers were influenced by DO, TOC, and NH 4 + ( ). Richness, OTUs, and rarefaction curves Shannon’s and Simpson’s diversity indices revealed that the diversity of nosZ denitrifiers was the highest during the SIM, followed by the FIM and NEM. Diversity was the highest at 250 m during all sampling seasons. The non-parametric estimators, Chao 1 and ACE showed a higher number of OTUs at 250 m during all seasons. The SIM (250 m) and NEM (250 m) had the highest number of OTUs ( ). A rarefaction curve was plotted based on the number of clones and OTUs to investigate the relationship between sampling efforts and diversity. The saturation of the rarefaction curve indicated that the sampling effort sufficiently covered nosZ -denitrifying bacteria ( ).
Temperature, salinity, pH, and total organic carbon (TOC) were consistent, while the concentrations of DO and nutrients varied at core OMZ depths ( ). The average DO concentration decreased from 15.73 μmol L –1 at 250 m to 5.85 μmol L –1 at 500 m. The concentration of nitrate was higher at core OMZ depths during all three seasons and ranged between 15.33 to 32.79 μmol L –1 . The highest concentrations of nitrite (2.57 μmol L –1 ) and ammonia (NH 4 + 1.01 μmol L –1 ) were noted at 250 and 500 m during the NEM and SIM, respectively. The concentrations of DO, NO 2 – NO 3 – , and NH 4 + during all three seasons are shown in .
Evolutionary differences between OTUs and representative sequences were assessed by phylogenetic analyses. PCR amplification of the nosZ gene (259-bp amplification product) was only positive for samples collected at 250 and 500 m in all seasons. Among the 210 clones sequenced, 171 non-chimeric nosZ gene sequences were obtained and the following clone libraries were built: SIM-250, SIM-500, FIM-250, FIM-500, NEM-250, and NEM-500.
nosZ denitrifiers The sequences of nosZ clones showed 80–93% identity with one another and 77–88% similarity to sequences in GeneBank. All OTUs obtained in the present study matched the sequences of the phylum Proteobacteria in the NCBI database. Taxonomic analyses reveals that most nosZ OTUs were associated with Pseudomonas , Rhodopseudomonas , Bradyrhizobium , and Alphaproteobacteria . A small percentage of bacteria was affiliated with Azospirillum , Achromobacter , Cupriavidus , Nisaea , Thalassobaculum , Sinorhizobium , Herbaspirillum , Burkholderia , and Alcaligenes . The phylogenetic analysis showed that OTUs in the present study closely matched the environmental sequences obtained from the suboxic zone of the AS, deep-sea waters of the Mediterranean Sea, the crop soil and wetland sediment of Mexico, terrestrial subsurface sediments, a marine aquaculture biofilter, and paddy soil. Among the 87 clones sequenced from nosZ250m, 17 OTUs were obtained ( ), the majority of which were affiliated with the genera Rhodopseudomonas (18%, 3 OTUs) and Pseudomonas (18%, 3 OTUs). Azospirillum (2 OTUs), Bradyrhizobium (2 OTUs), Achromobacter (2 OTUs), and Cupriavidus (2 OTUs) each contributed 12% of the total number of OTUs. Bacterial sequences belonging to the genera Nisaea (1 OTU), Thalassobaculum (1 OTU), and class Alphaproteobacteria (1 OTU) formed OTUs with a smaller number of clones. Thirteen OTUs were generated from nosZ500m (84 sequences) ( ), the majority of which belonged to the genera Bradyrhizobium (15%, 2 OTUs), Pseudomonas (15%, 2 OTUs), and Alphaproteobacteria (15%, 2 OTUs). Thalassobaculum (1 OTU), Rhodopseudomonas (1 OTU), Sinorhizobium (1 OTU), Azospirillum (1 OTU), Herbaspirillum (1 OTU), Burkholderia (1 OTU), and Alcaligenes (1 OTU) were the other minor OTUs present. Among the 171 clones sequenced from core OMZ depths, 27 OTUs were generated, which were affiliated with the classes Alpha- , Beta- , and Gammaproteobacteria ( ). The majority of OTUs were affiliated with the genera Bradyrhizobium (5 OTUs, 18.5%, 32 sequences), Azospirillum (5 OTUs, 18.5%, 32 sequences), and Pseudomonas (4 OTUs, 14.8%, 26 sequences). Three OTUs (11.1%) each were associated with Rhodopseudomonas (19 sequences), Sinorhizobium (19 sequences), and Achromobacter (19 sequences). One OTU (6 sequences) each was affiliated with the genera Thalassobaculum , Nisaea , and Burkholderia and the class Alphaproteobacteria ( ).
nosZ denitrifiers The seasonal distribution patterns of nosZ gene sequences at the class level from 250 and 500 m showed that the maximum number of sequences was affiliated with the class Alphaproteobacteria . At 250 m, Alphaproteobacteria , Oscillatoriophycideae , Bacilli , and Actinobacteria were detected during all three seasons. Acidobacteria was found during the SIM and FIM, Chloroflexia only during the SIM, and Clostridia and Cytophagia only during the NEM. The maximum number of classes was detected in the SIM (7), followed by the FIM (5) and NEM (5) ( ). At 500 m, Alphaproteobacteria was predominant, followed by Bacilli , Actinobacteria , and Oscillatoriophycideae . Clostridia and Mollicutes were the other classes present at small percentages. No significant variations among classes were observed between seasons, except for Clostridia and Mollicutes , which were only found during the NEM ( ). The number of classes detected at 250 m (8) was higher than that detected at 500 m (6).
nosZ gene Standard curves for the nosZ gene were plotted and qPCR efficiency (81.93%) was calculated. The efficiency reading was used as a reference to calculate the concentrations of the gene in environmental DNA samples. At surface and DCM depths, the copy numbers of the nosZ gene were negligible, ranging between 0.05 and 0.09×10 6 copies L –1 and between 0.07 and 0.09×10 6 copies L –1 , respectively. At 250 m, the abundance of nosZ was the highest during the SIM (1.32×10 6 copies L –1 ), followed by the NEM (0.63×10 6 copies L –1 ) and FIM (0.49×10 6 copies L –1 ). At 500 m, the abundance of the nosZ gene was the highest during the SIM (1.50×10 6 copies L –1 ), followed by the NEM (0.63×10 6 copies L –1 ) and FIM (0.41×10 6 copies L –1 ). Overall, the highest copy number of the nosZ gene was detected at 500 m during the SIM (1.50×10 6 copies L –1 ). The abundance of the nosZ gene at 1,000 m varied from 0.05 to 0.08×10 6 copies L –1 . shows the seasonal abundance and distribution of nosZ genes.
During the SIM, DO (r=–0.701; P <0.01), nitrite (r=–0.985; P <0.001), and TOC (r=–0.593; P <0.05) negatively correlated with nosZ gene copy numbers, whereas ammonia (r=0.682; P <0.01) showed a positive correlation. Environmental parameters did not correlate with the abundance of the nosZ gene during the FIM, except for nitrite (r=–0.576; P <0.05), which showed a negative correlation. DO (r=–0.747; P <0.01) negatively correlated with the distribution of the nosZ gene during the NEM, while this correlation was positive for nitrite (r=0.827; P <0.001) ( ). A one-way ANOVA confirmed that physicochemical parameters, OTUs, and copy numbers significantly varied between depths ( ). Tukey’s post-hoc tests showed significant differences in DO, TOC, NO 3 – , and copy numbers between 250 and 500 m ( ) during all three seasons. PCA revealed three principal components with eigenvalues >1, which explained 92.39% of all variations in physicochemical parameters. The first component (PC1) explained a total variance of 42.20% and reflected a strong gradient caused by DO (0.943) and TOC (0.916). The second component (PC2) accounted for 27.06% of variance and was reflected by NH 4 + (0.958). The third component (PC3) explained 23.13% of variance and included NO 2 – (0.947) ( ). Multiple regression analyses showed that OTUs and copy numbers were influenced by DO, TOC, NO 2 – , and NO 3 – , while copy numbers were influenced by DO, TOC, and NH 4 + ( ).
Shannon’s and Simpson’s diversity indices revealed that the diversity of nosZ denitrifiers was the highest during the SIM, followed by the FIM and NEM. Diversity was the highest at 250 m during all sampling seasons. The non-parametric estimators, Chao 1 and ACE showed a higher number of OTUs at 250 m during all seasons. The SIM (250 m) and NEM (250 m) had the highest number of OTUs ( ). A rarefaction curve was plotted based on the number of clones and OTUs to investigate the relationship between sampling efforts and diversity. The saturation of the rarefaction curve indicated that the sampling effort sufficiently covered nosZ -denitrifying bacteria ( ).
In the highly productive surface waters (0–150 m) of the AS-OMZ, the in situ oxidation and rapid decomposition of organic matter leads to the near-exhaustion of DO at intermediate depths (~200–1,200 m column) triggering intense denitrification ( ), which, in turn, causes the copious efflux of N 2 O into the atmosphere ( ; ). Although present in small numbers, nosZ denitrifiers are diverse ( ) and play a critical role in nitrous oxide production. However, most studies conducted on nosZ bacteria associated with denitrification are performed on cultures and do not represent the total microbial community. Therefore, detection and characterization using a metagenomic approach are important for recognizing the nosZ bacterial community, which is critical in denitrification. Although mostly detected in the soil ecosystem ( ), recent metagenomic sequencing studies identified Clades I and II nosZ genes in regions associated with the OMZ of the Eastern Tropical Pacific, the AS, and in oxygenated surface waters of the Arctic and Southern Oceans ( ). The majority of studies that attempted to characterize nosZ gene diversity using DNA-based PCR approaches mainly focused on Clade I nosZ because the high diversity of Clade II nosZ makes it challenging to design a universal primer set that effectively amplifies all nosZ genes in this clade. The primer pair (nosZ1F/nosZ1R; ) used in the present study was suitable for both marine and terrestrial nosZ sequences. This primer amplifies a shorter region of 259 bps than those developed by others (1,100 bp; ) for marine targets, and, thus, was more suitable for the present study. Although these primers may not efficiently amplify and cover the sequence divergence of Clade II nosZ sequences ( ), previous studies using microarrays showed that they amplified some of these sequences ( ). Rhodobacteraceaea affiliated with Clade II nosZ genes detected in the present study was identified in the anoxic section of a wastewater treatment plant using Clade II nosZ gene-specific primers (DaeHyun, D.K., et al. , 2019 Development of group-specific nosZ quantification method targeting active nitrous oxide reducing population in complex environmental samples. bioRxiv https://doi.org/10.1101/710483 ). Nevertheless, we acknowledge that the PCR primers used in the present study may have been biased towards the detection of Clade I nosZ genes and, thus, we may have underestimated the real abundance of nosZ genes in our samples. Nevertheless, the present results indicate that marine nosZ denitrifiers (Clade I) inhabit core AS-OMZ depths and play an equal and significant role at the ASTS, leading to a high percentage of fixed nitrogen loss. The nosZ phylogeny The absence of nosZ -denitrifying bacteria from the surface, DCM, and 1,000 m, and its invariable presence in samples from 250 and 500 m during all three seasons is consistent with previous findings from the AS-OMZ ( ). As the level of oxygen falls below the detection limit ( ), conditions become favorable for denitrification ( ). The persistence of low oxygen levels in the core of the OMZ may be a factor limiting nosZ -denitrifying bacteria to depths of 250 and 500 m ( ). The AS-OMZ with intense upwelling and the low exchange of intermediate waters provides a high redox environment for the growth and multiplication of denitrifying bacteria. The present results demonstrated that nosZ denitrifiers inhabited the core of the AS-OMZ with DO levels fluctuating between 0.76–11 μmol L –1 , similar to findings from the Colombian Pacific Bay (CPB; ). All nosZ sequences obtained from the ASTS were affiliated with the phylum Proteobacteria, as was also reported by , , and . Proteobacteria have been identified as the dominant phylum and play a significant role in denitrification in the AS-OMZ ( ; ). OTUs obtained from the ASTS aligned into three classes: Alphaproteobacteria , Betaproteobacteria , and Gammaproteobacteria , which is consistent with findings from the OMZ of the Subtropical Deep Reservoir ( ) and sediments of the AS ( ). nosZ denitrifiers from the present study were predominant within the class Alphaproteobacteria , which is in accordance with the findings of from the AS. The nosZ denitrifiers identified in the present study were dominated by phylotypes affiliated to Pseudomonas , Rhodopseudomonas , Bradyrhizobium , and Alphaproteobacteria . This is similar to the findings of and . It is important to note that Bradyrhizobium , an aerobic anoxygenic phototrophic bacterium typically reported in oxic waters ( ), was found at core OMZ depths in the present study. Few OTUs from the ASTS were homologous with the novel sequence of Nisaea denitrificans (class Alphaproteobacteria ) isolated from the Mediterranean Sea, which is potentially involved in denitrification ( ). Sequences from the ASTS showed similarity to the nosZ gene isolates of Achromobacter detected from the OMZ of the CPB ( ), Herbaspirillum identified in anaerobic wastewater treatment plants ( ), Sinorhizobium reported in the sediments of the Atlantic ( ), and Azospirillum found in eutrophic freshwater lakes ( ). We also identified NosZ denitrifiers affiliated with Burkholderia , Alcaligenes , and Cupriavidus , which were previously detected in boreal peat moss ( ), wastewater treatment plants ( ), and marsh soils ( ), respectively. The majority of sequences identified in the present study showed homogeneity to those reported by from the seasonal OMZ in the AS. All of the nosZ denitrifiers identified in the present study were actively involved in nitrous oxide production. Some OTU sequences from this study aligned with the cultured, facultative anaerobe Thalassobaculum (family Rhodospirillaceae ), the role of which remains unknown ( ). Abundance and distribution of the nosZ gene NosZ denitrifiers were uncommon at 5 m, DCM, and 1,000 m. Quantitative analyses of the nosZ gene from these depths showed lower Ct values than those at core OMZ depths (250 and 500 m). PCR with a higher concentration of DNA did not yield positive amplification for any of the samples taken from depths of 5 m, DCM, or 1,000 m. Therefore, nosZ genes at these depths were limited and hard to detect via conventional PCR. Additionally, ; , and reported that the sensitivity of qPCR was higher than that of conventional PCR. Melt curves, the melting temperature, and all other protocols confirmed that qPCR amplification at 5 m, DCM, and 1,000 m was not an artefact. Although differences were observed in abundance, the highest copy numbers during all three seasons were detected at core OMZ depths. Therefore, the oxygen concentration at these depths was the most suitable for denitrifying bacteria, indicating the perennial survival of nosZ denitrifiers in the AS-OMZ. The higher abundance of nosZ denitrifiers during the SIM may be attributed to organic carbon in the OMZ being a significant substrate that supports the existence of denitrifying and anammox bacteria ( ). During the SIM, bacterial communities in the AS-OMZ are sustained by slow-to-degrade dissolved organic carbon (DOC) ( ) i.e. , the SIM is a transitional phase with low primary productivity ( ) due to the persistence of oligotrophic conditions and stratification. We herein reported a higher abundance of N 2 O-reducing bacteria from the ASTS than nirS denitrifiers ( ). The nirS and nosZ genes are both assumed to be present in the genome as single-copy genes; however, there are exceptions for nosZ genes ( ). One possible explanation for differences in abundance is that not all N 2 O-consuming bacteria contain a complete denitrification gene sequence ( ). nosZ gene-associated bacteria lack the other steps required for conventional denitrification. In comparisons with other ecosystems, bacteria with only nosZ genes are over-represented in the genomes of marine bacteria ( ). However, the nirS gene, is associated with bacteria that contain a complete denitrification pathway ( ). Another contributing factor that may explain this difference is the specificity of PCR primers. The primers used in the present study represent a more extensive database of nosZ sequences (terrestrial and marine sequences), whereas the nirS primers used in previous studies ( ) are potentially biased towards marine sequences ( ). Diversity and richness estimation reported that DO and organic matter were important factors affecting the microbial community composition in the OMZ. The present results suggest that DO and TOC play a critical role in influencing the diversity and abundance of nosZ denitrifiers during different seasons. The presence of denitrifiers at core OMZ depths (as is implicit in derived diversity indices and the richness estimators ACE and Chao 1) indicated that low concentrations of oxygen, nitrite, and ammonia provided an ideal environment for the presence of nosZ denitrifiers ( ). In contrast to the findings of , the present results showed a higher diversity of nosZ denitrifiers at 250 m than at 500 m. Although the sampling size in the present study was not very large, the saturation of rarefaction curves indicates that the diversity of nosZ denitrifiers was adequately covered.
nosZ phylogeny The absence of nosZ -denitrifying bacteria from the surface, DCM, and 1,000 m, and its invariable presence in samples from 250 and 500 m during all three seasons is consistent with previous findings from the AS-OMZ ( ). As the level of oxygen falls below the detection limit ( ), conditions become favorable for denitrification ( ). The persistence of low oxygen levels in the core of the OMZ may be a factor limiting nosZ -denitrifying bacteria to depths of 250 and 500 m ( ). The AS-OMZ with intense upwelling and the low exchange of intermediate waters provides a high redox environment for the growth and multiplication of denitrifying bacteria. The present results demonstrated that nosZ denitrifiers inhabited the core of the AS-OMZ with DO levels fluctuating between 0.76–11 μmol L –1 , similar to findings from the Colombian Pacific Bay (CPB; ). All nosZ sequences obtained from the ASTS were affiliated with the phylum Proteobacteria, as was also reported by , , and . Proteobacteria have been identified as the dominant phylum and play a significant role in denitrification in the AS-OMZ ( ; ). OTUs obtained from the ASTS aligned into three classes: Alphaproteobacteria , Betaproteobacteria , and Gammaproteobacteria , which is consistent with findings from the OMZ of the Subtropical Deep Reservoir ( ) and sediments of the AS ( ). nosZ denitrifiers from the present study were predominant within the class Alphaproteobacteria , which is in accordance with the findings of from the AS. The nosZ denitrifiers identified in the present study were dominated by phylotypes affiliated to Pseudomonas , Rhodopseudomonas , Bradyrhizobium , and Alphaproteobacteria . This is similar to the findings of and . It is important to note that Bradyrhizobium , an aerobic anoxygenic phototrophic bacterium typically reported in oxic waters ( ), was found at core OMZ depths in the present study. Few OTUs from the ASTS were homologous with the novel sequence of Nisaea denitrificans (class Alphaproteobacteria ) isolated from the Mediterranean Sea, which is potentially involved in denitrification ( ). Sequences from the ASTS showed similarity to the nosZ gene isolates of Achromobacter detected from the OMZ of the CPB ( ), Herbaspirillum identified in anaerobic wastewater treatment plants ( ), Sinorhizobium reported in the sediments of the Atlantic ( ), and Azospirillum found in eutrophic freshwater lakes ( ). We also identified NosZ denitrifiers affiliated with Burkholderia , Alcaligenes , and Cupriavidus , which were previously detected in boreal peat moss ( ), wastewater treatment plants ( ), and marsh soils ( ), respectively. The majority of sequences identified in the present study showed homogeneity to those reported by from the seasonal OMZ in the AS. All of the nosZ denitrifiers identified in the present study were actively involved in nitrous oxide production. Some OTU sequences from this study aligned with the cultured, facultative anaerobe Thalassobaculum (family Rhodospirillaceae ), the role of which remains unknown ( ).
nosZ gene NosZ denitrifiers were uncommon at 5 m, DCM, and 1,000 m. Quantitative analyses of the nosZ gene from these depths showed lower Ct values than those at core OMZ depths (250 and 500 m). PCR with a higher concentration of DNA did not yield positive amplification for any of the samples taken from depths of 5 m, DCM, or 1,000 m. Therefore, nosZ genes at these depths were limited and hard to detect via conventional PCR. Additionally, ; , and reported that the sensitivity of qPCR was higher than that of conventional PCR. Melt curves, the melting temperature, and all other protocols confirmed that qPCR amplification at 5 m, DCM, and 1,000 m was not an artefact. Although differences were observed in abundance, the highest copy numbers during all three seasons were detected at core OMZ depths. Therefore, the oxygen concentration at these depths was the most suitable for denitrifying bacteria, indicating the perennial survival of nosZ denitrifiers in the AS-OMZ. The higher abundance of nosZ denitrifiers during the SIM may be attributed to organic carbon in the OMZ being a significant substrate that supports the existence of denitrifying and anammox bacteria ( ). During the SIM, bacterial communities in the AS-OMZ are sustained by slow-to-degrade dissolved organic carbon (DOC) ( ) i.e. , the SIM is a transitional phase with low primary productivity ( ) due to the persistence of oligotrophic conditions and stratification. We herein reported a higher abundance of N 2 O-reducing bacteria from the ASTS than nirS denitrifiers ( ). The nirS and nosZ genes are both assumed to be present in the genome as single-copy genes; however, there are exceptions for nosZ genes ( ). One possible explanation for differences in abundance is that not all N 2 O-consuming bacteria contain a complete denitrification gene sequence ( ). nosZ gene-associated bacteria lack the other steps required for conventional denitrification. In comparisons with other ecosystems, bacteria with only nosZ genes are over-represented in the genomes of marine bacteria ( ). However, the nirS gene, is associated with bacteria that contain a complete denitrification pathway ( ). Another contributing factor that may explain this difference is the specificity of PCR primers. The primers used in the present study represent a more extensive database of nosZ sequences (terrestrial and marine sequences), whereas the nirS primers used in previous studies ( ) are potentially biased towards marine sequences ( ).
reported that DO and organic matter were important factors affecting the microbial community composition in the OMZ. The present results suggest that DO and TOC play a critical role in influencing the diversity and abundance of nosZ denitrifiers during different seasons. The presence of denitrifiers at core OMZ depths (as is implicit in derived diversity indices and the richness estimators ACE and Chao 1) indicated that low concentrations of oxygen, nitrite, and ammonia provided an ideal environment for the presence of nosZ denitrifiers ( ). In contrast to the findings of , the present results showed a higher diversity of nosZ denitrifiers at 250 m than at 500 m. Although the sampling size in the present study was not very large, the saturation of rarefaction curves indicates that the diversity of nosZ denitrifiers was adequately covered.
In the OMZ layers of the Arabian Sea, denitrification is a crucial pathway ( ; ; ; ) that has enabled the occurrence and sizable abundance of a diverse group of microbial communities, including those not taking part in the process, per se . The present results showed that the diversity nosZ denitrifiers was low and limited to core OMZ depths, suggesting that low concentrations of organic matter in OMZs not only reduce the number of available niches for microbes ( ), but also unfavorably influence the denitrifying microbial community structure. While nirS -possessing denitrifiers control N 2 O emissions ( ; ), the higher abundance of the nosZ gene than nirS from the ASTS in the present study suggests otherwise. Furthermore, our results on hydrographic parameters indicate that the concentrations of DO and TOC influence the abundance and distribution of nosZ -denitrifying bacteria at core OMZ depths of the AS.
Bandekar, M., Ramaiah, N., Seleyi, S. C.., Nazareth, D. R.., and Kekäläinen, J. (2023) Diversity and Quantitative Detection of Clade I Type nosZ Denitrifiers in the Arabian Sea Oxygen Minimum Zone. Microbes Environ 38 : ME22056. https://doi.org/10.1264/jsme2.ME22056
Supplementary Material
|
The opinions of GDPs on the challenges of managing tooth wear in primary dental care | f36a6ce7-e081-4361-b599-994f30ea2bb8 | 10037362 | Dental[mh] | Three general dental practitioners (GDPs), each with over 20 years' experience in a variety of healthcare settings, but predominantly in general dental practice, were interviewed about their opinions of how patients with tooth wear are managed in primary care. The interviewees were included on the basis that they worked in differing primary dental care settings in England, had significant professional experience themselves and worked with other colleagues of varying ages and experiences. Their practices models were: 1) majority NHS-funded; 2) majority private, fee-per-item; and 3) majority private, capitation and insurance scheme ( ). The interviewees had mostly worked in either the North or the South of England and their practice colleagues had experience in several other areas of the country. None had pursued post-graduate studies related to the management of tooth wear, which was important to attempt to avoid opinions of those which are likely to differ from most GDPs. A series of ten questions to be discussed was developed in advance, with contribution from and agreement with the interviewees. They were encouraged to discuss the questions and likely responses with other colleagues working at their practices before the interviews. Two weeks later, an initial semi-structured group discussion with the three interviewees was held via a virtual meeting, with the questions presented and discussions moderated by the authors and the dialogue recorded for later transcription. The transcription was shared with the interviewees. Two weeks later, a second group discussion was held via a virtual meeting, with each question and previous response discussed again. The transcription was again shared within the group, to ensure the responses correctly represented their views.
Are you seeing younger people with a lot more tooth wear presenting or is it older people who are just keeping their teeth for longer? 'People are certainly keeping their teeth longer, so they are wearing out more and we see some more tooth wear, sensitivity and facial pain' 'Over my 20-plus years in practice, there is a pattern of increasing numbers of younger patients and some worrying features, such as combined erosion and attrition with really damaged occlusal surfaces of their posterior teeth. The older patients tend to either have a mouth full of restorations or a lifetime of tooth wear. Tooth wear is a long-term condition for our older patients, with an occasional acute problem' 'In our practices, despite the majority of patients having at least some tooth wear, most just don't see it as a problem. If they have a cavity and toothache, they know they have something wrong. But the tooth wear is rarely an issue for them, until they have symptoms or an aesthetic concern'. What would prompt you to start discussing your patients' tooth wear with them? 'I rarely consider low-level wear to be an issue and don't often mention it - it is usually just physiological tooth wear. For moderate and severe tooth wear, whether the patient has noticed a problem or not, it is important to mention this with them. I offer the opportunity to have further discussions about treatment, when the patient is ready to consider it' 'NHS practices aren't funded for working like this. We think about whether to get into a conversation about tooth wear, including explaining the diagnosis and how to prevent and treat it. Considering how quickly we all have to work in general practice, I have to just ask myself "am I worried by what I'm seeing?" due the extent of the damage or because of their symptoms: "is this active tooth wear or are we seeing the long-term gradual changes in a stable mouth?" Most patients are usually just not that bothered by tooth wear. The challenge is to explain the condition and to go through the prevention and consequences of treating or not treating it, with the minority who are concerned' 'I ask myself that if I don't talk about the tooth wear and it deteriorates, will I be in trouble? Secondly, is there going to be a reasonably easy way to treat this, or will I wish I'd never mentioned it?' How do you and your colleagues monitor a patient's tooth wear? 'I don't have a structured, "scientific" way of monitoring it. For example, I don't take serial study models. If someone has significant attrition, then the only treatment I would feel capable of giving is a soft splint. I don't think I'm qualified to do anything more than that. The only time I intervene is when people have an aesthetic issue with their teeth' 'We don't repeatedly monitor tooth wear at all. We are a mainly UDA-based [Units of Dental Activity] practice. We struggle to do everything we need to do in a check-up under one UDA, never mind monitoring tooth wear. It is mostly based around highlighting and informing the patient about the condition. Rarely, if everything was done and we had a very interested patient, you might offer to do some study models and photographs for their own awareness and to be kept for monitoring' 'We don't often use the Smith and Knight index, BEWE [Basic Erosive Wear Examination] or anything similar. I just describe it as early, moderate or severe and comment whether it is into dentine, pulp seen or at gingival level. There is just no funding or resource for it under UDAs, unfortunately'. If you thought the patient may have an undiagnosed gastroesophageal problem, would you contact their general medical practitioner? 'Perhaps, especially if the teeth are showing signs of erosion but the diet isn't particularly acidic' 'Only if I am really concerned and just advise them to discuss it next time they attend with their GP. I wouldn't refer them myself unless I was concerned that they weren't capable of doing so themselves' 'I wouldn't write to the GP myself. Whether it is reflux problems for tooth wear or assessing diabetes in a patient with periodontal disease, I would just advise the patient to speak to their GP'. What would prompt you to start offering treatment? 'The likelihood to intervene is much higher in a younger person. I'd be more worried about a 17-year-old with erosion and attrition that a 70-year-old with buccal abrasion cavities. We are less inclined to intervene with the older patient for two reasons: they don't see it as an issue and they question whether it is going to make a significant difference to them keeping their teeth' 'If a patient isn't interested, motivated or even aware when we mention it, they are not normally going to want to do anything about it. I feel uneasy talking to them about treatments that they have said they feel are unnecessary or unimportant for them. However, when they are aware of an aesthetic issue, it is much easier to discuss treatment' 'My initial treatment step would be preventive, to stop the tooth wear getting worse. I would ask about diet if I was worried about erosion. The only other intervention I might do for a bruxist is to give a splint. If the patient is not motivated, if they don't see the benefit of the splint, they are never going to wear it. The only tooth wear patients who seem to wear a splint really well, are either those with facial pain or those who chose cosmetic treatment and want to protect it' 'If I am considering offering treatment to a patient, I am also thinking about what I will do when this fails in the future. For a younger patient, this is the first step in their cycle of treatment through their life. It is a big issue when treating a chronic condition such as tooth wear. Who has responsibility to maintaining a case?' Is tooth wear something your practice team has a focus on? 'No, it is low on our list of priorities. There are no national policies to push practices to think about it. We do have assessment of tooth wear on our practice template. Our patients have other problems, such as caries and periodontal disease. I don't know of any NHS practices that focus on tooth wear, although some private practices take specialist referrals for patients with tooth wear and occlusion problems' 'I agree. Tooth wear is probably the last thing we would focus on, at the bottom of the list of other priorities' 'This week we had our first ever practice meeting in over 20 years related to tooth wear and that is because I was discussing this with you all. We have sometimes spoken about a case we are struggling with, but not about tooth wear as a subject' 'From my point of view, when I see tooth wear, apart from making a soft splint, there isn't really anything I can do. I will restore teeth for a patient who has an issue with aesthetics. I don't know anything about things such as deprogrammers etc. I wouldn't know where to start with complex occlusal problems and severe tooth wear'. It is important that we as professionals still focus on an issue such as tooth wear, even if it is probably irrelevant for many patients. Our experience normally tells us who are the 10% of patients we need to ask more questions to. What is an acceptable standard of dental care for tooth wear? 'This isn't spoken about enough. There needs to be some level of public health understanding. Not everyone is going to be perfectly healthy. People are going to have all sorts of chronic conditions and it is then just a case of making them aware. As long as the patients are aware and have some input into how to manage it, we shouldn't feel a pressure to fix everything. Some patients are always going to grind their teeth, they are not going to wear a splint and they're always going to break most of what we use to restore their teeth. As long as they know their options and agree with our approach, we are practising safely and professionally' 'Many dentists are trained with a big medico-legal worry that if we can't fix it, we are not treating the patient correctly. We have got to accept that there are a lot of people who will be affected by tooth wear, for the rest of their life' 'From a medico-legal perspective, we are much more focused on periodontal disease. Diagnosing it, making sure the patient is aware of it, treating it or referring the patient' 'Record keeping for tooth wear isn't as comprehensive as is it for periodontal disease, soft tissue assessment and medical history'. What restrictions are there for managing tooth wear in your practices? 'I am not sure that many GDPs have the experience and knowledge of what to do for their patients. The cost of treatment and the time required to complete the treatment are also problems. Our associates said that as they can't do any comprehensive treatment for tooth wear under UDAs, they feel less confident and are probably deskilled. They wouldn't want to try some treatment they're not going to do very often. If treating tooth wear cases with composite and using direct Dahl approaches was more commonly performed or was an allowable thing within our NHS contract, for a severe tooth wear case, they would be more confident to do it' 'I think we have the same concerns and I'm sure many of our younger dentists will also be worried: "once I start this, where am I going with it?" Managing tooth wear just doesn't fit with the NHS UDA system at all' 'I'm quite patient-led with tooth wear cases. Someone is only going to have that treatment if they feel their tooth wear is an issue for them, especially a cosmetic problem. If it is a complex case, even from the perspective of private practice, most aren't likely to be able to spend that sort of money on their treatment' 'My practice is mainly private, and we have more time with our patients, but I still don't treat tooth wear as you may think we would do, without the NHS UDA constraints. I would only treat them if they didn't like the appearance of their teeth. I wouldn't even talk about a splint unless they were a severe bruxist' 'We use a capitation system for our patients. This lends itself better to managing some issues such as tooth wear, as we have more time and fewer financial restrictions'. What have been your experiences of referring tooth wear patients into a dental hospital for treatment? What is your confidence with maintaining these patients? 'Over the years since I qualified, I have had some good experiences and usually just expect a treatment plan to be sent back. More recently, most referrals are now returned, and the patients are rarely, if ever, taken on for treatment. We've not been given a lot of help' 'For a patient treated elsewhere, I would usually offer minor repairs myself. Anything more than that and I have to consider the consequences of what will I become responsible for, if I do any more. I usually advise the patient that they need to go back to whoever provided their treatment, as that colleague will know what was planned, in case the first treatment approach failed' 'We don't have a dental hospital nearby for that sort of referral, so I can't really comment. Our nearest hospital would be at least an hour away. We only refer privately to a local specialist prosthodontist' 'My practice colleagues said they have almost given up referring. Since COVID-19, just about everything gets rejected. Even before COVID, the patient would be returned with a treatment plan that was beyond what they are capable of doing or that was impossible to do in primary care, within the UDA system. Primary dental care is meant to be for prevention and for more simple and achievable treatments'. Can tooth wear be effectively managed in primary dental care? 'There are two aspects to this: managing prevention of tooth wear and managing treatment of tooth wear. At present, neither of these are resourced well in this country and the biggest NHS focus is improving access for any dental care' 'I don't know if there are any healthcare systems in the world that have worked out how to fund or incentivise prevention. So, we either work out a way to improve the prevention side or we must accept we can only treat some patients who have an advanced condition. These patients must be motivated through either symptoms or appearance or have enough understanding of why intervening earlier is worthwhile. That makes it such a challenging area. The public has a better understanding of tooth decay and to some extent, periodontal disease. But not tooth wear' 'There isn't enough funding for Tier 2 services and Managed Clinical Networks for tooth wear. Under the UDA system, there would need to be a separate band for complex treatments, such as this. It needs something like the old "prior approval" system, where we do an additional, more detailed assessment, demonstrating why this is an advanced case, requiring extra UDAs and get that approved before starting' 'At a public health level, we need to consider what the public money is spent on. We need to ask: "should NHS dentistry only be for the essentials, such as tooth decay, gum disease, simple tooth replacement, mouth cancer screening etc?" If you happen to come across some tooth wear when examining a patient, is the dentist's only role to explain that it is happening and make the patient aware of the problem but not actually to try to fix it? Perhaps the NHS has got more important things to spend the public's money on than tooth wear' 'Let us be pragmatic about the budget required to manage tooth wear, the complexity of the treatments, the longevity of restorations and the burden of maintenance. Is it "only tooth wear" for most people? Most people don't complain about their ability to eat. Very few are really affected by sensitivity. It really is just a cosmetic problem for most of these patients'.
'People are certainly keeping their teeth longer, so they are wearing out more and we see some more tooth wear, sensitivity and facial pain' 'Over my 20-plus years in practice, there is a pattern of increasing numbers of younger patients and some worrying features, such as combined erosion and attrition with really damaged occlusal surfaces of their posterior teeth. The older patients tend to either have a mouth full of restorations or a lifetime of tooth wear. Tooth wear is a long-term condition for our older patients, with an occasional acute problem' 'In our practices, despite the majority of patients having at least some tooth wear, most just don't see it as a problem. If they have a cavity and toothache, they know they have something wrong. But the tooth wear is rarely an issue for them, until they have symptoms or an aesthetic concern'.
'I rarely consider low-level wear to be an issue and don't often mention it - it is usually just physiological tooth wear. For moderate and severe tooth wear, whether the patient has noticed a problem or not, it is important to mention this with them. I offer the opportunity to have further discussions about treatment, when the patient is ready to consider it' 'NHS practices aren't funded for working like this. We think about whether to get into a conversation about tooth wear, including explaining the diagnosis and how to prevent and treat it. Considering how quickly we all have to work in general practice, I have to just ask myself "am I worried by what I'm seeing?" due the extent of the damage or because of their symptoms: "is this active tooth wear or are we seeing the long-term gradual changes in a stable mouth?" Most patients are usually just not that bothered by tooth wear. The challenge is to explain the condition and to go through the prevention and consequences of treating or not treating it, with the minority who are concerned' 'I ask myself that if I don't talk about the tooth wear and it deteriorates, will I be in trouble? Secondly, is there going to be a reasonably easy way to treat this, or will I wish I'd never mentioned it?'
'I don't have a structured, "scientific" way of monitoring it. For example, I don't take serial study models. If someone has significant attrition, then the only treatment I would feel capable of giving is a soft splint. I don't think I'm qualified to do anything more than that. The only time I intervene is when people have an aesthetic issue with their teeth' 'We don't repeatedly monitor tooth wear at all. We are a mainly UDA-based [Units of Dental Activity] practice. We struggle to do everything we need to do in a check-up under one UDA, never mind monitoring tooth wear. It is mostly based around highlighting and informing the patient about the condition. Rarely, if everything was done and we had a very interested patient, you might offer to do some study models and photographs for their own awareness and to be kept for monitoring' 'We don't often use the Smith and Knight index, BEWE [Basic Erosive Wear Examination] or anything similar. I just describe it as early, moderate or severe and comment whether it is into dentine, pulp seen or at gingival level. There is just no funding or resource for it under UDAs, unfortunately'.
'Perhaps, especially if the teeth are showing signs of erosion but the diet isn't particularly acidic' 'Only if I am really concerned and just advise them to discuss it next time they attend with their GP. I wouldn't refer them myself unless I was concerned that they weren't capable of doing so themselves' 'I wouldn't write to the GP myself. Whether it is reflux problems for tooth wear or assessing diabetes in a patient with periodontal disease, I would just advise the patient to speak to their GP'.
'The likelihood to intervene is much higher in a younger person. I'd be more worried about a 17-year-old with erosion and attrition that a 70-year-old with buccal abrasion cavities. We are less inclined to intervene with the older patient for two reasons: they don't see it as an issue and they question whether it is going to make a significant difference to them keeping their teeth' 'If a patient isn't interested, motivated or even aware when we mention it, they are not normally going to want to do anything about it. I feel uneasy talking to them about treatments that they have said they feel are unnecessary or unimportant for them. However, when they are aware of an aesthetic issue, it is much easier to discuss treatment' 'My initial treatment step would be preventive, to stop the tooth wear getting worse. I would ask about diet if I was worried about erosion. The only other intervention I might do for a bruxist is to give a splint. If the patient is not motivated, if they don't see the benefit of the splint, they are never going to wear it. The only tooth wear patients who seem to wear a splint really well, are either those with facial pain or those who chose cosmetic treatment and want to protect it' 'If I am considering offering treatment to a patient, I am also thinking about what I will do when this fails in the future. For a younger patient, this is the first step in their cycle of treatment through their life. It is a big issue when treating a chronic condition such as tooth wear. Who has responsibility to maintaining a case?'
'No, it is low on our list of priorities. There are no national policies to push practices to think about it. We do have assessment of tooth wear on our practice template. Our patients have other problems, such as caries and periodontal disease. I don't know of any NHS practices that focus on tooth wear, although some private practices take specialist referrals for patients with tooth wear and occlusion problems' 'I agree. Tooth wear is probably the last thing we would focus on, at the bottom of the list of other priorities' 'This week we had our first ever practice meeting in over 20 years related to tooth wear and that is because I was discussing this with you all. We have sometimes spoken about a case we are struggling with, but not about tooth wear as a subject' 'From my point of view, when I see tooth wear, apart from making a soft splint, there isn't really anything I can do. I will restore teeth for a patient who has an issue with aesthetics. I don't know anything about things such as deprogrammers etc. I wouldn't know where to start with complex occlusal problems and severe tooth wear'.
'This isn't spoken about enough. There needs to be some level of public health understanding. Not everyone is going to be perfectly healthy. People are going to have all sorts of chronic conditions and it is then just a case of making them aware. As long as the patients are aware and have some input into how to manage it, we shouldn't feel a pressure to fix everything. Some patients are always going to grind their teeth, they are not going to wear a splint and they're always going to break most of what we use to restore their teeth. As long as they know their options and agree with our approach, we are practising safely and professionally' 'Many dentists are trained with a big medico-legal worry that if we can't fix it, we are not treating the patient correctly. We have got to accept that there are a lot of people who will be affected by tooth wear, for the rest of their life' 'From a medico-legal perspective, we are much more focused on periodontal disease. Diagnosing it, making sure the patient is aware of it, treating it or referring the patient' 'Record keeping for tooth wear isn't as comprehensive as is it for periodontal disease, soft tissue assessment and medical history'.
'I am not sure that many GDPs have the experience and knowledge of what to do for their patients. The cost of treatment and the time required to complete the treatment are also problems. Our associates said that as they can't do any comprehensive treatment for tooth wear under UDAs, they feel less confident and are probably deskilled. They wouldn't want to try some treatment they're not going to do very often. If treating tooth wear cases with composite and using direct Dahl approaches was more commonly performed or was an allowable thing within our NHS contract, for a severe tooth wear case, they would be more confident to do it' 'I think we have the same concerns and I'm sure many of our younger dentists will also be worried: "once I start this, where am I going with it?" Managing tooth wear just doesn't fit with the NHS UDA system at all' 'I'm quite patient-led with tooth wear cases. Someone is only going to have that treatment if they feel their tooth wear is an issue for them, especially a cosmetic problem. If it is a complex case, even from the perspective of private practice, most aren't likely to be able to spend that sort of money on their treatment' 'My practice is mainly private, and we have more time with our patients, but I still don't treat tooth wear as you may think we would do, without the NHS UDA constraints. I would only treat them if they didn't like the appearance of their teeth. I wouldn't even talk about a splint unless they were a severe bruxist' 'We use a capitation system for our patients. This lends itself better to managing some issues such as tooth wear, as we have more time and fewer financial restrictions'.
'Over the years since I qualified, I have had some good experiences and usually just expect a treatment plan to be sent back. More recently, most referrals are now returned, and the patients are rarely, if ever, taken on for treatment. We've not been given a lot of help' 'For a patient treated elsewhere, I would usually offer minor repairs myself. Anything more than that and I have to consider the consequences of what will I become responsible for, if I do any more. I usually advise the patient that they need to go back to whoever provided their treatment, as that colleague will know what was planned, in case the first treatment approach failed' 'We don't have a dental hospital nearby for that sort of referral, so I can't really comment. Our nearest hospital would be at least an hour away. We only refer privately to a local specialist prosthodontist' 'My practice colleagues said they have almost given up referring. Since COVID-19, just about everything gets rejected. Even before COVID, the patient would be returned with a treatment plan that was beyond what they are capable of doing or that was impossible to do in primary care, within the UDA system. Primary dental care is meant to be for prevention and for more simple and achievable treatments'.
'There are two aspects to this: managing prevention of tooth wear and managing treatment of tooth wear. At present, neither of these are resourced well in this country and the biggest NHS focus is improving access for any dental care' 'I don't know if there are any healthcare systems in the world that have worked out how to fund or incentivise prevention. So, we either work out a way to improve the prevention side or we must accept we can only treat some patients who have an advanced condition. These patients must be motivated through either symptoms or appearance or have enough understanding of why intervening earlier is worthwhile. That makes it such a challenging area. The public has a better understanding of tooth decay and to some extent, periodontal disease. But not tooth wear' 'There isn't enough funding for Tier 2 services and Managed Clinical Networks for tooth wear. Under the UDA system, there would need to be a separate band for complex treatments, such as this. It needs something like the old "prior approval" system, where we do an additional, more detailed assessment, demonstrating why this is an advanced case, requiring extra UDAs and get that approved before starting' 'At a public health level, we need to consider what the public money is spent on. We need to ask: "should NHS dentistry only be for the essentials, such as tooth decay, gum disease, simple tooth replacement, mouth cancer screening etc?" If you happen to come across some tooth wear when examining a patient, is the dentist's only role to explain that it is happening and make the patient aware of the problem but not actually to try to fix it? Perhaps the NHS has got more important things to spend the public's money on than tooth wear' 'Let us be pragmatic about the budget required to manage tooth wear, the complexity of the treatments, the longevity of restorations and the burden of maintenance. Is it "only tooth wear" for most people? Most people don't complain about their ability to eat. Very few are really affected by sensitivity. It really is just a cosmetic problem for most of these patients'.
Even by involving experienced colleagues, it is not possible for these discussions to accurately represent the views of all GDPs, especially those much more recently qualified. However, each of those that contributed, work alongside several other colleagues and have awareness of a wider view within the profession. The participants all confirmed that their patients are affected by tooth wear, but in general, justify their decisions in how they manage tooth wear by their belief that: The large majority of their patients have no concerns and express no symptoms related to tooth wear A smaller proportion of patients present with tooth wear that requires intervention, due to their age, the extent of tooth wear, or their concern about their symptoms and appearance Dentists have other, more important priorities to focus on, such as patient access, caries and periodontal disease Other public health issues may be of a greater priority for public funding Current practice business models and remuneration systems limit the clinical management of tooth wear Dentists may lack clinical knowledge and confidence in treatment methods Inadequate capacity for support from secondary care services. Other authors have identified similar themes related to management of tooth wear in primary care. O'Hara and Millar in 2020 evaluated currently available methods for assessing and monitoring tooth wear in a general dental practice environment. They concluded 'dentists do not seem to be aware of the current guidelines but do make reasonable attempts to monitor tooth wear'. Condon and Eaton in 2020, recognised that 'restoring complex tooth wear cases is technically challenging and not well-remunerated under the NHS general dental service contract. Therefore, numbers of referrals to secondary care are increasing, but these are often rejected as dental hospitals have a high workload. This may make it difficult for patients with tooth wear to access appropriate care unless paying privately, which may be costly for them. Their study found low confidence in restoring complex tooth wear cases: only 21% of practitioners stated they would treat complex cases under the current NHS contract and 62% reported that they had experienced difficulty referring these cases to hospital. In 2018, O'Toole and co-workers assessed charting, risk assessment and treatment-planning of tooth wear between four recently qualified and seven experienced dentists in general dental practice. Their findings identified that: there are significant differences between patient management between recently qualified and experienced dentists; improvements are required in recording (48% versus 5%), risk assessing (51% versus 1%) and preventive treatment planning (62% versus 1%) of erosive tooth wear; and experienced GDPs may benefit from re-training in this area. In 2020, Mehta and co-workers assessed the habits of tooth wear risk assessment and charting using a tooth wear index, by UK and non-UK dental practitioners. Based on a sample of 297 responses, 81% agreed to the need to undertake risk assessment for all dental patients attending for a first-time consultation. In total, 59% undertook risk assessments for patients previously identified with signs of severe tooth wear. The routine use of a clinical index to perform tooth wear charting was described by 13.5%, with 5% documenting the frequent use of the BEWE tool. The paper found that specialist dental practitioners or those with further post-graduate training were more likely to use a tooth wear index.
Understanding the opinions and perceptions of experienced GDPs is important when the management of patients presenting with tooth wear is considered, as almost all patients will be either initially or only ever seen in the primary care environment. A number of challenges to delivery of treatment are identified and discussed. There is a large body of academic work related to tooth wear, including text books and guidelines from the Royal College of Surgeons of England, much of which explains and recommends what would be considered 'best practice'. The opinions expressed by the interviewees in this study and other recent studies suggest that a notable and perhaps alarming gap exists between the management of tooth wear in general dental practice in England compared to the published guidelines. The opinions of the interviewees in this study can also be interpreted as suggesting, controversially, that, as the management of most patients with tooth wear is not prioritised within commissioned NHS funding models, by GDPs, by dental hospitals receiving referrals or by most patients, perhaps it is not currently of importance outside academic environments. Further, cross-profession discussions are required, related to both commissioning NHS primary and secondary care services and to addressing the lack of dentists' confidence to manage patients affected by tooth wear.
|
Integrated Small Animal PET/CT/RT with Onboard PET/CT Image Guidance for Preclinical Radiation Oncology Research | b6af1d60-e9a7-4b17-a753-c674110b459b | 10037572 | Internal Medicine[mh] | Combined PET/CT functional and anatomic images are ubiquitously used in clinical radiotherapy (RT) for tumor diagnosis and delineation to guide the treatment plan and access the therapeutic effects. However, in the field of preclinical radiation research, all existing image-guided small animal irradiator systems are only equipped with onboard CT without PET due to various technical reasons. Although optical or other modality imaging technology can provide some functional information of the tumor, for most radiation oncology studies, particularly for those with orthotopically implemented animal tumor models, PET is still the modality that can provide the desired quantitative, functional/biological/molecular image to substantially improve the accuracy of radiation targeting to reduce the radiation margin for increasing the dose to the tumor and sparing the surrounding normal tissues. The lack of preclinical PET/CT image-guided RT capability has severely limited the precision of animal RT study to accurately investigate radiation’s biological effects and, more importantly in the era of translational RT research, the value to translate the findings from preclinical research to clinical applications or vice versa. We recently developed a compact and lightweight small animal positron emission tomography (PET) with uniform, high spatial resolution across its imaging field-of-view (FOV) that is suited for integration with a cone-beam computed tomography (CBCT, or CT for short) image-guided small animal radiation therapy (RT) irradiator (CT/RT) . Although functional and anatomical images can be acquired by the standalone PET and the CT inside the irradiator and the dual-modality images can be co-registered through software , such offboard PET/CT imaging requires the transportation and reposition of the animal over separated modalities that will lengthen the scan time and be prone to dual-modality alignment and associated image registration errors . On the other hand, the integrated onboard PET/CT/RT can streamline the process of animal positioning, dual-image acquisition, and data processing to minimize the scan time and alignment error, simplify the workflow, and even permit motion compensated or other advanced image-guided RT applications with the onboard PET/CT imaging . In this study, we report the integration of the PET with an existing CT/RT and the performance evaluation of this integrated preclinical PET/CT/RT for onboard PET/CT image-guided preclinical RT research.
2.1. PET As shown in a, the compact and lightweight PET is suited to be installed inside an existing small animal CT/RT for mechanical integration. The following are summaries of its specification and performance . The PET consists of a ring of 12 detector panels in a dodecagon configuration. Each detector panel has a 30 × 30 array of 1 × 1 × 20 mm 3 Ce-doped Lutetium-Yttrium Oxyorthosilicate (Lu 0.6 Y 1.4 SiO 0.5 :Ce, LYSO) scintillators. Each end of the scintillator array is optically coupled to an 8 × 8 silicon photomultiplier array (model MPPC S13361-2050-08, Hamamatsu Photonics K.K., Shizuoka, Japan) for depth-of-interaction (DOI) measurement based on the dual-ended readout . All six scintillator surfaces were lapped with 0.03 mm grade to provide balanced light output and good DOI resolution. Optical reflective films (ESR, 3M Corp, Saint Paul, MN, USA) with 0.06 mm thickness were used between scintillators to prevent inter-scintillator optical crosstalk and enhance the light output of each crystal. Each MPPC active pixel size is 2 × 2 mm 2 , with nominal operational bias and dark count rate around 54.4 v and ~500 Kcps. The outer diameter and axial length of the PET gantry are 33.0 cm and 11.0 cm, respectively, with a 11.0 cm diameter animal port. The total weight of the gantry is 6.5 kg, which includes the PET detector ring, front-end readout electronics boards, air-low fans, 3D-printed packaging holders, and metal plate for installation. The imaging field-of-view (FOV) is 8.0 cm in diameter and 3.5 cm axial extent, with an 11.0 cm diameter animal port. The mean spatial resolutions along radial, tangential, and axial directions were measured as 1.30, 1.18, and 0.96 mm with ~11.8% uniform activity background. A ~1.1 mm uniform spatial resolution was achieved within 20 mm FOV radius without the resolution recovery, while the same ~1.1 mm uniform spatial resolution within 30 mm FOV radius can be achieved by including the resolution recovery with a ~1.0 mm full-width-half-maximum (FWHM) Gaussian . The maximum sensitivity at the center of FOV was ~1.8% with a 350–650 keV energy window. 2.2. Integration of PET and CT/RT With the mechanical support of off-the-shelf metal bars, the PET gantry was stationarily installed inside the CT/RT (X-RAD 225Cx, Precision X-ray Irradiation, Madison, CT, USA) in a tandem PET/CT imaging configuration . Laser beams were used to ensure that the orientations and centers of both the PET and CT imaging FOVs were closely aligned with each other. To minimize the impact of X-ray radiations, a 10 mm thick aluminum alloy metal plate was placed at the end of the PET gantry that faced the incoming scattered X-rays. For data transmission, signal synchronization, and system command, twelve low-voltage-differential-signals (LVDS) cables were used to connect 12 front-end detector readout boards to the system electronics board sitting outside the CT/RT. The PET data acquisition computer was also placed outside and next to the CT/RT control console. By placing the system electronics board and acquisition computer outside the CT/RT, we eased the problem of the limited interior CT/RT space and minimized the heat generated from PET components that could raise the temperature inside the CT/RT and affect the detector performance. To enable the translation of an animal between PET and CT for dual imaging acquisitions, an additional linear translational bed was constructed and attached to the existing 3D stage to extend the horizontal translation range that was limited by the original 3D bed motion mechanical system. This additional bed consisted of a thin carbon-fiber curved plate on plastic lightweight rails driven by a step motor for linear translational motion . The size and thickness of the plate were 15.5 cm long, 5.0 cm wide, and 2 mm thick, which had approximately the same attenuation effect to CT image as that with the original animal bed. PET data were independently acquired and processed with the system’s PC computer. A user interface based on MATLAB (MathWorks, Natick, MA, USA) was developed to control the motion of the linear translational animal bed, start and stop the data acquisition, and monitor the detector performance. Acquired data were processed offline for data calibration, selection, and correction. 2.3. PET/CT Coordinate System Alignment for Dual-Modality Image Registration An existing method was applied to align the PET and CT coordinate systems for image registration . It was based on rigid-body transformation of dual-modality fiducial markers to measure the misalignment between the PET and CT coordinate systems and transform the original PET coordinate system to accurately align it with the CT coordinate system. In this study, five independent PET and CT scans were conducted to measure the dual modality positions of a 22 Na point source with 251 kBq activity [MMS03 Multimodal Imaging Source, Eckert & Ziegler Isotope Products, Valencia, CA, USA]. The physical 22 Na radioactivity material was sealed at the center of a plastic cube with 10 × 10 × 10 mm 3 volume. The spherical radioactivity source with around 0.25 mm diameter was precisely at the geometric center of the cube and was used as the CT-measured point source position. The PET-measured point source position was calculated from the intensity-weighted centroid of the corresponding reconstructed PET image. For each PET/CT scan, the point source was placed at a different non-coplanar position inside the PET and CT image FOVs, which was achieved by inserting the point source into a size-matching cubic hole at the surface of a holder that was 3D-printed with low density material and a hollowed center to minimize the attenuation. The holder had five such cubic holes at its different surfaces and locations for five different PET/CT scans. The holder was rigidly fixed on the bed during all five PET/CT scans. The differences among corresponding PET-measured and CT-measured point source positions were used to measure the misalignment between the PET and CT coordinates and, correspondingly, to transform and align the PET coordinate system with the CT coordinate system that had its center of FOV (CFOV) at the RT isocenter. 2.4. Initial Onboard PET/CT Imaging Study 2.4.1. Phantom Study An ultra-micro hot-rods phantom (Data Spectrum Corporation, Durham, NC, USA) filled with [ 18 F]NaF was used to evaluate the dual-modality acquisitions and the accuracy of the registered PET/CT images. The diameter and length of the phantom insert were 26.0 and 10.0 mm. The diameters of the through holes (rods), which were arranged in six sections within the insert, are 0.75, 1.0, 1.35, 1.7, 2.0, and 2.4 mm, respectively. The phantom was placed at the CFOV of each imaging modality for its acquisition. The PET image was acquired for 30 min with a start radioactivity at 3.8 MBq, 350–650 keV energy window, and 10 ns coincidence timing window. An open-source code (CASToR) based on an ordered subset expectation maximization (OSEM) algorithm was used to reconstruct the image with 10 subsets and 10 iterations . CT image data were acquired from 301 projections of 40 kVp and 5 mA X-rays and were reconstructed with standard a FDK algorithm . PET and CT phantom images were registered with the aligned PET/CT coordinate systems. The accuracy of the image registration was assessed by the difference between the rod centers measured from the PET and CT images. 2.4.2. Animal Study An initial onboard PET/CT animal tumor imaging study was conducted. shows a ~21 g mouse bearing a tumor being placed at the positions for PET and CT imaging acquisitions. The tumor (human lung carcinoma cell, H460), which had been subcutaneously implanted in the hind leg of a female athymic nude mouse, was measured with a size of ~3.9 mm width and ~7.1 mm length at the time of imaging. For PET imaging, ~5.6 MBq of 2-deoxy-2-[ 18 F]fluoro-D-glucose ([ 18 F]FDG) was injected via the tail vein for data acquisition (25 min) after radiotracer uptake (~30 min). Coincidence events were selected with a 350–650 KeV energy window and a 10 ns time window. The images were reconstructed using CASToR with 4 subsets and 4 iterations, 128 × 128 × 128 matrix, and 0.5 × 0.5 × 0.5 mm 3 voxel size. For CT imaging, 301 projection data were acquired with 40 kVp and 5 mA X-rays and were reconstructed with standard a FDK algorithm with 350 × 350 × 350 matrix and 0.4 × 0.4 × 0.4 mm 3 voxel size. In addition to the one-bed tumor focused image, a whole-body animal PET image was also achieved by acquiring data with the animal being at three sequential bed positions and joining their corresponding reconstructed images based on the known imaging positions. The same data acquisition parameters, data process, and image reconstruction used in the one-bed tumor imaging were applied, except a radioisotope decay correction was also applied to the dataset acquired at each bed position to ensure the similar count statistics among the datasets .
As shown in a, the compact and lightweight PET is suited to be installed inside an existing small animal CT/RT for mechanical integration. The following are summaries of its specification and performance . The PET consists of a ring of 12 detector panels in a dodecagon configuration. Each detector panel has a 30 × 30 array of 1 × 1 × 20 mm 3 Ce-doped Lutetium-Yttrium Oxyorthosilicate (Lu 0.6 Y 1.4 SiO 0.5 :Ce, LYSO) scintillators. Each end of the scintillator array is optically coupled to an 8 × 8 silicon photomultiplier array (model MPPC S13361-2050-08, Hamamatsu Photonics K.K., Shizuoka, Japan) for depth-of-interaction (DOI) measurement based on the dual-ended readout . All six scintillator surfaces were lapped with 0.03 mm grade to provide balanced light output and good DOI resolution. Optical reflective films (ESR, 3M Corp, Saint Paul, MN, USA) with 0.06 mm thickness were used between scintillators to prevent inter-scintillator optical crosstalk and enhance the light output of each crystal. Each MPPC active pixel size is 2 × 2 mm 2 , with nominal operational bias and dark count rate around 54.4 v and ~500 Kcps. The outer diameter and axial length of the PET gantry are 33.0 cm and 11.0 cm, respectively, with a 11.0 cm diameter animal port. The total weight of the gantry is 6.5 kg, which includes the PET detector ring, front-end readout electronics boards, air-low fans, 3D-printed packaging holders, and metal plate for installation. The imaging field-of-view (FOV) is 8.0 cm in diameter and 3.5 cm axial extent, with an 11.0 cm diameter animal port. The mean spatial resolutions along radial, tangential, and axial directions were measured as 1.30, 1.18, and 0.96 mm with ~11.8% uniform activity background. A ~1.1 mm uniform spatial resolution was achieved within 20 mm FOV radius without the resolution recovery, while the same ~1.1 mm uniform spatial resolution within 30 mm FOV radius can be achieved by including the resolution recovery with a ~1.0 mm full-width-half-maximum (FWHM) Gaussian . The maximum sensitivity at the center of FOV was ~1.8% with a 350–650 keV energy window.
With the mechanical support of off-the-shelf metal bars, the PET gantry was stationarily installed inside the CT/RT (X-RAD 225Cx, Precision X-ray Irradiation, Madison, CT, USA) in a tandem PET/CT imaging configuration . Laser beams were used to ensure that the orientations and centers of both the PET and CT imaging FOVs were closely aligned with each other. To minimize the impact of X-ray radiations, a 10 mm thick aluminum alloy metal plate was placed at the end of the PET gantry that faced the incoming scattered X-rays. For data transmission, signal synchronization, and system command, twelve low-voltage-differential-signals (LVDS) cables were used to connect 12 front-end detector readout boards to the system electronics board sitting outside the CT/RT. The PET data acquisition computer was also placed outside and next to the CT/RT control console. By placing the system electronics board and acquisition computer outside the CT/RT, we eased the problem of the limited interior CT/RT space and minimized the heat generated from PET components that could raise the temperature inside the CT/RT and affect the detector performance. To enable the translation of an animal between PET and CT for dual imaging acquisitions, an additional linear translational bed was constructed and attached to the existing 3D stage to extend the horizontal translation range that was limited by the original 3D bed motion mechanical system. This additional bed consisted of a thin carbon-fiber curved plate on plastic lightweight rails driven by a step motor for linear translational motion . The size and thickness of the plate were 15.5 cm long, 5.0 cm wide, and 2 mm thick, which had approximately the same attenuation effect to CT image as that with the original animal bed. PET data were independently acquired and processed with the system’s PC computer. A user interface based on MATLAB (MathWorks, Natick, MA, USA) was developed to control the motion of the linear translational animal bed, start and stop the data acquisition, and monitor the detector performance. Acquired data were processed offline for data calibration, selection, and correction.
An existing method was applied to align the PET and CT coordinate systems for image registration . It was based on rigid-body transformation of dual-modality fiducial markers to measure the misalignment between the PET and CT coordinate systems and transform the original PET coordinate system to accurately align it with the CT coordinate system. In this study, five independent PET and CT scans were conducted to measure the dual modality positions of a 22 Na point source with 251 kBq activity [MMS03 Multimodal Imaging Source, Eckert & Ziegler Isotope Products, Valencia, CA, USA]. The physical 22 Na radioactivity material was sealed at the center of a plastic cube with 10 × 10 × 10 mm 3 volume. The spherical radioactivity source with around 0.25 mm diameter was precisely at the geometric center of the cube and was used as the CT-measured point source position. The PET-measured point source position was calculated from the intensity-weighted centroid of the corresponding reconstructed PET image. For each PET/CT scan, the point source was placed at a different non-coplanar position inside the PET and CT image FOVs, which was achieved by inserting the point source into a size-matching cubic hole at the surface of a holder that was 3D-printed with low density material and a hollowed center to minimize the attenuation. The holder had five such cubic holes at its different surfaces and locations for five different PET/CT scans. The holder was rigidly fixed on the bed during all five PET/CT scans. The differences among corresponding PET-measured and CT-measured point source positions were used to measure the misalignment between the PET and CT coordinates and, correspondingly, to transform and align the PET coordinate system with the CT coordinate system that had its center of FOV (CFOV) at the RT isocenter.
2.4.1. Phantom Study An ultra-micro hot-rods phantom (Data Spectrum Corporation, Durham, NC, USA) filled with [ 18 F]NaF was used to evaluate the dual-modality acquisitions and the accuracy of the registered PET/CT images. The diameter and length of the phantom insert were 26.0 and 10.0 mm. The diameters of the through holes (rods), which were arranged in six sections within the insert, are 0.75, 1.0, 1.35, 1.7, 2.0, and 2.4 mm, respectively. The phantom was placed at the CFOV of each imaging modality for its acquisition. The PET image was acquired for 30 min with a start radioactivity at 3.8 MBq, 350–650 keV energy window, and 10 ns coincidence timing window. An open-source code (CASToR) based on an ordered subset expectation maximization (OSEM) algorithm was used to reconstruct the image with 10 subsets and 10 iterations . CT image data were acquired from 301 projections of 40 kVp and 5 mA X-rays and were reconstructed with standard a FDK algorithm . PET and CT phantom images were registered with the aligned PET/CT coordinate systems. The accuracy of the image registration was assessed by the difference between the rod centers measured from the PET and CT images. 2.4.2. Animal Study An initial onboard PET/CT animal tumor imaging study was conducted. shows a ~21 g mouse bearing a tumor being placed at the positions for PET and CT imaging acquisitions. The tumor (human lung carcinoma cell, H460), which had been subcutaneously implanted in the hind leg of a female athymic nude mouse, was measured with a size of ~3.9 mm width and ~7.1 mm length at the time of imaging. For PET imaging, ~5.6 MBq of 2-deoxy-2-[ 18 F]fluoro-D-glucose ([ 18 F]FDG) was injected via the tail vein for data acquisition (25 min) after radiotracer uptake (~30 min). Coincidence events were selected with a 350–650 KeV energy window and a 10 ns time window. The images were reconstructed using CASToR with 4 subsets and 4 iterations, 128 × 128 × 128 matrix, and 0.5 × 0.5 × 0.5 mm 3 voxel size. For CT imaging, 301 projection data were acquired with 40 kVp and 5 mA X-rays and were reconstructed with standard a FDK algorithm with 350 × 350 × 350 matrix and 0.4 × 0.4 × 0.4 mm 3 voxel size. In addition to the one-bed tumor focused image, a whole-body animal PET image was also achieved by acquiring data with the animal being at three sequential bed positions and joining their corresponding reconstructed images based on the known imaging positions. The same data acquisition parameters, data process, and image reconstruction used in the one-bed tumor imaging were applied, except a radioisotope decay correction was also applied to the dataset acquired at each bed position to ensure the similar count statistics among the datasets .
An ultra-micro hot-rods phantom (Data Spectrum Corporation, Durham, NC, USA) filled with [ 18 F]NaF was used to evaluate the dual-modality acquisitions and the accuracy of the registered PET/CT images. The diameter and length of the phantom insert were 26.0 and 10.0 mm. The diameters of the through holes (rods), which were arranged in six sections within the insert, are 0.75, 1.0, 1.35, 1.7, 2.0, and 2.4 mm, respectively. The phantom was placed at the CFOV of each imaging modality for its acquisition. The PET image was acquired for 30 min with a start radioactivity at 3.8 MBq, 350–650 keV energy window, and 10 ns coincidence timing window. An open-source code (CASToR) based on an ordered subset expectation maximization (OSEM) algorithm was used to reconstruct the image with 10 subsets and 10 iterations . CT image data were acquired from 301 projections of 40 kVp and 5 mA X-rays and were reconstructed with standard a FDK algorithm . PET and CT phantom images were registered with the aligned PET/CT coordinate systems. The accuracy of the image registration was assessed by the difference between the rod centers measured from the PET and CT images.
An initial onboard PET/CT animal tumor imaging study was conducted. shows a ~21 g mouse bearing a tumor being placed at the positions for PET and CT imaging acquisitions. The tumor (human lung carcinoma cell, H460), which had been subcutaneously implanted in the hind leg of a female athymic nude mouse, was measured with a size of ~3.9 mm width and ~7.1 mm length at the time of imaging. For PET imaging, ~5.6 MBq of 2-deoxy-2-[ 18 F]fluoro-D-glucose ([ 18 F]FDG) was injected via the tail vein for data acquisition (25 min) after radiotracer uptake (~30 min). Coincidence events were selected with a 350–650 KeV energy window and a 10 ns time window. The images were reconstructed using CASToR with 4 subsets and 4 iterations, 128 × 128 × 128 matrix, and 0.5 × 0.5 × 0.5 mm 3 voxel size. For CT imaging, 301 projection data were acquired with 40 kVp and 5 mA X-rays and were reconstructed with standard a FDK algorithm with 350 × 350 × 350 matrix and 0.4 × 0.4 × 0.4 mm 3 voxel size. In addition to the one-bed tumor focused image, a whole-body animal PET image was also achieved by acquiring data with the animal being at three sequential bed positions and joining their corresponding reconstructed images based on the known imaging positions. The same data acquisition parameters, data process, and image reconstruction used in the one-bed tumor imaging were applied, except a radioisotope decay correction was also applied to the dataset acquired at each bed position to ensure the similar count statistics among the datasets .
3.1. PET and CT Coordinate System Alignment shows the measured PET, CT, and PET/CT images of the 22 Na point source at three different FOV positions. The differences among all five source positions measured between PET and CT were used to calculate the PET/CT coordinate misalignment for PET coordinate transformation and alignment. The registered images show that PET and CT coordinates can be accurately aligned with the described method and procedure. 3.2. Phantom Study with Onboard PET and CT Acquisitions shows the [ 18 F]NaF PET, CT, and registered PET/CT images of the ultra-micro hot-rods phantom acquired with the onboard PET and CT. For the PET image that is shown in the transformed PET coordinate system, all hot-rods from 1.0 mm to 2.4 mm diameter are clearly separated. It also shows that uniform spatial resolution can be achieved by the PET detectors with the DOI measurement capability that is desired for a compact PET geometry . For the CT image, the image contrast and quality are relatively low due to the use of cone-beam CT. There are also image artifacts at the edge of the phantom due to the additional attenuation from the rails that mechanically support the translational bed , and darker spots in smaller rods and the area between the insert and outer holder of the phantom due to air bubbles produced during the radiotracer filling. Although the contrast and image quality are relatively low, all rods from 1.35 mm to 2.4 mm diameter rods can still be clearly identified and separated. The registered PET/CT image shows that both images are well aligned with each other. The centers of the identified rods were measured with the PET and CT images, and the mean difference among the PET-measured and CT-measured centers was 0.28 ± 0.27 mm, with 0.07 mm minimum and 0.67 mm maximum, which demonstrated that a sufficiently accurate onboard PET/CT image registration can be achieved. 3.3. Initial Animal Study with Onboard PET/CT Imaging shows the onboard PET/CT acquired [ 18 F]FDG PET, CT, and registered PET/CT animal images. The PET image was acquired with a one-bed position that covered the tumor volume within the PET FOV. The tumor can be clearly identified from the PET in all image slices. However, it is difficult to identify the tumor from the CT image. The registered PET/CT images provide the expected functional and anatomical information for tumor identification and boundary determination. shows the profiles across the PET/CT images at the positions as indicated in . It is obvious that the outer edge of the tumor can be well determined from both the PET and CT images because the tumor was on the skin surface. However, the inner boundary of the tumor is rather difficult to be determined from the CT image, while it can be clearly determined from the PET image. It shows that PET/CT imaging can significantly improve tumor boundary determination, as anticipated, which should lead to enhanced RT accuracy with improved precision of beam targeting. shows the PET/CT whole-body images of the same animal with [ 18 F]FDG PET acquisitions over three sequential bed positions. These selected slices show cardiac images with ventricle of the animal heart. It demonstrates the capability of acquiring onboard whole-body PET/CT animal images for image-guided preclinical RT study.
shows the measured PET, CT, and PET/CT images of the 22 Na point source at three different FOV positions. The differences among all five source positions measured between PET and CT were used to calculate the PET/CT coordinate misalignment for PET coordinate transformation and alignment. The registered images show that PET and CT coordinates can be accurately aligned with the described method and procedure.
shows the [ 18 F]NaF PET, CT, and registered PET/CT images of the ultra-micro hot-rods phantom acquired with the onboard PET and CT. For the PET image that is shown in the transformed PET coordinate system, all hot-rods from 1.0 mm to 2.4 mm diameter are clearly separated. It also shows that uniform spatial resolution can be achieved by the PET detectors with the DOI measurement capability that is desired for a compact PET geometry . For the CT image, the image contrast and quality are relatively low due to the use of cone-beam CT. There are also image artifacts at the edge of the phantom due to the additional attenuation from the rails that mechanically support the translational bed , and darker spots in smaller rods and the area between the insert and outer holder of the phantom due to air bubbles produced during the radiotracer filling. Although the contrast and image quality are relatively low, all rods from 1.35 mm to 2.4 mm diameter rods can still be clearly identified and separated. The registered PET/CT image shows that both images are well aligned with each other. The centers of the identified rods were measured with the PET and CT images, and the mean difference among the PET-measured and CT-measured centers was 0.28 ± 0.27 mm, with 0.07 mm minimum and 0.67 mm maximum, which demonstrated that a sufficiently accurate onboard PET/CT image registration can be achieved.
shows the onboard PET/CT acquired [ 18 F]FDG PET, CT, and registered PET/CT animal images. The PET image was acquired with a one-bed position that covered the tumor volume within the PET FOV. The tumor can be clearly identified from the PET in all image slices. However, it is difficult to identify the tumor from the CT image. The registered PET/CT images provide the expected functional and anatomical information for tumor identification and boundary determination. shows the profiles across the PET/CT images at the positions as indicated in . It is obvious that the outer edge of the tumor can be well determined from both the PET and CT images because the tumor was on the skin surface. However, the inner boundary of the tumor is rather difficult to be determined from the CT image, while it can be clearly determined from the PET image. It shows that PET/CT imaging can significantly improve tumor boundary determination, as anticipated, which should lead to enhanced RT accuracy with improved precision of beam targeting. shows the PET/CT whole-body images of the same animal with [ 18 F]FDG PET acquisitions over three sequential bed positions. These selected slices show cardiac images with ventricle of the animal heart. It demonstrates the capability of acquiring onboard whole-body PET/CT animal images for image-guided preclinical RT study.
Compared with a different approach with a pair of rotated PET detectors affixed to the CT gantry to provide PET/CT imaging , the integrated PET/CT/RT described in this study provides a stationary full-ring PET that can have substantial advantages which include significantly higher sensitivity and image quality, substantially shortened acquisition time with high counting rate, simplified operation without extra rotation of entire CT gantry solely for PET acquisition, more compatibility to the clinical PET/CT image-guided RT, and therefore practicality for routine preclinical RT research and translation to clinical applications . In our study, there was no measurable PET detector performance degradation from the imaging acquisition with PET radioactive sources. However, PET detector performance is sensitive to the exposure of external X-ray radiations, and any such exposure could potentially degrade the detector performance by impacting its gain and background noise. That is because the interaction probability to the SiPM arrays, which are semiconductor photon sensors from low-energy X-ray photons, are much higher than that from the 511 keV coincidence gamma photons. The severity of the detector performance degradation is X-ray energy, intensity, and exposure time dependent. With a CT imaging associated collimator attached to the X-ray tube that shielded scattered X-ray photons, there was no measurable PET detector performance degradation after multiple CT acquisitions. However, without the shielding from a collimator, such as during the flood field X-ray radiation for CT detector calibration, there was measurable PET detector performance degradation. Therefore, the current prototype PET works fine with a routine CT imaging acquisition that requires a collimator anyway, but it is not suited to stay inside the CT/RT during X-ray radiations without a collimator attached, or a more rigorous X-ray shielding would be required. Fortunately, such flood field X-ray radiation is rarely performed. In the current prototype development, two carbon-fiber flat plates with 53 cm long, 2.5 cm wide, and 2 mm thick, each plate was used as the rail to support the additional bed and guide its translational motion. One technical issue to be addressed in the next step is the attenuation of the rails to X-rays that led to visible CT image artifacts, such as those seen in . Although these artifacts did not seriously impact the focused studies in this research on demonstrating the feasibility of integrated PET/CT/RT for onboard PET/CT imaging and they can be corrected with known attenuations, for practical application with streamed PET and CT acquisition processes and required image qualities, it is important to overcome this issue without an extra and lengthy data correction. The potential solutions include using lower density yet still mechanically strong material to construct the rails to minimize the attenuation to X-rays, or to extend the horizontal translational range of the existing 3D bed motion to avoid adding an additional translational bed. The latter is the best approach and is engineeringly feasible, although it will require the modification of the existing bed motion mechanical system. The current prototype PET has a 3.5 cm axial FOV with one detector ring. Since detector and front-end readout electronics have modular designs, there are no fundamental technical challenges to extend the axial FOV by adding more detector rings to increase the system sensitivity. However, for the radiotherapy guidance with known tumor location and size, the current 3.5 cm axial FOV should be sufficient for most preclinical RT studies. The one detector ring also has the advantage of being light weight for PET integration, and it is also practical to acquire whole-body animal images with multiple bed acquisitions. On the other hand, the extended axial FOV with high sensitivity, improved image quality, and shortened scan time will be more suited for the imaging applications of diagnosis, RT monitoring, and therapy effectiveness assessment. Thus, a trade-off should be carefully considered for the pros and cons to extend the axial PET FOV. Another potentially required improvement to the prototype onboard PET/CT integration is to minimize the number of signal transmission cables. Currently, there are 12 LVDS cables that are cumbersome to handle and install inside the CT/RT, and the number of cables will be increased with increased detector rings. Some inter-detector signal processing, multiplexing, and transmission needs to be investigated, particularly if a larger number of detectors will be implemented. The PET performance with ambient temperature variation inside the CT/RT is another potential concern to be addressed. During our imaging and radiation study, the ambient temperature inside the CT/RT irradiator was increased ~1–2 °C after ~30 min since PET powered on, mainly due to the heat generated from PET FPGA electronics within the enclosed environment, but which stabilized after reaching the heat equilibrium. The PET gain and noise were also changed and stabilized accordingly. With appropriate energy window selection, PET imaging capability in our study was not affected. However, for routine applications with extended imaging time and repeated opening and closing of the irradiator enclosure, it is worth it to implement a mechanism to stabilize the ambient temperature inside the irradiator or adaptively adjust the detector voltage bias to stabilize the detector performance. Although PET and CT images can be accurately registered as demonstrated, the dual-modality coordinates alignment is a lengthy and complex process. Different from a clinical PET/CT where both scanners are tightly integrated together, the prototype preclinical PET was inserted inside the CT/RT and fixed with metal bars; quite often the PET needed to be removed and be inserted again, which could lead to coordinates misalignment and performing a new alignment procedure. Thus, either an accurate and reliable PET insert method beyond the current approach or a fast and accurate coordinates alignment method is needed for robust and routine preclinical PET/CT image-guided RT applications. With its compact size, light weight, and relatively large size animal port, the PET can also be integrated with other popular and latest animal CT/RT irradiators . Besides hardware integration by affixing PET inside a CT/RT, it is also feasible and potentially advantageous to mechanically insert the PET inside CT/RT for PET imaging and to optionally move it outside CT/RT after the imaging. This approach can increase the flexibility in terms of using the insert PET and CT/RT separately and reduce the interference between them, potentially have PET and CT acquisitions with overlapped image FOV without moving the animal, and minimize or even eliminate the radiation shielding to PET. In addition, it will also minimize the PET temperature-dependent performance variation, as the entire PET imaging session will be conducted with open irradiator enclosure under a stable room temperature. On the other hand, the challenges with respect to the accurate alignment of the inserted PET to CT/RT stably and repeatably for routine applications will have to be addressed. The future studies in this research will mainly include implementing and evaluating full data corrections and quantitative PET image processes, potentially integrating PET/CT acquisitions and processes under the same console, and most importantly guiding the treatment plan with the PET/CT functional and anatomical images and evaluating the effectiveness of this new paradigm of image guidance for preclinical radiation oncology research, such as the dosimetry evaluation including some guidelines for the target volume delineation and the calculation of CTV-PTV uncertainty interval depending on the possible sources of errors and uncertainties. After all the above improvements, it is our belief that the animal PET/CT/RT can play a critical role in accelerating preclinical RT research and expanding the investigation of tumor radiation biology and the related clinical translation at the level driven by applying tumor functional images and information, which will permit researchers to directly understand the impact of radiation with different doses, target volumes, and fractionations to the tumor and normal tissues, and therefore develop the optimized treatment plan. By providing functional PET images to understand the intricate systematic biological effects of different radiation deliveries and exploring various novel RT approaches, the onboard PET/CT/RT may also play an important role in the forefront RT research and applications, such radiomics, immuno-radiotherapy, and ultrahigh dose rate (FLASH) radiotherapy . For example, with the growing interest in preclinical PET radiomics studies , animal PET/CT/RT can provide a missing dataset of quantitative, biological image-guided radiotherapy to facilitate the radiomics analysis and modeling in preclinical studies. Additionally, it will also enable co-clinical radiomics investigations and potentially radiomics guided radiotherapy applications by comparing the clinical radiomics analysis and the preclinical outcome resulted from high-precision and biologically targeted animal radiation studies with a known disease model .
The first small animal integrated PET/CT/RT with the integration of a compact and lightweight stationary PET within a CT/RT has been developed and evaluated for its onboard PET/CT imaging capability. The initial study has shown that the prototype can achieve accurately registered onboard PET/CT phantom and animal images with normal PET and CT imaging conditions and acquisitions. The study demonstrated that an integrated PET/CT/RT can provide practical onboard functional/biological/molecular and anatomical image-guided preclinical radiation research with significantly improved accuracy of tumor delineation and radiation targeting, which will enhance the existing study, enable potentially new and breakthrough investigations, and ultimately expand and accelerate the radiation oncology research.
|
Comparative Study of Postmortem Concentrations of Benzodiazepines and Z-Hypnotics in Several Different Matrices | d08413ec-ca4b-4aa6-a63f-3beba8e227a9 | 10037634 | Forensic Medicine[mh] | Toxicological analyses are important in forensic autopsy cases. Several matrices can be utilized. Peripheral blood (PB) is, however, considered to be the gold standard in these cases ( ) as concentrations measured in blood can indicate impairment or intoxication. In special cases where the body has a high degree of autolysis/putrefaction, severe blood loss or burns/charring, collection of PB might be impossible. However, other matrices might still be available for drug analysis. Since the 1960s, benzodiazepines have been used as anxiolytics, hypnotics, sedatives, muscle relaxants and anticonvulsants ( ). The hypnotic drugs such as zopiclone and zolpidem (z-hypnotics) act via the same γ-aminobutyric acid type-A receptor as the benzodiazepines, although their chemical structures are different ( ). Both the benzodiazepines and z-hypnotics are drugs with abuse potential ( ), which can cause respiratory depression and death, especially when combined with opioids or alcohol ( ). Correspondingly, a benzodiazepine prevalence of 80–90% in fatal intoxication cases has been reported in Norway ( ). There is only limited amount of literature pertaining to concentrations of benzodiazepines and z-hypnotics in alternative postmortem matrices. Some studies reporting concentrations in different body fluids and tissues have been published, but these contain only a handful of cases ( , ) or one single benzodiazepine/z-hypnotic ( , ). Other studies also report concentrations of several benzodiazepines but do not compare the results to PB ( , ). The available amount of vitreous humor (VH) and pericardial fluid (PF) in postmortem cases is limited, but modern mass spectrometric techniques facilitate analysis of numerous substances in a limited amount of sample. Scott and Oliver compared temazepam, diazepam and nordiazepam in the blood and VH and found some correlation between concentrations for temazepam and diazepam but no correlation for nordiazepam ( ). Robertson and Drummer studied the distribution of nitrobenzodiazepines and their 7-amino metabolites ( ). They found higher concentrations of 7-amino metabolites than parent compounds in VH. For both parent and metabolite compounds, concentrations were about one-third of the concentrations in the blood. Diazepam was among the compounds first investigated for correlation between concentrations in blood and PF ( ). A study of 218 autopsy cases found relatively similar concentrations of alprazolam, diazepam, estazolam, bromazepam and zolpidem in heart blood compared to PF, while the concentration of midazolam was lower ( ). A correlation coefficient of 0.870 between concentrations in PF and PB was found for benzodiazepines as a group, based on 12 samples, in a study that included nine benzodiazepines ( ). Moriya and Hashimoto evaluated postmortem drug concentrations in cerebrospinal fluid compared with blood and PF ( ). Although muscle might be a relevant matrix for putrefied samples, available literature is mainly based on case reports and might not include blood for comparison ( ). A study of diazepam in rabbit tissue and body fluids showed lower concentrations in skeletal muscle and VH compared to blood for diazepam and nordiazepam ( ). Two studies of autopsy samples, each with four cases, found both higher and lower concentrations of diazepam and nordiazepam in muscle compared to blood. The ratios of concentrations in muscle relative to blood varied between 0.2 and 3.6 for diazepam and 0.4 and 1.5 for nordiazepam in the study by Garriot ( ) and 0.43 and 3.4 for diazepam and 0.86 and 1.9 for nordiazepam in the study by Christensen et al. ( ). A meta-analysis performed by Ketola and Kriikku ( ) in 2019 established median concentrations of drugs in various postmortem specimens and compared them with median concentrations of the same drugs in postmortem femoral blood. Diazepam, oxazepam, zopiclone and zolpidem were included in the meta-analysis. The number of published results for the different compounds available was however limited, from 1 to 11 studies. This limited amount of available data for evaluation of concentrations of benzodiazepines and z-hypnotics in other matrices compared to PB accentuates the need for systematic studies. Our study was designed to provide new knowledge of postmortem drug concentrations in different matrices for a wide range of drugs. By analyzing six different postmortem matrices from the same case and using the same analytical methods, we sought to provide comparable data. The aim of the present study was to compare concentrations of benzodiazepines and z-hypnotics in five alternative matrices to assess whether these concentrations are comparable to concentrations in PB.
Postmortem matrices collection and preparation Samples from medicolegal autopsies performed between June 2013 and November 2016 with suspected use of medicinal or illegal drugs were collected by forensic pathologists during autopsy. The autopsies were in general performed within a week after death. PB, cardiac blood (CB), psoas muscle (PM), lateral vastus muscle (LVM), PF and VH samples were collected for the study along with routine samples of urine and PB. PB, with a volume of 40–50 mL, was collected in the following manner: The iliac vein was transected. After initial blood flow from the vessel stopped, a ladle was placed on the superior part of the cross-section, and blood from the femoral vein was “milked” on to the ladle by compression to the inner thigh and groin. Thus, no blood from the caval vein or superior part of inguinal vein was collected. To enable collection of 40–50 mL, collection from both sides was necessary in a majority of cases. The study protocol has been described previously ( ). After arrival at our laboratory, in the Department of Forensic Sciences at Oslo University Hospital, the samples were stored refrigerated (4°C) until analysis, with a storage time of typically 1–2 weeks, to allow for the screening of PB. For cases where benzodiazepines or z-hypnotics were detected in PB in the routine cases, all six matrices collected as project samples (including PB) were analyzed together to minimize differences due to instrument or storage time variation. The muscle analysis was performed in aliquots of muscle homogenate, made from ∼3.5 g of the muscle in 14 mL of water. Corrections due to dilution were performed using the following formula: [12pt]{minimal}
}{}$$ & { Concentration\ in\ muscle} \\ & = (({ weighted\ gram\ muscle} + { amount\ of\ water}) . \\ & . / { gram\ weighted\ muscle}) \\ & { concentration\ in\ homogenized\ sample} $$ The forensic pathologist estimated the degree of autolysis, using a scale from 0 to 2, where 0 equaled no putrefaction, 1 slight putrefaction and 2 moderate putrefaction. Analytical methods PB collected for the case work was analyzed for ethanol, psychoactive medicinal drugs and common drugs of abuse. The benzodiazepines and z-hypnotics included in this screening method were clonazepam, 7-aminoclonazepam, flunitrazepam, 7-aminoflunitrazepam, nitrazepam, 7-aminonitrazepam, diazepam, nordiazepam, oxazepam, midazolam, bromazepam, lorazepam, phenazepam, zopiclone and zolpidem, and the screening was performed with a UHPLC–MS-MS method ( ) using an acidic mobile phase. Confirmation was performed with three different ultra high performance liquid chromatography--tandem mass spectrometry (UHPLC–MS-MS) methods using alkaline mobile phases. The methods for benzodiazepines and z-hypnotics have been published elsewhere ( , ), while the method for 7-amino metabolites can be found in the . Whole blood from the blood bank at Oslo University Hospital was used to prepare calibrator and quality control (QC) samples, and all matrices were quantified using the same calibration curve. Precision and accuracy for spiked QC samples in all six matrices calculated using whole blood calibration curves were tested in a separate experiment. To minimize differences in concentrations due to storage time or instrument variations, all matrices from each case were analyzed in the same series as far as possible. Spiked whole blood standards were used to prepare the calibration curve, however recovery and matrix effects were examined for all matrices. We report results above the limit of quantification (LOQ) of the confirmation methods ( ), while concentrations below LOQ were not included in the data evaluation. Matrix effects and extraction recovery We evaluated matrix effects using post-extraction addition as described by Matuszewski et al. ( ) at two concentration levels for all six matrices, using negative samples from five different autopsy cases. For zopiclone and zolpidem, one sample used to test for matrix effects and extraction recovery was evaluated by the pathologist to have slight putrefaction while the others had no putrefaction. For the benzodiazepines and 7-amino metabolites, three samples had no putrefaction and two samples had slight putrefaction. Three different sets were made: (A) blank extracted matrix samples spiked after extraction, (B) neat standards in mobile phase and (C) matrix samples spiked before extraction. Internal standard was added to all sets after extraction. Matrix effects were calculated by dividing the mean peak height for the samples spiked after extraction (A) with the mean peak height found for the neat solutions (B) and given as percentage: Matrix effect = (A/B)× 100%. Extraction recovery was found by comparing set C with set A: Recovery = (C/A)× 100%. Statistics We performed data analysis using IBM SPSS Statistics for Windows version 23.0 and calculated concentration ratios for all the matrices relative to PB. Correlations between concentrations in the alternative matrices and PB were examined using Spearman’s rank correlation, with two-tailed evaluation of significance. P -values < 0.05 were considered statistically significant, and missing values were excluded pairwise. Ethics The study was approved by the Regional Committee for Medical Research Ethics (reference number 2012/2173) and by the Higher Prosecuting Authority (reference number 2012/02455). The cases were not listed in the National Registry of Withdrawal from Biological Research Consent, where people can deny inclusion into biological research projects. Additionally, for cases included after the middle of 2015, the next of kin was given the opportunity to decline inclusion of the deceased into research projects.
Samples from medicolegal autopsies performed between June 2013 and November 2016 with suspected use of medicinal or illegal drugs were collected by forensic pathologists during autopsy. The autopsies were in general performed within a week after death. PB, cardiac blood (CB), psoas muscle (PM), lateral vastus muscle (LVM), PF and VH samples were collected for the study along with routine samples of urine and PB. PB, with a volume of 40–50 mL, was collected in the following manner: The iliac vein was transected. After initial blood flow from the vessel stopped, a ladle was placed on the superior part of the cross-section, and blood from the femoral vein was “milked” on to the ladle by compression to the inner thigh and groin. Thus, no blood from the caval vein or superior part of inguinal vein was collected. To enable collection of 40–50 mL, collection from both sides was necessary in a majority of cases. The study protocol has been described previously ( ). After arrival at our laboratory, in the Department of Forensic Sciences at Oslo University Hospital, the samples were stored refrigerated (4°C) until analysis, with a storage time of typically 1–2 weeks, to allow for the screening of PB. For cases where benzodiazepines or z-hypnotics were detected in PB in the routine cases, all six matrices collected as project samples (including PB) were analyzed together to minimize differences due to instrument or storage time variation. The muscle analysis was performed in aliquots of muscle homogenate, made from ∼3.5 g of the muscle in 14 mL of water. Corrections due to dilution were performed using the following formula: [12pt]{minimal}
}{}$$ & { Concentration\ in\ muscle} \\ & = (({ weighted\ gram\ muscle} + { amount\ of\ water}) . \\ & . / { gram\ weighted\ muscle}) \\ & { concentration\ in\ homogenized\ sample} $$ The forensic pathologist estimated the degree of autolysis, using a scale from 0 to 2, where 0 equaled no putrefaction, 1 slight putrefaction and 2 moderate putrefaction.
PB collected for the case work was analyzed for ethanol, psychoactive medicinal drugs and common drugs of abuse. The benzodiazepines and z-hypnotics included in this screening method were clonazepam, 7-aminoclonazepam, flunitrazepam, 7-aminoflunitrazepam, nitrazepam, 7-aminonitrazepam, diazepam, nordiazepam, oxazepam, midazolam, bromazepam, lorazepam, phenazepam, zopiclone and zolpidem, and the screening was performed with a UHPLC–MS-MS method ( ) using an acidic mobile phase. Confirmation was performed with three different ultra high performance liquid chromatography--tandem mass spectrometry (UHPLC–MS-MS) methods using alkaline mobile phases. The methods for benzodiazepines and z-hypnotics have been published elsewhere ( , ), while the method for 7-amino metabolites can be found in the . Whole blood from the blood bank at Oslo University Hospital was used to prepare calibrator and quality control (QC) samples, and all matrices were quantified using the same calibration curve. Precision and accuracy for spiked QC samples in all six matrices calculated using whole blood calibration curves were tested in a separate experiment. To minimize differences in concentrations due to storage time or instrument variations, all matrices from each case were analyzed in the same series as far as possible. Spiked whole blood standards were used to prepare the calibration curve, however recovery and matrix effects were examined for all matrices. We report results above the limit of quantification (LOQ) of the confirmation methods ( ), while concentrations below LOQ were not included in the data evaluation.
We evaluated matrix effects using post-extraction addition as described by Matuszewski et al. ( ) at two concentration levels for all six matrices, using negative samples from five different autopsy cases. For zopiclone and zolpidem, one sample used to test for matrix effects and extraction recovery was evaluated by the pathologist to have slight putrefaction while the others had no putrefaction. For the benzodiazepines and 7-amino metabolites, three samples had no putrefaction and two samples had slight putrefaction. Three different sets were made: (A) blank extracted matrix samples spiked after extraction, (B) neat standards in mobile phase and (C) matrix samples spiked before extraction. Internal standard was added to all sets after extraction. Matrix effects were calculated by dividing the mean peak height for the samples spiked after extraction (A) with the mean peak height found for the neat solutions (B) and given as percentage: Matrix effect = (A/B)× 100%. Extraction recovery was found by comparing set C with set A: Recovery = (C/A)× 100%.
We performed data analysis using IBM SPSS Statistics for Windows version 23.0 and calculated concentration ratios for all the matrices relative to PB. Correlations between concentrations in the alternative matrices and PB were examined using Spearman’s rank correlation, with two-tailed evaluation of significance. P -values < 0.05 were considered statistically significant, and missing values were excluded pairwise.
The study was approved by the Regional Committee for Medical Research Ethics (reference number 2012/2173) and by the Higher Prosecuting Authority (reference number 2012/02455). The cases were not listed in the National Registry of Withdrawal from Biological Research Consent, where people can deny inclusion into biological research projects. Additionally, for cases included after the middle of 2015, the next of kin was given the opportunity to decline inclusion of the deceased into research projects.
We found 109 cases in our total of 183 cases with one or several of the benzodiazepines or z-hypnotics detected. All concentrations are given in . Clonazepam was the most frequently detected substance, followed by diazepam ( ). Detection of metabolites was accounted as ingestion of the parent compound except for oxazepam. Two to four compounds were found in 59 of the cases. The most common combination was intake of clonazepam together with diazepam, followed by clonazepam combined with alprazolam. Most cases showed no ( n = 77) or slight ( n = 24) putrefaction, while four cases showed a moderate degree of putrefaction. For four cases, no information about the degree of putrefaction was available. Matrix effects and recovery Matrix effects are shown in ; some ion suppression was evident, especially for the muscle samples, but this was in all cases corrected by the internal standard. The coefficients of variation (CVs) were <15% for all substances in all matrices. The recoveries are found in . For several compounds, somewhat lower extraction recoveries were found for the muscle samples. CVs were <20% except for diazepam in PB, which had a CV of 26%. Concentrations of benzodiazepines and z-hypnotics The median, minimum and maximum concentrations of the benzodiazepines and z-hypnotics in all matrices are shown in . Individual concentrations and degree of putrefaction are found in . The inclusion criterion for analysis in all study matrices was the finding of benzodiazepines or z-hypnotics in PB. For some compounds, the number of cases analyzed in an alternative matrix is higher than the number in PB. The initial inclusion is then due to a positive PB result of another compound in the analytical method. The concentration values in are based on values above LOQ, and values below LOQ were omitted. The measured muscle results in homogenate were evaluated against the LOQ, and only values above LOQ were corrected for dilution and included in the results. Benzodiazepines and z-hypnotics were detected in all matrices. However, systematically lower concentrations and a lower number of positive findings were seen in VH ( ), except for zopiclone, which was detected in VH in all the 16 cases. In addition, there were in general fewer findings in muscles. Correlation between concentrations measured in the alternative matrices and PB was investigated for substances with >10 cases ( ). We found statistically significant correlations for all these compounds. The oxazepam correlations are not given for VH, due to the low number of findings for this matrix. Correlations between concentrations in the alternative matrices and PB were medium to strong, with Spearman’s rank correlations ( ρ ) ranging from 0.51 to 0.99 as shown in . The precision and accuracy data from QC samples spiked in the six different matrices compared to calibrators in whole blood from Oslo University Blood Bank are found in . Results were within the ±20% recommended by the Scientific Working Group for Forensic Toxicology guidelines ( ) for most compounds and matrices. Flunitrazepam had a systematic negative bias of up to −24%, but the other matrices were comparable to PB. For diazepam, one deviating sample leads to deviation in precision and accuracy for the low QC sample for PB, heart blood and PF. The midazolam muscle samples had a positive bias of close to 30%. Ratios between the alternative matrices and PB shows median ratios between concentrations in CB, PF, muscles (PM and LVM) and VH compared to concentrations in PB. Three nitrobenzodiazepines were included in this study. Clonazepam showed a large variability between the different matrices ( ). Large variation was also seen for concentration ratios for the individual cases. Part of this large variation is due to some outliers with largely deviating ratios. The metabolite 7-aminoclonazepam had distinctively less variation in the median ratio to PB for the different matrices and had median ratios close to 1, except for VH which had a median ratio of 0.24. Flunitrazepam was only detected in two cases. No results were obtained for muscles or VH for flunitrazepam, while positive results were found for all matrices for 7-aminoflunitrazepam ( ). Just as for clonazepam, the median concentration ratios for nitrazepam in the PM and vastus muscle relative to PB were higher compared to the other alternative matrices. However, the results showed somewhat less variability than for clonazepam. As for flunitrazepam and clonazepam, the ratio between concentrations in muscle and PB was more similar to the other matrices for 7-amino metabolites. shows box plots of the concentration ratios for clonazepam, nitrazepam and their metabolites. Alprazolam had median concentration ratios close to 1 for all matrices compared to PB except VH ( ). For diazepam, the median concentration ratio for VH to PB was only 0.082, while the other median ratios varied between 0.77 and 1.7. As could be expected since the metabolite nordiazepam has very similar structure and characteristics, similar results were found. Oxazepam was found together with diazepam or nordiazepam in 5 cases and alone in 13 cases. The median oxazepam concentration ratios for the alternative matrices to PB varied between 1.1 and 1.4 ( ), while VH had a low ratio (0.11) as found for the other benzodiazepines. Only one case had a positive result for midazolam. shows box plots for the individual ratios for alprazolam, diazepam, nordiazepam and oxazepam. Although the median concentration ratios of alternative matrices to PB were close to 1 for the matrices except VH, individual cases showed a larger degree of variation. It can be noted that especially muscle concentrations for diazepam had a very large range. Zopiclone is the only compound in our study with a median concentration ratio of VH to PB close to 1 (0.97). The median ratios for zopiclone for the other alternative matrices compared to PB were all close to 1 ( ), however individual cases showed large variability. Zolpidem had a lower ratio for VH compared to PB, 0.21, and ratios ranging from 0.63 to 1.9 for the other matrices. With only four cases, evaluation of variations is difficult. Box plots for zopiclone and zolpidem ratios are found in .
Matrix effects are shown in ; some ion suppression was evident, especially for the muscle samples, but this was in all cases corrected by the internal standard. The coefficients of variation (CVs) were <15% for all substances in all matrices. The recoveries are found in . For several compounds, somewhat lower extraction recoveries were found for the muscle samples. CVs were <20% except for diazepam in PB, which had a CV of 26%.
The median, minimum and maximum concentrations of the benzodiazepines and z-hypnotics in all matrices are shown in . Individual concentrations and degree of putrefaction are found in . The inclusion criterion for analysis in all study matrices was the finding of benzodiazepines or z-hypnotics in PB. For some compounds, the number of cases analyzed in an alternative matrix is higher than the number in PB. The initial inclusion is then due to a positive PB result of another compound in the analytical method. The concentration values in are based on values above LOQ, and values below LOQ were omitted. The measured muscle results in homogenate were evaluated against the LOQ, and only values above LOQ were corrected for dilution and included in the results. Benzodiazepines and z-hypnotics were detected in all matrices. However, systematically lower concentrations and a lower number of positive findings were seen in VH ( ), except for zopiclone, which was detected in VH in all the 16 cases. In addition, there were in general fewer findings in muscles. Correlation between concentrations measured in the alternative matrices and PB was investigated for substances with >10 cases ( ). We found statistically significant correlations for all these compounds. The oxazepam correlations are not given for VH, due to the low number of findings for this matrix. Correlations between concentrations in the alternative matrices and PB were medium to strong, with Spearman’s rank correlations ( ρ ) ranging from 0.51 to 0.99 as shown in . The precision and accuracy data from QC samples spiked in the six different matrices compared to calibrators in whole blood from Oslo University Blood Bank are found in . Results were within the ±20% recommended by the Scientific Working Group for Forensic Toxicology guidelines ( ) for most compounds and matrices. Flunitrazepam had a systematic negative bias of up to −24%, but the other matrices were comparable to PB. For diazepam, one deviating sample leads to deviation in precision and accuracy for the low QC sample for PB, heart blood and PF. The midazolam muscle samples had a positive bias of close to 30%.
shows median ratios between concentrations in CB, PF, muscles (PM and LVM) and VH compared to concentrations in PB. Three nitrobenzodiazepines were included in this study. Clonazepam showed a large variability between the different matrices ( ). Large variation was also seen for concentration ratios for the individual cases. Part of this large variation is due to some outliers with largely deviating ratios. The metabolite 7-aminoclonazepam had distinctively less variation in the median ratio to PB for the different matrices and had median ratios close to 1, except for VH which had a median ratio of 0.24. Flunitrazepam was only detected in two cases. No results were obtained for muscles or VH for flunitrazepam, while positive results were found for all matrices for 7-aminoflunitrazepam ( ). Just as for clonazepam, the median concentration ratios for nitrazepam in the PM and vastus muscle relative to PB were higher compared to the other alternative matrices. However, the results showed somewhat less variability than for clonazepam. As for flunitrazepam and clonazepam, the ratio between concentrations in muscle and PB was more similar to the other matrices for 7-amino metabolites. shows box plots of the concentration ratios for clonazepam, nitrazepam and their metabolites. Alprazolam had median concentration ratios close to 1 for all matrices compared to PB except VH ( ). For diazepam, the median concentration ratio for VH to PB was only 0.082, while the other median ratios varied between 0.77 and 1.7. As could be expected since the metabolite nordiazepam has very similar structure and characteristics, similar results were found. Oxazepam was found together with diazepam or nordiazepam in 5 cases and alone in 13 cases. The median oxazepam concentration ratios for the alternative matrices to PB varied between 1.1 and 1.4 ( ), while VH had a low ratio (0.11) as found for the other benzodiazepines. Only one case had a positive result for midazolam. shows box plots for the individual ratios for alprazolam, diazepam, nordiazepam and oxazepam. Although the median concentration ratios of alternative matrices to PB were close to 1 for the matrices except VH, individual cases showed a larger degree of variation. It can be noted that especially muscle concentrations for diazepam had a very large range. Zopiclone is the only compound in our study with a median concentration ratio of VH to PB close to 1 (0.97). The median ratios for zopiclone for the other alternative matrices compared to PB were all close to 1 ( ), however individual cases showed large variability. Zolpidem had a lower ratio for VH compared to PB, 0.21, and ratios ranging from 0.63 to 1.9 for the other matrices. With only four cases, evaluation of variations is difficult. Box plots for zopiclone and zolpidem ratios are found in .
Contrasting to the widespread use of benzodiazepines and z-hypnotics, limited data exist on the possible use of alternative postmortem matrices to PB in forensic toxicological analyses. In our study, the results from 13 different substances in six different matrices are presented for a total of 109 cases, where one to four compounds were detected in each case. This substantially increases the foundation for use of other matrices than PB to evaluate postmortem concentrations for benzodiazepines and z-hypnotics. In general, good correlation was found between the PB concentrations and concentrations in CB and PF. Good correlation was likewise found for concentration in muscle samples and PB for many of the compounds, although large variations were seen for some of the concentration ratios. Fewer findings were observed in muscles, possibly due to the dilution of the samples during homogenization, as an increasing tendency for obtaining missing values for muscle samples can be seen for decreasing PB concentrations ( ). The high concentrations seen in several of the muscle samples might be due to the lipophilic properties of the compounds and binding to peripheral tissues. For all compounds except zopiclone, concentrations in VH were very low compared to concentrations in PB. This is in concordance with the degree of protein binding of the substances. Benzodiazepines and zolpidem have high degrees of protein binding; ∼80–90% (alprazolam, flunitrazepam and nitrazepam) and 90–99% (clonazepam, zolpidem, midazolam, oxazepam, nordiazepam and diazepam) while zopiclone is known to be less protein bound, 45% ( ). Nitrobenzodiazepines Nitrazepam and flunitrazepam are prescribed as hypnotic drugs/sedatives and clonazepam as an antiepileptic drug in Norway, and in addition they might be used illegally. A clear correlation between benzodiazepine-positive driving under the influence of drugs, cases and reported seizures has previously been reported ( ). Clonazepam was the most prevalent substance of all the benzodiazepines/z-hypnotics analyzed; close to 60% of the cases in our study had findings of this benzodiazepine in one or more matrices. Considerably higher median concentrations were found for the 7-amino metabolites than the parent compounds in all matrices ( ). Likewise, a summary of a large number of cases from Finland reported a substantially higher median value for 7-aminoclonazepam than clonazepam ( ). This could be due to metabolism and the possible degradation of the nitrobenzodiazepines to the 7-amino metabolites in vitro ( ). The median concentration ratio between the metabolite and parent compound is given in . The results show that 7-aminoclonazepam is detected in higher concentrations in all the investigated matrices, compared to the parent drug. Clonazepam is, however, detected in higher concentrations in the LVMs, compared to the other alternative matrices. shows box plots of the individual concentration ratios for clonazepam, nitrazepam and their metabolites. For clonazepam, outliers with higher ratios were evident for several matrices. Both clonazepam and nitrazepam have considerably higher median ratios for muscles than the other matrices, while for the 7-amino metabolites, less variation and median concentration ratios to PB closer to one were found. A parallel can be found in a case published by Moriya and Hashimoto ( ) where muscle to femoral blood concentration ratios for nitrazepam and 7-aminonitrazepam were 4.7 and 0.67, respectively. A muscle to blood concentration ratio of 0.94 for nitrazepam ( n = 1) has likewise been reported ( ). Our study parallels the reported concentrations for VH of approximately one-third of PB concentrations found by Robertson and Drummer ( ) for clonazepam, flunitrazepam, 7-aminoclonazepam and 7-aminonitrazepam, while our two positive results for nitrazepam gave lower concentration ratios between VH and femoral blood of 0.018 and 0.19, respectively. Findings in a single case ( ) were also low, 0.2 for nitrazepam and 0.02 for the metabolite. The delayed transfer to VH and additionally less putrefaction in the shielded compartment of the eye could contribute to variable ratios and should be kept in mind. Both nitrobenzodiazepines and the 7-amino metabolites in our study showed median concentration ratios close to 1 for CB and PF relative to PB. Ratios of 0.4 and 0.5 for nitrazepam and 7-aminonitrazepam in PF and 2.2 and 0.58 in CB compared to femoral blood have been found previously ( ) and a concentration ratio of 0.8 for a single clonazepam case in PF ( ). In our study, >44 paired values were found for PF or CB and PB for clonazepam and 7-aminoclonazepam. Some extreme values are found, however as seen from the box plots, the 25–75 percentile range is closer to one, ranging from 0.35 to 1.55. Consequently, PF and CB will, for most cases, give results within the same order of magnitude as PB. Other benzodiazepines As seen from and , low ratios of 0.07–0.35 were found for VH for diazepam, nordiazepam, oxazepam and alprazolam. For the other matrices, the median concentration ratio values close to 1 and fairly narrow ranges can be observed. In their meta-analysis, Ketola and Kriikku reported that the concentration ratios of other blood compared to PB were 0.95 for diazepam ( n = 11) and 1.29 for a single oxazepam case, while VH to PB ratios were 0.26 for alprazolam ( n = 2) and 1.0 for diazepam ( n = 1) ( ). The ratio for diazepam in muscle was 0.56 ( n = 9). Garriot published muscle and blood results for four cases. From these results one can calculate muscle to blood ratios of 0.2–1.8 and 0.4–1.5 for diazepam and nordiazepam, respectively. Another case report gave ratios of 0.56 and 1.0 for diazepam and nordiazepam ( ). Based on concentration results published by Álvarez-Freire and coworkers for 12 cases we can calculate PF to blood ratios between 0.2 and 0.67 (diazepam, three cases), 0.25 and 1.35 (oxazepam, four cases) and 0.45 (midazolam, one case). For nordiazepam, six cases had values between 0.45 and 1.2, while one case had a very deviating value of 15. These values are in line with our findings. A case report found a ratio between VH and CB of 0.28 ( ) for alprazolam, close to our median value of 0.35. The very low values for VH to PB concentration ratios found for diazepam, nordiazepam and oxazepam in our study are in line with what we could expect due to the high degree of protein binding. Reported literature values vary and are, however, in some cases much higher. Scott and Oliver found median ratios between VH and blood concentrations of 0.8 and 0.9 based on 11 and 13 cases for diazepam and nordiazepam, respectively. Metushi et al. found mean concentration values of 0.37 and 0.13 in blood and VH, respectivley, which would correspond to a ratio of 0.35 ( ). A study including 19 cases of diazepam and nordiazepam in VH and femoral blood performed by Holmgren et al. does, in addition, have results in line with our findings ( ). The median concentration ratios to PB were close to 1 for diazepam in all matrices, except VH. Individual cases showed a larger variation and especially muscle concentrations for diazepam. The concentration ratio between muscle and PB will be affected by the homogenization and dilution of the samples, and for low concentrations, variations will affect the ratios more. In cases where the concentrations in PB are higher compared to muscle samples, time of ingestion before death might be relevant. Comparing the corresponding nordiazepam concentrations in PB in these cases, we can however not conclude that diazepam had been ingested within a short time prior to death. The concentration ratios for nordiazepam between PM and PB show less variation compared to the corresponding diazepam ratios. If one looks at box plots in , the 25 and 75 percentile ratio values fall between 0.55 and 2.1 for CB, PF and the two muscles for all four compounds and are in line with the limited existing literature. This leads us to conclude that these four matrices, to a larger degree than we have previously considered, may give an indication of the PB concentration. Concentrations in VH are very low for the most protein bound substances and might be less suited for evaluation of these benzodiazepines. Z-hypnotics Zopiclone is a commonly used drug in Norway, with defined daily dose of 28.6 in 2015 ( ). Zolpidem is less frequently used, which was reflected in the low number of cases. Zaleplone is not prescribed in Norway and was not included in the study. shows box plots of the ratio between concentrations in the other matrices and PB. For both z-hypnotics, the median concentrations ratios are close to 1, except for the ratio of zolpidem in VH to PB, which was 0.21. A few case reports describe findings of zopiclone in postmortem matrices. Ratios between concentrations in CB and VH relative to femoral blood were found to be 1.6 and 0.37 by van Bocxlaer et al. ( ). Pounder and Davies found a concentration ratio of 0.8–1 for heart blood relative to femoral blood for zopiclone and a ratio of 2.8 for PM to PB ( ). The case report found zopiclone to be stable in postmortem blood for 40 h, while long-term instability in blood has been demonstrated by other researchers ( , ). The tendency for higher values for PF and VH compared to PB that can be seen in could therefore indicate an increased stability compared to whole blood. From our results, one can infer that all the tested matrices could be used to gain insight into zopiclone intake in cases with lack of PB. Due to the low number of cases, clear conclusions are difficult to draw for zolpidem, but for our four cases, trends of lower concentration ratios for muscles and VH and equal or somewhat higher values for CB and PF can be found. This is in line with case reports finding median concentration ratio values of 0.37 for muscle ( n = 8) and 0.29 for VH ( n = 2) ( ) and 2.1 ( ), 1.4 ( ) and 0.84 ( n = 4) ( ) for CB when compared to PB. In the meta-analysis by Ketola and Kriikku, a ratio of 1.71 for unspecified other blood is given ( n = 3) ( ). Limitations While blood, muscle and PF were available for analysis from all cases, and muscle was unsuitable for analysis in only one case, the availability of VH was unfortunately limited. Most of the cases had multiple drug findings, and the distribution of VH between several methods was necessary. The analytical method for benzodiazepines uses 0.5 mL material, and in some cases, the sampled volume of VH was not sufficient for this method. Finding of the 7-amino metabolite could have prompted analysis by the clonazepam confirmation method, even though PB concentrations of clonazepam were reported as negative in the initial screening. For some benzodiazepines and zolpidem, the number of cases is low, making it difficult to draw conclusions. No information of ante-mortem blood concentrations was available to indicate which of the matrices represents the concentration at the time of death, and we do not know how postmortem changes influence the different matrices. As the inclusion aimed for cases with all matrices available, further studies are necessary to include cases with extended time between time of death and autopsy, severe putrefaction, injuries or burns. The use of whole blood to make calibrators used for all matrices instead of dedicated calibrators in all six matrices, as is the preferred method for forensic work, is a limitation to the study. As we are not permitted to sample these other matrices from routine cases to use for research, this unfortunately was the only option available. We did however use material from project cases negative for our compounds to evaluate recovery and matrix effects and test accuracy and precision for spiked QC samples. Based on the results in and , we believe that it is feasible to use whole blood calibration curves for the other matrices when applying stable isotope-labeled internal standards. This is in addition in line with other recent research that has demonstrated the validity of analyzing brain and muscle tissue samples using blood calibrators ( , ) For flunitrazepam and midazolam where the accuracy and precision study revealed a systematic bias for all matrices or muscle samples, respectively, the use of carbon 13-labeled internal standards might correct the results better than deuterated internal standards ( ).
Nitrazepam and flunitrazepam are prescribed as hypnotic drugs/sedatives and clonazepam as an antiepileptic drug in Norway, and in addition they might be used illegally. A clear correlation between benzodiazepine-positive driving under the influence of drugs, cases and reported seizures has previously been reported ( ). Clonazepam was the most prevalent substance of all the benzodiazepines/z-hypnotics analyzed; close to 60% of the cases in our study had findings of this benzodiazepine in one or more matrices. Considerably higher median concentrations were found for the 7-amino metabolites than the parent compounds in all matrices ( ). Likewise, a summary of a large number of cases from Finland reported a substantially higher median value for 7-aminoclonazepam than clonazepam ( ). This could be due to metabolism and the possible degradation of the nitrobenzodiazepines to the 7-amino metabolites in vitro ( ). The median concentration ratio between the metabolite and parent compound is given in . The results show that 7-aminoclonazepam is detected in higher concentrations in all the investigated matrices, compared to the parent drug. Clonazepam is, however, detected in higher concentrations in the LVMs, compared to the other alternative matrices. shows box plots of the individual concentration ratios for clonazepam, nitrazepam and their metabolites. For clonazepam, outliers with higher ratios were evident for several matrices. Both clonazepam and nitrazepam have considerably higher median ratios for muscles than the other matrices, while for the 7-amino metabolites, less variation and median concentration ratios to PB closer to one were found. A parallel can be found in a case published by Moriya and Hashimoto ( ) where muscle to femoral blood concentration ratios for nitrazepam and 7-aminonitrazepam were 4.7 and 0.67, respectively. A muscle to blood concentration ratio of 0.94 for nitrazepam ( n = 1) has likewise been reported ( ). Our study parallels the reported concentrations for VH of approximately one-third of PB concentrations found by Robertson and Drummer ( ) for clonazepam, flunitrazepam, 7-aminoclonazepam and 7-aminonitrazepam, while our two positive results for nitrazepam gave lower concentration ratios between VH and femoral blood of 0.018 and 0.19, respectively. Findings in a single case ( ) were also low, 0.2 for nitrazepam and 0.02 for the metabolite. The delayed transfer to VH and additionally less putrefaction in the shielded compartment of the eye could contribute to variable ratios and should be kept in mind. Both nitrobenzodiazepines and the 7-amino metabolites in our study showed median concentration ratios close to 1 for CB and PF relative to PB. Ratios of 0.4 and 0.5 for nitrazepam and 7-aminonitrazepam in PF and 2.2 and 0.58 in CB compared to femoral blood have been found previously ( ) and a concentration ratio of 0.8 for a single clonazepam case in PF ( ). In our study, >44 paired values were found for PF or CB and PB for clonazepam and 7-aminoclonazepam. Some extreme values are found, however as seen from the box plots, the 25–75 percentile range is closer to one, ranging from 0.35 to 1.55. Consequently, PF and CB will, for most cases, give results within the same order of magnitude as PB.
As seen from and , low ratios of 0.07–0.35 were found for VH for diazepam, nordiazepam, oxazepam and alprazolam. For the other matrices, the median concentration ratio values close to 1 and fairly narrow ranges can be observed. In their meta-analysis, Ketola and Kriikku reported that the concentration ratios of other blood compared to PB were 0.95 for diazepam ( n = 11) and 1.29 for a single oxazepam case, while VH to PB ratios were 0.26 for alprazolam ( n = 2) and 1.0 for diazepam ( n = 1) ( ). The ratio for diazepam in muscle was 0.56 ( n = 9). Garriot published muscle and blood results for four cases. From these results one can calculate muscle to blood ratios of 0.2–1.8 and 0.4–1.5 for diazepam and nordiazepam, respectively. Another case report gave ratios of 0.56 and 1.0 for diazepam and nordiazepam ( ). Based on concentration results published by Álvarez-Freire and coworkers for 12 cases we can calculate PF to blood ratios between 0.2 and 0.67 (diazepam, three cases), 0.25 and 1.35 (oxazepam, four cases) and 0.45 (midazolam, one case). For nordiazepam, six cases had values between 0.45 and 1.2, while one case had a very deviating value of 15. These values are in line with our findings. A case report found a ratio between VH and CB of 0.28 ( ) for alprazolam, close to our median value of 0.35. The very low values for VH to PB concentration ratios found for diazepam, nordiazepam and oxazepam in our study are in line with what we could expect due to the high degree of protein binding. Reported literature values vary and are, however, in some cases much higher. Scott and Oliver found median ratios between VH and blood concentrations of 0.8 and 0.9 based on 11 and 13 cases for diazepam and nordiazepam, respectively. Metushi et al. found mean concentration values of 0.37 and 0.13 in blood and VH, respectivley, which would correspond to a ratio of 0.35 ( ). A study including 19 cases of diazepam and nordiazepam in VH and femoral blood performed by Holmgren et al. does, in addition, have results in line with our findings ( ). The median concentration ratios to PB were close to 1 for diazepam in all matrices, except VH. Individual cases showed a larger variation and especially muscle concentrations for diazepam. The concentration ratio between muscle and PB will be affected by the homogenization and dilution of the samples, and for low concentrations, variations will affect the ratios more. In cases where the concentrations in PB are higher compared to muscle samples, time of ingestion before death might be relevant. Comparing the corresponding nordiazepam concentrations in PB in these cases, we can however not conclude that diazepam had been ingested within a short time prior to death. The concentration ratios for nordiazepam between PM and PB show less variation compared to the corresponding diazepam ratios. If one looks at box plots in , the 25 and 75 percentile ratio values fall between 0.55 and 2.1 for CB, PF and the two muscles for all four compounds and are in line with the limited existing literature. This leads us to conclude that these four matrices, to a larger degree than we have previously considered, may give an indication of the PB concentration. Concentrations in VH are very low for the most protein bound substances and might be less suited for evaluation of these benzodiazepines.
Zopiclone is a commonly used drug in Norway, with defined daily dose of 28.6 in 2015 ( ). Zolpidem is less frequently used, which was reflected in the low number of cases. Zaleplone is not prescribed in Norway and was not included in the study. shows box plots of the ratio between concentrations in the other matrices and PB. For both z-hypnotics, the median concentrations ratios are close to 1, except for the ratio of zolpidem in VH to PB, which was 0.21. A few case reports describe findings of zopiclone in postmortem matrices. Ratios between concentrations in CB and VH relative to femoral blood were found to be 1.6 and 0.37 by van Bocxlaer et al. ( ). Pounder and Davies found a concentration ratio of 0.8–1 for heart blood relative to femoral blood for zopiclone and a ratio of 2.8 for PM to PB ( ). The case report found zopiclone to be stable in postmortem blood for 40 h, while long-term instability in blood has been demonstrated by other researchers ( , ). The tendency for higher values for PF and VH compared to PB that can be seen in could therefore indicate an increased stability compared to whole blood. From our results, one can infer that all the tested matrices could be used to gain insight into zopiclone intake in cases with lack of PB. Due to the low number of cases, clear conclusions are difficult to draw for zolpidem, but for our four cases, trends of lower concentration ratios for muscles and VH and equal or somewhat higher values for CB and PF can be found. This is in line with case reports finding median concentration ratio values of 0.37 for muscle ( n = 8) and 0.29 for VH ( n = 2) ( ) and 2.1 ( ), 1.4 ( ) and 0.84 ( n = 4) ( ) for CB when compared to PB. In the meta-analysis by Ketola and Kriikku, a ratio of 1.71 for unspecified other blood is given ( n = 3) ( ).
While blood, muscle and PF were available for analysis from all cases, and muscle was unsuitable for analysis in only one case, the availability of VH was unfortunately limited. Most of the cases had multiple drug findings, and the distribution of VH between several methods was necessary. The analytical method for benzodiazepines uses 0.5 mL material, and in some cases, the sampled volume of VH was not sufficient for this method. Finding of the 7-amino metabolite could have prompted analysis by the clonazepam confirmation method, even though PB concentrations of clonazepam were reported as negative in the initial screening. For some benzodiazepines and zolpidem, the number of cases is low, making it difficult to draw conclusions. No information of ante-mortem blood concentrations was available to indicate which of the matrices represents the concentration at the time of death, and we do not know how postmortem changes influence the different matrices. As the inclusion aimed for cases with all matrices available, further studies are necessary to include cases with extended time between time of death and autopsy, severe putrefaction, injuries or burns. The use of whole blood to make calibrators used for all matrices instead of dedicated calibrators in all six matrices, as is the preferred method for forensic work, is a limitation to the study. As we are not permitted to sample these other matrices from routine cases to use for research, this unfortunately was the only option available. We did however use material from project cases negative for our compounds to evaluate recovery and matrix effects and test accuracy and precision for spiked QC samples. Based on the results in and , we believe that it is feasible to use whole blood calibration curves for the other matrices when applying stable isotope-labeled internal standards. This is in addition in line with other recent research that has demonstrated the validity of analyzing brain and muscle tissue samples using blood calibrators ( , ) For flunitrazepam and midazolam where the accuracy and precision study revealed a systematic bias for all matrices or muscle samples, respectively, the use of carbon 13-labeled internal standards might correct the results better than deuterated internal standards ( ).
With over 100 positive cases and application of the same analytical methods, comparable data for concentrations of benzodiazepines and z-hypnotics in six different postmortem matrices have been provided in this study. Benzodiazepines and z-hypnotics can be detected in all the six matrices investigated, however VH is a less suitable matrix for detecting benzodiazepines or zolpidem. The main reasons for this are the very low concentrations found in VH for highly protein bound substances as well as the limited volume available. The correlations between concentrations in VH and PB were better for zopiclone. A larger variability in the concentration ratios between the alternative matrices and PB was displayed for clonazepam, 7-aminoclonazepam, diazepam and zopiclone than the other compounds in this study. A large part of this variation was, however, due to a few outliers. Both CB and PF seem to be viable alternative matrices to PB, although a larger variation was seen for PF. A tendency for higher concentrations in PF could be observed, with the highest number of detected substances in total. For the nitrobenzodiazepines, the concentrations in the muscles were higher compared to PB, while for the 7-amino metabolites and the other compounds, concentrations in muscle revealed good correlation with PB. It is important to emphasize that interpretation of concentrations in different postmortem matrices must still be performed in the context of full case history, medical history, case circumstances and scene/autopsy findings.
bkac106_Supp Click here for additional data file.
|
Mitral valve infective endocarditis in a dialysis patient with a tunneled dialysis catheter and prior MitraClip® implantation: an autopsy case | 1590b616-aa3d-462a-a942-7b8efec68575 | 10037769 | Forensic Medicine[mh] | Cardiac device-related infective endocarditis (CDRIE) is an infection that has spread to the endocardium, including device leads and heart valves . The main routes of device infection are device pocket infection during implantation surgery and exposure of the lead or catheter to skin surfaces. CDRIE occurs when inflammation spreads from a device infection to the intracardiac space via an intravascular lead. Risk factors for CDRIE include renal failure (including hemodialysis patients), steroid use, congestive heart failure, hematoma formation in the device pocket, diabetes mellitus, and anticoagulant use [ – ]. Staphylococcus , especially coagulase-negative Staphylococcus (CNS), is the most common causative organism, accounting for more than half of all CDRIE cases . Device infections may be mixed infections caused by multiple causative organisms, and care must be taken to identify the causative organisms . As with other cases of infective endocarditis (IE), blood cultures and cardiac echocardiography are central to the diagnosis for CDRIE. The principle of treatment for CDRIE, with reference to the ESC guidelines 2015, is continuous administration of antimicrobial agents and complete removal of the device including the lead . In patients with CDRIE, the rate of IE recurrence is high and prognosis is poor when medical treatment alone is performed without device removal. In recent years, transcatheter aortic valve implantation ( TAVI) has been performed in clinical practice as a treatment for severe aortic stenosis in very old patients and those at high surgical risk . However, in patients who underwent TAVI, male gender, diabetes, and moderate to severe residual aortic regurgitation were showed to be significantly associated with an increased risk of IE, and it was also reported that patients who developed endocarditis had higher in-hospital and 2-year mortality rates . Percutaneous catheter mitral valve repair, which is minimally invasive like TAVI, has also been developed as a treatment for patients with inoperable or high-risk mitral regurgitation (MR). MitraClip® (Abbott Vascular, Santa Clara, CA, USA) therapy is currently the most popular therapy. MitraClip® therapy is very promising for secondary MR with heart failure [ – ]. There have been limited reports of CDRIE in patients treated with MitraClip® and even fewer reports of autopsy cases. In addition, there have been few case studies in which Elastica-Masson staining was performed to assess the degree of valve destruction caused by infective endocarditis. In this article, we report a very rare experience of treating a case of IE at the implantation site of MitraClip® and performing an autopsy on a patient who unfortunately died.
The patient, a 66-year-old male, had been undergoing outpatient hemodialysis treatment for one year for end-stage renal failure with autosomal dominant polycystic kidney disease. Transthoracic echocardiography performed at the start of hemodialysis showed a left ventricular ejection fraction of approximately 30% and severe MR from between the middle scallop of anterior leaflet (A2) and posterior leaflet (P2). The patient had functional MR due to valve ring enlargement and tethering. Tricuspid regurgitation was mild. There was no vegetation on the mitral valve or other valves. Hence, Bio-Flex® Tesio® Cath for long-term hemodialysis was inserted through the left subclavian vein and the patient was hemodialyzed using it as a blood access. The patient initially consulted the department of cardiology because of decreased left ventricular contraction shown by transthoracic echocardiography, shortness of breath, cardiac enlargement and pulmonary congestion on a chest X-ray. Although coronary angiography showed that there was no significant stenosis, transthoracic echocardiography revealed a shallow valve coaptation and marked progression of tethering MR due to left ventricular dilation. The tricuspid valve coaptation was also very shallow and tricuspid regurgitation was severe. According to the 2020 guidelines of the Japanese Circulation Society, the patient had secondary MR, which was very frail and difficult to operate on. Transesophageal ultrasonography showed that the mitral valve morphology was suitable for the MitraClip® procedure, and intervention was expected to relieve the symptoms. In the heart team conference with a cardiovascular surgeon, intervention with MitraClip® was recommended, and percutaneous catheter mitral valve repair was performed. A 1 Clip (MITRACLIP® NT Device) was implanted between the middle scallop of the anterior leaflet (A2) and the posterior leaflet (P2) of the mitral valve after preoperative administration of antibiotics to prevent surgical site infection. There were no intraoperative complications, including adverse events that impaired valve function. MR improved from severe to mild after mitral valve junction repair with placement of a single MitraClip® between the middle scallop of the anterior leaflet and the middle scallop of the posterior leaflet. The patient continued to have good fluid management on outpatient hemodialysis at another hospital after discharge from our hospital. However, 8 months after MitraClip® implantation, the patient had a fever of 38 °C or higher and hypotension, and dialysis could not be performed even with the use of continuous intravenous administration of noradrenaline during dialysis. He was referred to our hospital with suspicion of septic shock. MRSA was cultured from all blood cultures and also from the tip of the Bio-Flex® Tesio® Cath, which was used as a hemodialysis blood access. Transthoracic echocardiography performed in our hospital showed a large vegetation at the site of the MitraClip® implantation. The vegetation had extended to the medial site and lateral site and was mobile. MR was mild to moderate with outflow from both sides of the clip. (Fig. ). Head magnetic resonance imaging performed as a screening test also showed multiple cerebral infarctions. After admission, blood culture tests were performed, and methicillin‐resistant Staphylococcus aureus (MRSA) was cultured in all three sets. The diagnosis was CDRIE at the site of MitraClip® implantation, and the causative organism was MRSA based on the clinical course and examination. During the course of the disease, the patient developed disturbance of consciousness, which was suspected to be another cerebral infarction caused by embolization from the vegetation. Transthoracic echocardiography showed that the left ventricular ejection fraction had decreased to about 10%, and adequate water removal with dialysis treatment was difficult. A heart team conference was held with the cardiovascular surgeon, and it was decided to perform only conservative treatment with antibiotics, gentamicin and vancomycin, and not to perform surgery. On the 7th day, the patient died of peripheral circulatory failure due to septic shock. After his death, we received consent from his family to perform an autopsy. The autopsy specimen showed a large vegetation of 5 cm in size. The vegetation had spread to the left and right sides around the tip of the MitraClip® arm where it was joined (Fig. ). No vegetation was found on valves other than the mitral valve The autopsy pathology showed severe destruction of the mitral valve. Hematoxylin and eosin (H-E) staining showed marked infiltration of neutrophils in the mitral valve tissue at the site of contact with the verruca and a large amount of bacterial mass in the darkly stained area of the valve tissue. Gram staining of the pathology specimen showed positive cocci, a finding consistent with MRSA cultured by various culture tests. Elastica-Masson staining (a staining method that clearly shows connective tissues such as elastic fibers and collagen fibers) also showed a high degree of neutrophilic infiltration in the atrial and trabecular layers, and the valve structure was destroyed. The fibrous layer showed a high degree of neutrophilic infiltration, and all layers, including the myocardial layer, were destroyed (Fig. ).
The autopsy pathology showed severe destruction of the mitral valve. Hematoxylin and eosin (H-E) staining showed marked infiltration of neutrophils in the mitral valve tissue at the site of contact with the verruca and a large amount of bacterial mass in the darkly stained area of the valve tissue. Gram staining of the pathology specimen showed positive cocci, a finding consistent with MRSA cultured by various culture tests. Elastica-Masson staining (a staining method that clearly shows connective tissues such as elastic fibers and collagen fibers) also showed a high degree of neutrophilic infiltration in the atrial and trabecular layers, and the valve structure was destroyed. The fibrous layer showed a high degree of neutrophilic infiltration, and all layers, including the myocardial layer, were destroyed (Fig. ).
Although there have been many reports of CDRIE of native valves, prosthetic valves, valves after TAVI, pacemakers, implantable cardioverter-defibrillators, and biventricular pacemakers, there have been only a few reports of IE after MitraClip® implantation [ – ]. Furthermore, to the best of the authors' knowledge, there has been no autopsy report of CDRIE after MitraClip® implantation. There has also been no report showing the extent of tissue damage caused by IE in histopathology. Autopsy pathology findings also showed Gram-positive cocci. Clinical findings suggested that MRSA had entered the vessels percutaneously through the tunnel hemodialysis catheter or tunnel infection around the catheter, leading to bacteremia and IE. Prior to MitraClip® implantation, the patient had severe mitral regurgitation and poor fluid management. Tricuspid valve regurgitation was also functionally severe. In view of his general condition, a decision was made to perform MitraClip® implantation instead of surgical intervention. After MitraClip® implantation, the MR became mild and cardiac output increased. The problem of difficulty in hemodialysis was resolved and fluid management was good. Tricuspid regurgitation also improved from severe to mild. The clinical course demonstrated that IE occurred only in the mitral valve, which may have been related to the recent implantation of the MitraClip® . The MitraClip® was covered with neointima and the intrinsic mitral valve itself was strongly disrupted in the autopsy specimen. There was thick vegetation mainly on the middle scallop joined by the MitraClip®. The vegetation on the middle scallop had spread to the medial and lateral sides and also extended toward the left ventricle through the two unjoined and open native mitral valve entries. The synergistic effect of placing the Mitraclip® between the middle scallop of the anterior leaflet and the middle scallop of the posterior leaflet in a compromised host may have caused bacteria to adhere to the central area of the native mitral valve, creating an environment in which colonies could easily form and proliferate. H-E staining of the autopsy pathology specimen showed marked neutrophilic infiltration, and Gram-positive cocci were detected by Gram staining despite antibiotic treatment. Elastica-Masson staining showed a high degree of neutrophilic infiltration in the atrial and trabecular layers and a high degree of neutrophilic infiltration in the fibrous layer toward the valve cusps, indicating that the valve structure was destroyed in all layers including the ventricular layer . Our patient underwent the less invasive MitraClip® procedure for functional MR, which makes it difficult to maintain hemodynamics during dialysis. However, our patient had low cardiac function and was an easily infected host dialyzed with an indwelling dialysis catheter rather than an internal shunt, which predisposed him to CDRIE and IE. We hypothesized that the verruca was refractory to antibiotic therapy after the onset of IE and further increased in size. In addition, histopathology confirmed that there was significant tissue destruction due to IE. In conclusion, even in cases in which highly invasive treatment such as mitral valvuloplasty or mitral valve replacement is not feasible, the indication for even minimally invasive treatment with MitraClip® should be carefully considered in compromised hosts such as our patient. In addition, we should consult doctors in an infectious disease department and consider performing the MitraClip® procedure after screening for MRSA or antibiotic-resistant bacteria and administering appropriate antibiotic prophylaxis.
|
A health communication campaign for prevention of osteoporosis in rural elderly women | 6f71230d-5fbb-4798-8ace-36c20136a06d | 10037817 | Health Communication[mh] | Nowadays, osteoporosis, is an important global public health problem. It is usually asymptomatic and often presents as a clinically evident fracture, so it may impose an increasing physical and economic burden on patients and the society . Osteoporosis is a skeletal system disease characterized by low bone mass and structural deterioration of the bone tissue, resulting in increased bone fragility and risk of fractures . Osteoporosis can be classified into two categories, primary and secondary. Primary osteoporosis can occur in both sexes at any age with aging, while it considerably increases in postmenopausal women and gradually increases among men. Secondary osteoporosis may occur due to certain medications or diseases, such as hyperthyroidism and celiac disease . While Salari et al. (2021) in their meta-analysis study estimated that the prevalence of osteoporosis in women of the world was 23.1% , in the last Iranian Multi-center Osteoporosis Study (IMOS), the prevalence of osteoporosis in Iranian postmenopausal women was 37%; about 24.6%, 28.8%, and 5.6% of them suffered from osteoporosis in the femoral neck, spine, and total hip, respectively . Eghbali et al. (2022) in a meta-analysis, which included 26 recent studies in Iran, reported that the prevalence of osteoporosis and osteopenia in Iranian postmenopausal women was 33.70%, and 47.60%, respectively . Another meta-analysis including 30 articles published from 2005 to 2019 reported that the prevalence of osteoporosis in Iranian women over 60 years old was 34% . Women, who make up half of the world’s population are key and productive members of society, and their health affects the health of the subsequent generation. Therefore, preservation of their health is essential. On the other hand, osteoporosis is easily preventable through health education interventions . The risk factors of osteoporosis could be classified as non-modifiable (female sex, advancing age, heredity, and race), and modifiable (poor diet such as reduced calcium intake, insufficient physical activity, smoking, caffeine and alcohol drinking) . However, some studies revealed that the awareness, attitude, and practice of people toward these risk factors were not at desirable levels . The results of IMOS in Iran showed that 81.3% of the females had a poor knowledge of different aspects of osteoporosis and its complications . Thus, it seems necessary to increase health literacy of populations at risk by public health measures. One of the common health education interventions which made important contributions to the advancement of public health in health systems is health communication campaign . Most of these programs occur with a focus on health issues worldwide. Health communication campaigns (HCC) are “purposive attempts to inform or influence behaviors in large audience within a specified time period, using an organized set of communication activities and featuring specific messages in multiple channels to motivate behavior change in the individual and society” . Because previous studies have shown that osteoporosis is more prevalent in women living in rural areas than those in urban areas and since few studies conducted in Iran have examined the impact of HCCs in urban areas ; the present study aimed to evaluate the effects of an educational intervention which was designed and conducted based on HCC principles, on knowledge, attitude, and practice of elderly women towards osteoporosis in rural areas in Fasa, Iran. Design In this multi-stage mixed methods study, which lasted six months (June to December 2019) from the time of receiving the code of ethics to completion of data analysis, the intervention was conducted based on a guide entitled “Overview of Health Communication Campaigns” which was developed by The Health Communication Unit (THCU) at the Centre for Health Promotion University of Toronto (2007) . These stages were as follow: Review background information: Investigating the prevalence and the severity of osteoporosis and associated factors based on literature review (mentioned in the introduction section of this article). Set communication objectives: Three specific goals of this project included examining the impact of the communication campaigns on (1) knowledge, (2) attitude, and (3) practice of elderly women. Analyze and segment target audiences: This was done to determine socioeconomic status, lifestyle, and cultural and religious beliefs affecting health of the elderly. For this purpose, three focus groups with totally 7 local health staffs, eight old peoples and, 10 health volunteers were conducted. In this stage, the audience were categorized into primary audience (elderly women) and secondary audience (health volunteers, health staff, and family members of the elderly women). Channel analysis and selection of appropriate communication channel: The aim of this stage was to identify the preferred communication channels of the elderly and determine verbal and non-verbal communication barriers which may cause disruption in communication process at individual and group levels. Since several interpersonal and media channels should be used in HCCs, through consultation with the health education experts and health staff of the studied villages as well as reviewing the literature, three categories of communication channels were determined for message transmission to the elderly women, including Mass communication channels (pamphlets, billboards, posters, booklets), Interpersonal channels (Health volunteers, health staffs, physicians), and group channels (walking programs, lectures in mosques, exhibitions). Identify message concepts and pretest: According to the analysis of the audience and channel, proper messages were designed for each target group and channel. The main message and slogan of the campaign was “love our bones”, and “know how to prevent osteoporosis”. The contents of the messages were checked in terms of readability, comprehensibility and in accordance with the cultural values of the community in consultation with local health workers and 10 local trusted health volunteers. Create messages and materials: based on the results of step 5, 1000 pamphlets, 1000 leaflets, 500 booklets, and 15 billboards were produced, mainly for secondary audience, all of which were in the Persian language. A flip chart was designed for use in the educational sessions for the primary audience. It contained a picture or illustration on each paper sheet, so it communicated a single, distinct message in each page. Develop promotion plan: Planning for executive and educational activities was compiled for a 30 days’ period; the time schedule was developed, and one person in charge was appointed for each program. Activities included setting up a health exhibition for three days, in which health services such as face-to-face education and blood pressure, blood glucose level, and body mass index monitoring were provided for the elderly women, holding four 45-minute education sessions for the elderly (especially illiterate ones), in rural health homes, healthcare centers and local mosques, distribution of pamphlets and leaflets, installation of billboards in public places, implementing an education session for health volunteers and holding a one day walking plan for the public population of the villages. Implement communication strategies: Communication campaigns were launched for 30 days, based on the programs which were designed step 7, in the intervention group village. During this time, the control group received routine health education and health care services. Assess effects: The tools and methods of assessments will be explained in detail in the following section, and the related findings are presented in the results section of this paper. Sample Based on a similar study and considering 95% confidence level (α = 0.05), 90% power (1-β = 0.9), and 20% attrition rate, using the NCSS PASS 15 software, the sample size was computed at least 90 participants in each group. For the recruitment of the study population, at first from 55 total villages of Fasa city, seven villages which had at least 100 over 60-year-old female population were detected. After that, in order to ensure that the control group is not exposed to the campaign messages, we divided these villages into two groups based on their geographical location, so that each group was located in one of the geographic poles of the Fasa city with at least 60-kilometer distances. Then, from the villages of each pole, one village was randomly selected. All women aged 60–75 years within the selected villages were recruited for the study. In the next step, the pretest questionnaires were completed; then, the selected villages were allocated into two interventions (102 subjects) and control groups (106 subjects) using simple randomization. Three subjects in the control group and four in the intervention group were excluded from the study due to lost to follow-up (Fig. ). The inclusion criteria for the participants in the study were 60-75-year-old women, signing informed consent form to participate in the study, use of the Persian language, lack of severe and debilitating osteoporosis-related complications or other major limiting diseases, such as severe osteoarthritis and heart failure, ability to speak and communicate, and lack of severe cognitive or psychological problems. The participants were excluded from the study if they died, migrated, or traveled during the campaign. Measures Data were collected by a researcher-made questionnaire which consisted of two parts. In the first part, questions about the participants’ demographic information (age, education level), underlying chronic diseases (hypertension, diabetes, or cancer), and diagnosis of osteoporosis in themselves and their relatives were asked; in the second part, a researcher- made questionnaire regarding knowledge, attitude, and practice about the prevention and control of osteoporosis was used. The knowledge questionnaire consisted of 14 items with yes/ no / I don’t know scales (10 items for nutrition, 2 items for physical activity, and 2 items for smoking). Correct answers received score 1, incorrect and I don’t know answers received score 0. Attitude toward the effectiveness of osteoporosis prevention recommendations assessed by 11 items (6 items for nutrition, 3 items for physical activity, and 2 items for smoking) with a scale of low (score = 1), moderate (score = 2), high (score = 3), and seven items were designed for practice (5 items for nutrition, 1 item for physical activity, and 1 item for smoking). To measure the practice, we asked the participants to answer the question of how many days of the past week they did each of the seven behaviors (scores for each item were 0–7), including consumption of dairy products, fruits and vegetables, fish, and calcium and vitamin D supplements, regular walking, and avoidance of smoking. Before the intervention and two months after it, the questionnaires were completed by the participants at their home. If the participants did not understand the meaning of some questions, the researcher explained the meaning of those questions to the participants. For illiterate participants, the questionnaires were completed by a trained health care provider through 20 minutes’ face-to-face interviews. Validity and reliability Content and construct validity of the questionnaire were evaluated and confirmed through a panel of experts including seven health education specialists, a gerontologist, two health care providers, and two health volunteers from the selected villages. The internal consistency of the questionnaires was measured by Cronbach’s alpha, and values between 0.85 and 0.97 were obtained, indicating the acceptable reliability of the questionnaire. External reliability of the questionnaire was also evaluated by test-retest method over a two-week interval on a sample of 30 elderly women. Pearson correlation coefficient was obtained 0.73, indicating appropriate external reliability of the questionnaire. Analytic strategy All statistical analyses were performed using the SPSS 25. The normality assumption of the variables was assessed and confirmed by the Kolmogorov-Smirnov test (P > 0.2), descriptive and Chi-square analyses were used to assess the characteristics of the samples. In order to compare the mean scores of the variables, we used independent sample t-test for between group comparisons and paired t-test for within group comparisons. The significance level was set at < 0.05. In this multi-stage mixed methods study, which lasted six months (June to December 2019) from the time of receiving the code of ethics to completion of data analysis, the intervention was conducted based on a guide entitled “Overview of Health Communication Campaigns” which was developed by The Health Communication Unit (THCU) at the Centre for Health Promotion University of Toronto (2007) . These stages were as follow: Review background information: Investigating the prevalence and the severity of osteoporosis and associated factors based on literature review (mentioned in the introduction section of this article). Set communication objectives: Three specific goals of this project included examining the impact of the communication campaigns on (1) knowledge, (2) attitude, and (3) practice of elderly women. Analyze and segment target audiences: This was done to determine socioeconomic status, lifestyle, and cultural and religious beliefs affecting health of the elderly. For this purpose, three focus groups with totally 7 local health staffs, eight old peoples and, 10 health volunteers were conducted. In this stage, the audience were categorized into primary audience (elderly women) and secondary audience (health volunteers, health staff, and family members of the elderly women). Channel analysis and selection of appropriate communication channel: The aim of this stage was to identify the preferred communication channels of the elderly and determine verbal and non-verbal communication barriers which may cause disruption in communication process at individual and group levels. Since several interpersonal and media channels should be used in HCCs, through consultation with the health education experts and health staff of the studied villages as well as reviewing the literature, three categories of communication channels were determined for message transmission to the elderly women, including Mass communication channels (pamphlets, billboards, posters, booklets), Interpersonal channels (Health volunteers, health staffs, physicians), and group channels (walking programs, lectures in mosques, exhibitions). Identify message concepts and pretest: According to the analysis of the audience and channel, proper messages were designed for each target group and channel. The main message and slogan of the campaign was “love our bones”, and “know how to prevent osteoporosis”. The contents of the messages were checked in terms of readability, comprehensibility and in accordance with the cultural values of the community in consultation with local health workers and 10 local trusted health volunteers. Create messages and materials: based on the results of step 5, 1000 pamphlets, 1000 leaflets, 500 booklets, and 15 billboards were produced, mainly for secondary audience, all of which were in the Persian language. A flip chart was designed for use in the educational sessions for the primary audience. It contained a picture or illustration on each paper sheet, so it communicated a single, distinct message in each page. Develop promotion plan: Planning for executive and educational activities was compiled for a 30 days’ period; the time schedule was developed, and one person in charge was appointed for each program. Activities included setting up a health exhibition for three days, in which health services such as face-to-face education and blood pressure, blood glucose level, and body mass index monitoring were provided for the elderly women, holding four 45-minute education sessions for the elderly (especially illiterate ones), in rural health homes, healthcare centers and local mosques, distribution of pamphlets and leaflets, installation of billboards in public places, implementing an education session for health volunteers and holding a one day walking plan for the public population of the villages. Implement communication strategies: Communication campaigns were launched for 30 days, based on the programs which were designed step 7, in the intervention group village. During this time, the control group received routine health education and health care services. Assess effects: The tools and methods of assessments will be explained in detail in the following section, and the related findings are presented in the results section of this paper. Based on a similar study and considering 95% confidence level (α = 0.05), 90% power (1-β = 0.9), and 20% attrition rate, using the NCSS PASS 15 software, the sample size was computed at least 90 participants in each group. For the recruitment of the study population, at first from 55 total villages of Fasa city, seven villages which had at least 100 over 60-year-old female population were detected. After that, in order to ensure that the control group is not exposed to the campaign messages, we divided these villages into two groups based on their geographical location, so that each group was located in one of the geographic poles of the Fasa city with at least 60-kilometer distances. Then, from the villages of each pole, one village was randomly selected. All women aged 60–75 years within the selected villages were recruited for the study. In the next step, the pretest questionnaires were completed; then, the selected villages were allocated into two interventions (102 subjects) and control groups (106 subjects) using simple randomization. Three subjects in the control group and four in the intervention group were excluded from the study due to lost to follow-up (Fig. ). The inclusion criteria for the participants in the study were 60-75-year-old women, signing informed consent form to participate in the study, use of the Persian language, lack of severe and debilitating osteoporosis-related complications or other major limiting diseases, such as severe osteoarthritis and heart failure, ability to speak and communicate, and lack of severe cognitive or psychological problems. The participants were excluded from the study if they died, migrated, or traveled during the campaign. Data were collected by a researcher-made questionnaire which consisted of two parts. In the first part, questions about the participants’ demographic information (age, education level), underlying chronic diseases (hypertension, diabetes, or cancer), and diagnosis of osteoporosis in themselves and their relatives were asked; in the second part, a researcher- made questionnaire regarding knowledge, attitude, and practice about the prevention and control of osteoporosis was used. The knowledge questionnaire consisted of 14 items with yes/ no / I don’t know scales (10 items for nutrition, 2 items for physical activity, and 2 items for smoking). Correct answers received score 1, incorrect and I don’t know answers received score 0. Attitude toward the effectiveness of osteoporosis prevention recommendations assessed by 11 items (6 items for nutrition, 3 items for physical activity, and 2 items for smoking) with a scale of low (score = 1), moderate (score = 2), high (score = 3), and seven items were designed for practice (5 items for nutrition, 1 item for physical activity, and 1 item for smoking). To measure the practice, we asked the participants to answer the question of how many days of the past week they did each of the seven behaviors (scores for each item were 0–7), including consumption of dairy products, fruits and vegetables, fish, and calcium and vitamin D supplements, regular walking, and avoidance of smoking. Before the intervention and two months after it, the questionnaires were completed by the participants at their home. If the participants did not understand the meaning of some questions, the researcher explained the meaning of those questions to the participants. For illiterate participants, the questionnaires were completed by a trained health care provider through 20 minutes’ face-to-face interviews. Content and construct validity of the questionnaire were evaluated and confirmed through a panel of experts including seven health education specialists, a gerontologist, two health care providers, and two health volunteers from the selected villages. The internal consistency of the questionnaires was measured by Cronbach’s alpha, and values between 0.85 and 0.97 were obtained, indicating the acceptable reliability of the questionnaire. External reliability of the questionnaire was also evaluated by test-retest method over a two-week interval on a sample of 30 elderly women. Pearson correlation coefficient was obtained 0.73, indicating appropriate external reliability of the questionnaire. All statistical analyses were performed using the SPSS 25. The normality assumption of the variables was assessed and confirmed by the Kolmogorov-Smirnov test (P > 0.2), descriptive and Chi-square analyses were used to assess the characteristics of the samples. In order to compare the mean scores of the variables, we used independent sample t-test for between group comparisons and paired t-test for within group comparisons. The significance level was set at < 0.05. The mean age of the participants in the control and intervention groups was 65.70 ± 7.87 and 67.05 ± 5.33 years, respectively (p = 0.161). Table shows that demographic variables were not significantly different between the intervention and control groups. Knowledge The total knowledge mean score (possible range = 0–14) of all women who participated in the study was 7.85 ± 3.30. Based on inter-group comparisons, there were no significant differences in the mean scores of total knowledge, knowledge about nutrition (p = 0.447), physical activity (p = 0.126), and smoking (p = 0.157) prior to the intervention, between the control and intervention groups. However, after the intervention, the mean scores of total knowledge (p < 0.001), knowledge about nutrition (p < 0.001), physical activity (p = 0.038), and smoking (p = < 0.001) were significantly higher in the intervention group compared to the control group. Intra-group comparisons showed a significant increase in the mean score of total knowledge (p < 0.001), knowledge about nutrition (p < 0.001), physical activity (p = 0.001), and smoking (p = 0.001) in the intervention group after the intervention. However, there was no significant difference in the control group (Table ). Attitude The total attitude mean score (possible range = 11–33) of all women who participated in the study was 17.89 ± 3.75. Inter-group comparisons of the mean score of total attitude (P = 0.170) and its subscales including attitudes toward nutrition (p = 0.062), physical activity (p = 0.73), and smoking (p = 0.053) showed no significant differences between the two groups, before the intervention; however, the total mean score of attitude (P < 0.001), attitudes toward nutrition (p < 0.001), physical activity (p < 0.001), and smoking (p < 0.001) was significantly higher in the intervention group compared to the control group, after the intervention. Intra-group comparisons revealed that unlike the control group, the intervention group showed a significant increase in total mean scores of attitude (p < 0.001), attitude toward nutrition (p < 0.001), physical activity (p < 0.001), and smoking (p < 0.001) after the intervention (Table ). Practice The total attitude mean score (possible range = 0–49) of all women who participated in the study was 11.82 ± 5.38. As shown in Table , in inter-group analysis, there were no significant differences in the total mean score of osteoporosis prevention behaviors, practices in nutrition, physical activity, and smoking between the two study groups, both before and after the intervention (p > 0.05). In intra-group analysis, it was determined that, unlike the control group, the intervention group displayed a significant increase in the total mean scores of total practice (p < 0.001), nutritional practices (p < 0.001), and physical activity (p = 0.001) after the intervention; however, it was not enough to make a significant difference with the control group. According to the results of this study, no significant difference was observed in smoking behaviors in either intervention or control groups before and after the intervention. The total knowledge mean score (possible range = 0–14) of all women who participated in the study was 7.85 ± 3.30. Based on inter-group comparisons, there were no significant differences in the mean scores of total knowledge, knowledge about nutrition (p = 0.447), physical activity (p = 0.126), and smoking (p = 0.157) prior to the intervention, between the control and intervention groups. However, after the intervention, the mean scores of total knowledge (p < 0.001), knowledge about nutrition (p < 0.001), physical activity (p = 0.038), and smoking (p = < 0.001) were significantly higher in the intervention group compared to the control group. Intra-group comparisons showed a significant increase in the mean score of total knowledge (p < 0.001), knowledge about nutrition (p < 0.001), physical activity (p = 0.001), and smoking (p = 0.001) in the intervention group after the intervention. However, there was no significant difference in the control group (Table ). The total attitude mean score (possible range = 11–33) of all women who participated in the study was 17.89 ± 3.75. Inter-group comparisons of the mean score of total attitude (P = 0.170) and its subscales including attitudes toward nutrition (p = 0.062), physical activity (p = 0.73), and smoking (p = 0.053) showed no significant differences between the two groups, before the intervention; however, the total mean score of attitude (P < 0.001), attitudes toward nutrition (p < 0.001), physical activity (p < 0.001), and smoking (p < 0.001) was significantly higher in the intervention group compared to the control group, after the intervention. Intra-group comparisons revealed that unlike the control group, the intervention group showed a significant increase in total mean scores of attitude (p < 0.001), attitude toward nutrition (p < 0.001), physical activity (p < 0.001), and smoking (p < 0.001) after the intervention (Table ). The total attitude mean score (possible range = 0–49) of all women who participated in the study was 11.82 ± 5.38. As shown in Table , in inter-group analysis, there were no significant differences in the total mean score of osteoporosis prevention behaviors, practices in nutrition, physical activity, and smoking between the two study groups, both before and after the intervention (p > 0.05). In intra-group analysis, it was determined that, unlike the control group, the intervention group displayed a significant increase in the total mean scores of total practice (p < 0.001), nutritional practices (p < 0.001), and physical activity (p = 0.001) after the intervention; however, it was not enough to make a significant difference with the control group. According to the results of this study, no significant difference was observed in smoking behaviors in either intervention or control groups before and after the intervention. Health communication campaign is a type of media campaigns that seeks to promote public health by developing educational health interventions. The purpose of such campaigns is to increase the individual’s awareness about the impacts of diseases and to provide them with more information regarding prevention methods . The present study aimed to investigate the effect of an osteoporosis prevention campaign on knowledge, attitude, and practices of elderly women towards osteoporosis in Fasa, Iran. In the same line with the studies of Parandeh et al. (2019) in Iran , Oumer et al. (2020) in China , and Nohra et al. (2022) in Lebanon , in this study, the total knowledge score of the participants was assessed at a moderate level. However, Senthilraja et al. (2019) and Hussein & Wahdan (2021) reported that the knowledge scores of women were poor and very poor in India and Egypt, respectively. According to the results, educational interventions increased the mean score of knowledge about the prevention and control of osteoporosis in the intervention group compared to the control group. Consistent with our study, some other studies have also shown an increase in the knowledge of participants about osteoporosis following educational intervention in Iran and South Korea . The results about the mean attitude score of the participants showed that elderly women had a relatively favorable attitude towards the prevention of osteoporosis. Parandeh et al. (2019), reported that the perceived benefits of Iranian middle-aged (30–45 years old) women about the role of diet and physical activity in preventing osteoporosis were higher than moderate . The findings of Oumer et al. (2020) in China also revealed that more than 74%, 75%, and 48% of the participants believed that physical activity, diet, and smoking cessation were effective in prevention of osteoporosis, respectively. However, Ungan & Tumer (2001) stated that elderly women displayed no proper attitude towards prevention of osteoporosis . This difference may be due to the activities conducted through the past 20 years in Iran such as osteoporosis mounting public education campaings in world osteoporosis days, and and care services of several osteoporosis specialty clinics which have been developed in recent years . The mean score of attitudes about osteoporosis prevention behaviors in the intervention group increased significantly more than the control group in this study. Several studies in Iran such as those of Parandeh et al. (2019) , Shafieinia et al. (2016) , and Hazavehi et al. (2007) and studies in other countries including Parrott et al. in the United States (2008), and Pinar and Pinar (2020) in a rural population in Turkey also showed the positive effect of educational intervention on the attitudes and motivations of the participants towards osteoporosis prevention. Consistent with the study of Nohra et al. (2022) in Lebanon , the findings of the present study revealed that the mean score of the participants’ practices toward preventing osteoporosis was relatively low. This is inconsistent with the findings of Ahn and Oh (2021) in South Korea which reported that nutritional behaviors and physical activities in women at early old age were at moderate level and above. On the other hand, while, consistent with the studies of Ahn and Oh (2021) and Parandeh et al. (2019) , the results of our study indicated that the mean score of overall practices as well as the participants’ practices in adherence to nutritional and physical activity advices to prevent osteoporosis increased significantly in the intervention group. In line with some other studies , this increase was not enough to make a significant difference with the control group. Furthermore, educational intervention in the present study did not significantly change tobacco use; however, given that the participants in the study were women, and a few of them (3.5%) had reported tobacco use at baseline, there was little expectation about changing their practice in this regard. Limitations This is one of the few studies that was designed and implemented completely based on the principles of conducting health communication campaigns, especially for osteoporosis, but it had some limitations, one of the most important of which is that most of the participants were illiterate and the study audience was only rural elderly women, so the results may be not generalizable to the urban, male and more literate populations. Since the recommended behaviors of the participants were not observed by researchers, and their adherence to these behaviors was assessed using self-reporting questionnaire, it should be considered as another limitation of the study. This is one of the few studies that was designed and implemented completely based on the principles of conducting health communication campaigns, especially for osteoporosis, but it had some limitations, one of the most important of which is that most of the participants were illiterate and the study audience was only rural elderly women, so the results may be not generalizable to the urban, male and more literate populations. Since the recommended behaviors of the participants were not observed by researchers, and their adherence to these behaviors was assessed using self-reporting questionnaire, it should be considered as another limitation of the study. In conclusion, the results of the present study suggest that health communication campaign improved knowledge, attitude, and practice of the participants towards osteoporosis. Given that the primary target group in this study was the elderly women, one of the most important challenges in this study was communicating information to participants who almost all of whom were illiterate or low-literate. The results of this study showed that in developing countries such as Iran, where most of the elderly people are illiterate, developing and implementing educational programs in the form of health communication campaigns using different educational methods, engaging secondary audience who influence the low-literate elderly, and developing educational materials appropriate to educational levels of audience through careful analysis of the audience can be an effective way to improve their knowledge, attitude and behavior. On the other hand, the designed questionnaire was a good tool to collect the necessary information in this field in the population with low education level, and the participants communicated well with it, so it can be used in similar studies. |
Suture-tape augmentation of anterior cruciate ligament reconstruction: a prospective, randomised controlled trial (STACLR) | c856e540-c2fb-4b2f-9815-2774ca8fa60f | 10037835 | Suturing[mh] | Note: the numbers in curly brackets in this protocol refer to SPIRIT checklist item numbers. The order of the items has been modified to group similar items (see http://www.equator-network.org/reporting-guidelines/spirit-2013-statement-defining-standard-protocol-items-for-clinical-trials/ ).
Background and rationale {6a} The anterior cruciate ligament (ACL) acts as the primary static constraint to anterior translation and internal rotation of the tibia . ACL rupture is a common knee injury, affecting 0.03–3.67% of athletes annually , and most frequently the result of non-contact pivoting sporting injury . ACL rupture in a young person has been established to result in significant loss of participation in sport, the need for reconstructive surgery and prolonged rehabilitation , whilst long-term osteoarthritis risk can be increased . Historically, reconstruction with autograft has been the most common pathway to return to sport . Recent reports suggest, however, that return to sport rates are as low as 55% , whilst only 40% make a complete functional recovery . Secondary to this, graft failure remains a significant problem, particularly in at-risk populations such as younger and female athletes and those who yield smaller autograft widths at surgical reconstruction. Given these current limitations of ACL reconstruction, Nagelli and Hewett have suggested that return to sport should be delayed to up to 2 years postoperatively . Successful restoration of anterior tibial translatory stability with ACL reconstruction is dependent on the capacity of the ACL graft to withstand appropriate loads during rehabilitation, and the return to the sporting period. As such, synthetic augmentation of native autograft or indeed wholesale replacement of autograft has been of interest for many years. Primarily, these augments were the Ligament Augmentation and Reconstruction System (LARS) prosthesis, the Kennedy Ligament Augmentation Device (LAD), the Leeds-Keio device, the Dacron device and GORE-TEX devices and have been primarily utilised as complete substitutes for autograft , with their use as graft augments also reported . However, failure rates are up to 33%, particularly in earlier devices, whilst more recent devices such as the LARS ligament have been reported to be associated with higher rates of persistent effusion and foreign body synovitis . Recently, the use of suture tape (a modified suture composed of non-absorbable, braided polyethylene/polyester suture) acting as a ‘seatbelt’ for in situ autograft has been proposed [ – ]. It differs from historical graft augments in its ‘addition’ to autograft rather than replacement of it, and its successful use in ligamentous reconstruction and repair in other joints is encouraging [ – ]. Notably, the use of suture tape may improve initial graft stability and protect the graft prior to ligamentisation. Significantly, it has been studied to exhibit improved biomechanical stability relative to non-augmented grafts [ – ]. Additionally, animal studies have reported equivalent functional, arthroscopic and histological findings between augmented and unaugmented knees [ – ]. Importantly, rates of persistent synovitis, rates of graft incorporation and incidence of adverse histological findings are not reported to differ between suture-tape augmented grafts and control grafts among animal models . A recent human model report of 36 ACL reconstructions with suture tape augmentation recorded no persistent joint effusion or evidence of synovitis . However, prospective clinical data is lacking. A 2019 retrospective comparative study of 60 knees exhibited reduced postoperative pain, as well as improved change from baseline PROMs at 1 year after all-inside ACLR with suture-tape augmentation . However, this study is limited both in its retrospective design and its lack of objective assessment of knee stability. In contrast to this study, Parkes et al. have reported on 36 suture tape-augmented reconstructions, finding no significant differences in rates of return to sport, mean IKDC scores or post-operative examination findings compared with a 2:1 matched cohort . Whilst these studies offer promise regarding possible benefits of suture tape augmentation, they lack a control group and complete clinical assessment to the standard 2-year mark post-operatively. Significantly, the influence of graft augmentation upon residual anterior knee laxity after reconstruction of the ACL has not been reported across the literature. This study aims to address these deficits. The primary research objective is to determine whether the use of suture-tape augmentation of a primary hamstring autograft ACLR improves post-operative residual knee laxity. At present, the literature surrounding suture tape augmentation of ACLR is primarily biomechanical and animal model-based, with two retrospective cohort, non-randomised studies reporting subjective outcomes published. This study aims to collect a higher level of evidence, as well as objective measures of post-operative knee laxity among suture tape ACLR patients. The null hypothesis of this study is that there will be no significant difference in side-to-side residual anterior knee laxity, as measured with the GNRB arthrometer, as compared with the contralateral knee at 2 years post-operatively. Secondary outcomes will be graft failure, complication rates, clinical examination findings and patient-reported outcome measures (PROMs) including Knee Osteoarthritis and Outcomes Score Quality of Life subscale (KOOS-QOL) , International Knee Documentation Committee scale (IKDC) , Marx activity scale , EQ-5D-5L scale and ACL-Return to Sport Index (ACL-RSI) scores measured at 2 years post-operatively. The study aims to investigate whether suture-tape augmented anterior cruciate ligament reconstruction has superior results to non-augmented reconstructions.
The anterior cruciate ligament (ACL) acts as the primary static constraint to anterior translation and internal rotation of the tibia . ACL rupture is a common knee injury, affecting 0.03–3.67% of athletes annually , and most frequently the result of non-contact pivoting sporting injury . ACL rupture in a young person has been established to result in significant loss of participation in sport, the need for reconstructive surgery and prolonged rehabilitation , whilst long-term osteoarthritis risk can be increased . Historically, reconstruction with autograft has been the most common pathway to return to sport . Recent reports suggest, however, that return to sport rates are as low as 55% , whilst only 40% make a complete functional recovery . Secondary to this, graft failure remains a significant problem, particularly in at-risk populations such as younger and female athletes and those who yield smaller autograft widths at surgical reconstruction. Given these current limitations of ACL reconstruction, Nagelli and Hewett have suggested that return to sport should be delayed to up to 2 years postoperatively . Successful restoration of anterior tibial translatory stability with ACL reconstruction is dependent on the capacity of the ACL graft to withstand appropriate loads during rehabilitation, and the return to the sporting period. As such, synthetic augmentation of native autograft or indeed wholesale replacement of autograft has been of interest for many years. Primarily, these augments were the Ligament Augmentation and Reconstruction System (LARS) prosthesis, the Kennedy Ligament Augmentation Device (LAD), the Leeds-Keio device, the Dacron device and GORE-TEX devices and have been primarily utilised as complete substitutes for autograft , with their use as graft augments also reported . However, failure rates are up to 33%, particularly in earlier devices, whilst more recent devices such as the LARS ligament have been reported to be associated with higher rates of persistent effusion and foreign body synovitis . Recently, the use of suture tape (a modified suture composed of non-absorbable, braided polyethylene/polyester suture) acting as a ‘seatbelt’ for in situ autograft has been proposed [ – ]. It differs from historical graft augments in its ‘addition’ to autograft rather than replacement of it, and its successful use in ligamentous reconstruction and repair in other joints is encouraging [ – ]. Notably, the use of suture tape may improve initial graft stability and protect the graft prior to ligamentisation. Significantly, it has been studied to exhibit improved biomechanical stability relative to non-augmented grafts [ – ]. Additionally, animal studies have reported equivalent functional, arthroscopic and histological findings between augmented and unaugmented knees [ – ]. Importantly, rates of persistent synovitis, rates of graft incorporation and incidence of adverse histological findings are not reported to differ between suture-tape augmented grafts and control grafts among animal models . A recent human model report of 36 ACL reconstructions with suture tape augmentation recorded no persistent joint effusion or evidence of synovitis . However, prospective clinical data is lacking. A 2019 retrospective comparative study of 60 knees exhibited reduced postoperative pain, as well as improved change from baseline PROMs at 1 year after all-inside ACLR with suture-tape augmentation . However, this study is limited both in its retrospective design and its lack of objective assessment of knee stability. In contrast to this study, Parkes et al. have reported on 36 suture tape-augmented reconstructions, finding no significant differences in rates of return to sport, mean IKDC scores or post-operative examination findings compared with a 2:1 matched cohort . Whilst these studies offer promise regarding possible benefits of suture tape augmentation, they lack a control group and complete clinical assessment to the standard 2-year mark post-operatively. Significantly, the influence of graft augmentation upon residual anterior knee laxity after reconstruction of the ACL has not been reported across the literature. This study aims to address these deficits. The primary research objective is to determine whether the use of suture-tape augmentation of a primary hamstring autograft ACLR improves post-operative residual knee laxity. At present, the literature surrounding suture tape augmentation of ACLR is primarily biomechanical and animal model-based, with two retrospective cohort, non-randomised studies reporting subjective outcomes published. This study aims to collect a higher level of evidence, as well as objective measures of post-operative knee laxity among suture tape ACLR patients. The null hypothesis of this study is that there will be no significant difference in side-to-side residual anterior knee laxity, as measured with the GNRB arthrometer, as compared with the contralateral knee at 2 years post-operatively. Secondary outcomes will be graft failure, complication rates, clinical examination findings and patient-reported outcome measures (PROMs) including Knee Osteoarthritis and Outcomes Score Quality of Life subscale (KOOS-QOL) , International Knee Documentation Committee scale (IKDC) , Marx activity scale , EQ-5D-5L scale and ACL-Return to Sport Index (ACL-RSI) scores measured at 2 years post-operatively. The study aims to investigate whether suture-tape augmented anterior cruciate ligament reconstruction has superior results to non-augmented reconstructions.
Primary objective The objective of this study is to compare residual anterior knee laxity after primary hamstring autograft ACLR with or without suture tape (ST) augmentation, as measured by the GNRB ligament arthrometer at 2 years post-operatively. The primary outcome measure will be the difference between operative and non-operative limbs (side-to-side), between groups, at 2 years. Secondary objectives The secondary objectives are to compare PROMs, complication rates, return to sport rates and examination findings between SA and ST ACLR. The PROMs assessed will include the Marx activity scale, ACL RSI, IKDC, the KOOS QOL and the EQ5D-5L scale. Complications assessed will include post-operative pain levels, graft failure rates, contralateral knee ACL rupture, return to theatre and findings including rates of particulate-related synovitis, sterile effusion, persistent effusion and symptomatic arthrofibrosis of the knee. Return to sport data will include rates of return to sport, timing of return to sport and rates of return to the previous level of the sport. Examination findings assessed will include the presence of effusion, knee range of motion and the presence and grade of the Lachman test, anterior drawer and pivot shift tests.
The objective of this study is to compare residual anterior knee laxity after primary hamstring autograft ACLR with or without suture tape (ST) augmentation, as measured by the GNRB ligament arthrometer at 2 years post-operatively. The primary outcome measure will be the difference between operative and non-operative limbs (side-to-side), between groups, at 2 years.
The secondary objectives are to compare PROMs, complication rates, return to sport rates and examination findings between SA and ST ACLR. The PROMs assessed will include the Marx activity scale, ACL RSI, IKDC, the KOOS QOL and the EQ5D-5L scale. Complications assessed will include post-operative pain levels, graft failure rates, contralateral knee ACL rupture, return to theatre and findings including rates of particulate-related synovitis, sterile effusion, persistent effusion and symptomatic arthrofibrosis of the knee. Return to sport data will include rates of return to sport, timing of return to sport and rates of return to the previous level of the sport. Examination findings assessed will include the presence of effusion, knee range of motion and the presence and grade of the Lachman test, anterior drawer and pivot shift tests.
This study is a randomised controlled single-blind interventional 2-arm parallel-group superiority trial utilising 1:1 allocation ratio comparing ACLR with suture tape augmentation (ST-ACLR) to standard hamstring ACLR technique (ACLR) with femoral adjustable cortical fixation and tibial interference screw. This study is a superiority study with the hypothesis that ACL reconstruction with suture tape augmentation is superior to standard ACL reconstruction technique. All patients meeting inclusion criteria will be randomised to a treatment arm in the operating room following confirmation of exclusion criteria at arthroscopy. All components of the study will be undertaken at a single institution. All patients will be followed up to 2 years post-operatively.
Study setting {9} This study will be conducted across three subsidiary hospitals, within a single major metropolitan Australian academic hospital (site details available on ANZ Trial Registry). All patients will be recruited from the adult orthopaedic outpatient clinic. A total of 66 participants will be recruited and randomised to a treatment arm. Owing to the nature of the randomisation model, greater than 66 patients may be recruited to reach the eventual target randomisation number. All data will be collected within Australia. Eligibility criteria {10} All inclusion criteria will be assessed by the treating surgeon and associate investigators. Participants can be included if they are: Waitlisted for ACLR with either of the associated investigators (S.T or L.B). Waitlisting is based on evidence of complete ACL rupture, based on clinical assessment and MRI imaging. Patients waitlisted will also have the appropriate lifestyle indications to warrant surgical reconstruction of the ligament. Able to give informed consent and to participate fully in the interventions and follow-up procedures. Aged 18 and over. Concomitant meniscal and/or osteochondral pathology can be included. Planned for surgery using an ipsilateral hamstring tendon autograft Patients will be excluded if they: Have had a previous ACL reconstruction on the ipsilateral knee. Have had a previous ACL injury to the non-operative knee. Are of a developmental age where the presence of open physes would otherwise alter the surgical technique utilised. Have grade 2 or 3 medial collateral ligament (MCL)/lateral collateral ligament (LCL) injury, associated posterior cruciate ligament (PCL)/ posterolateral corner (PLC) injury that requires surgical intervention. Have inflammatory arthritis. Are pregnant. Have an articular cartilage defect requiring treatment that would alter the post-operative rehabilitation protocol and timelines. Have a meniscal injury requiring treatment that would alter the post-operative rehabilitation protocol and timelines (i.e. meniscal root or bucket handle tear repair). Have an ACL re-rupture risk significant enough to warrant the addition of an osteotomy or deformity corrective procedure or lateral extraarticular tenodesis. Who will take informed consent? {26a} Participants will undergo assessment and provide written informed consent for ACLR surgery with their treating surgeon or one of the orthopaedic department’s trainee surgeons in consultation with the treating surgeon. Patients waitlisted for primary ACLR with either of the Associate Investigators (ST or LB) will be approached and offered participation in the study. Patients will either be approached in person or via phone after their appointment. All consent for trial involvement will be performed by an informed member of the research team. The patient will be advised that they have the right to privacy and any information obtained in connection with this project and that could identify them will remain strictly confidential. Information will only be disclosed with the patient’s permission, except as required by law. They will be informed that at the time that in the unlikely event that a patient is deceased, any original paper records kept will be treated according to standardised hospital policies. Given the intention-to-treat basis of this trial, data already collected will be included within the trial analysis. Information may be used in a deidentified manner within presentations and publications in peer-reviewed medical journals. In accordance with relevant Australian and/or Victorian privacy and other relevant laws, the patient will have the right to access the information collected and stored by the researchers. Patients will be informed that they are free to refuse participation, and if they decide or withdraw at any time, they will not compromise their future medical care. In either case, they will be given a physical or digital copy of the participant information and consent form (PICF) and offered the opportunity to ask questions of any member of the research team. They will be then sent (or given) a digital consent form via REDCAP to sign electronically, which will be then stored in our institution’s secure REDCap platform. A copy of the signed consent form will also be sent to the participant. Additional consent provisions for collection and use of participant data and biological specimens {26b} Participants will undergo extended consent for use of study information in ancillary studies emanating from this trial. Participants will be advised that the results of this study may be utilised in a de-identified manner within publications or scientific research presentations. Interventions Explanation for the choice of comparators {6b} This study will compare ACL reconstruction with hamstrings autograft with (ST-ACLR) and without (ACLR) suture tape augmentation. The suture tape will be looped through the proximal femoral button. Both procedures are standard care at our institution. The surgical technique has been decided at the discretion of the senior surgeons. Intervention description {11a} Graft fixation Fixation will be performed with an adjustable suspensory ACL TightRope® 2 RT device (Arthrex, Naples, FL, USA) on the femoral side and a PEEK interference screw (Arthrex, Naples, FL, USA) on the tibial side with a diameter the same size as the tibial tunnel. Graft harvest and preparation The semitendinosus and gracilis tendons will be harvested from the ipsilateral knee with a tendon harvester. Each tendon will be doubled over the ACL TightRope® 2 RT device (Arthrex, Naples, FL, USA) and the free ends sutured together with 2-Fibreloop suture (Arthrex, Naples, FL, USA) to create a 4-strand hamstring graft. If the graft diameter is less than 7mm in females or 7.5 mm in males, both tendons will be tripled to create a 6-strand construct. The length of the graft may be whipstitched with a 1 Vicryl suture to tubularise and compress the graft. Femoral tunnel preparation An anatomic single-bundle reconstruction will be performed. The ACL femoral footprint will be identified and a point between the AM and PL bundles will be selected, erring towards the AM bundle. The position will be confirmed from the medial portal. A Spade Tip guide wire (Arthrex, Naples, FL, USA) will be passed via the anteromedial portal with the knee in maximum flexion and the tunnel length measured off the Spade Tip drill guide. A femoral reamer corresponding to the graft femoral diameter will be drilled via the anteromedial portal over the SpadeTip wire (Arthrex, Naples, FL, USA) to a depth of 25 to 30mm. A shuttling suture will be placed. Tibia tunnel preparation The tibial footprint will be identified and any residual tibial stump will be preserved where possible. An Arthrex ACL aimer set to 55° will be passed via the medial portal to facilitate the passage of a guide wire through the mid-point of the tibial ACL attachment. The full tibial tunnel will be created with a cylindrical reamer corresponding with the graft tibial diameter . Graft passage, tensioning and fixation technique The femoral button sutures will be passed through the femoral tunnel using the shuttling suture. The cortical suspensory button will be passed and confirmed to have flipped with a toggle test and reference to a marking on the tightrope corresponding to the femoral tunnel length and a second mark 7 mm closer to the graft. The graft will then be docked into the femoral tunnel, 5 mm short of the tunnel depth by shortening the adjustable loop. Two markings on the graft corresponding to the tunnel depth and 5 mm short of the tunnel depth as measured from the femoral end of the graft will be used to assess the amount docked into the femoral tunnel. The tibial side of the graft is then cycled 15 times with maximal manual tension applied. Following this, a proprietary “fish scale” tensioning device (Arthrex Inc, Naples, FL) is used to apply tension to the graft at 80N and an interference screw (PEEK; Arthrex Inc.) is placed over a nitinol wire with the knee at 0° of flexion with a posterior force on the tibia. The knee is then cycled through range of motion to ensure full range. The graft re-tensioned from the femoral side at 30° of flexion. Suture tape augmentation technique Patients randomised to the SA arm will use the Tightrope 2 femoral cortical button, which has a single strand of Fibertape looped through the proximal Tightrope to be run alongside the graft to serve as an augment. The femoral button and graft will then be docked into the femoral tunnel, 5 mm short of the tunnel depth by shortening the adjustable loop in a similar fashion to the standard ACL reconstruction group. The tibial side of the graft is then cycled 15 times with maximal manual tension applied. Secondary tibial fixation of the suture tape will be performed before the tibial graft, with a 4.75-mm SwiveLock anchor in full extension. A guide pin was drilled 1.5 cm distal to the tibial tunnel to a depth of 20mm and overrreamed with a 3.4-mm diameter reamer and 4.75-mm tap. Ensuring the free ends of the Fibertape are separate from the tibial-sided graft sutures, a haemostat is then placed underneath the FiberTape limbs to ensure it is not tighter than the graft. The suture tape is then fixed with a 4.75-mm SwiveLock anchor with the knee in maximal knee extension. The graft will be tensioned at 0° of knee flexion with a proprietary “fish scale” tensioning device used to apply tension to the graft at 80N, and a posterior force on the tibia. The tibial side of the graft will then be secured with an interference screw (PEEK; Arthrex Inc.), with the suture tape running alongside the graft through the tibial tunnel, with tension on the graft only. The knee is then cycled through a range of motion to ensure the full range is present. In order to ensure greater tension on the graft than the tape, the femoral side of the graft is then re-tensioned to ensure fully docked within the femoral tunnel at 30° of knee flexion. Closure Closure is performed after irrigation and haemostasis in a layered fashion. Criteria for discontinuing or modifying allocated interventions {11b} Owing to the nature of the intervention, being a surgical technique and use of specific implant, and the timing following randomisation, it is not anticipated that any participants will be discontinued from allocated intervention following randomisation. If a participant suffers graft failure or contralateral ACL knee rupture, then analysis via GNRB assessment may be changed. Participants may withdraw from the study without citing a reason, but reasons may include the participant has chosen to withdraw, postoperative graft rupture or failure or the participant has experienced an adverse event. If a patient indicates that they wish to withdraw, this will be recorded in the REDCap project, and they will not be asked to complete any further research-only assessments. They will be reassured that they will continue to receive care and any subsequent management at Western Health. If a patient withdraws from the study, data collected to this point will remain within the study for analysis unless the participant specifically requests otherwise. Strategies to improve adherence to interventions {11c} All follow-up, short of one interaction, will be performed in conjunction with standard surgical care. All PROMs will be collected by easily accessible medium, including text message or email contact. This study will be co-ordinated by a dedicated research assistant in order to give participants the time to ask and have answered questions with regard to trial or concomitant care, as well as to ensure the workload of trial delivery is adequately met. The research team will remain in contact with all participants and encourage their attendance at standard care appointments where data collection will occur via SMS reminders and phone calls as required. Relevant concomitant care permitted or prohibited during the trial {11d} Post-operatively, all patients will undergo standardised medical and rehabilitative protocols according to the Fowler Kennedy Physiotherapy following ACL Reconstruction Protocol. All patients will receive day 1 physiotherapy and standard post-discharge management including wound review at 2 weeks, with clinical reviews at 6 weeks, 3 months and 12 months. Participants will be offered institutional physiotherapy or be free to pursue physiotherapy through an independent provider. Participants will not be asked to modify any of their medication or treatment relating to other medical conditions. Return to sport will not be permitted until criteria stipulated by the Fowler Kennedy protocol are met, as agreed by the senior investigators. As stipulated by the protocol, return to sport is initiated in a graduated format from the 6–9-month mark, provided the operative knee is without pain or effusion during, or after functional sports-related training drills. Lower extremity function scores (LEFS) should be 76 points or greater at this point in rehabilitation. The LEFS is a self-reported questionnaire used to evaluate the functional status of an individual with a lower extremity musculoskeletal dysfunction. The individual must also be able to demonstrate the appropriate strength and endurance needed for their specific sport. Provisions for post-trial care {30} All participants will be offered routine medical and surgical post-trial care commensurate with their condition at the discretion of the treating surgeon. There will be no compensation offered to participants either for their involvement in the study. This is outlined in the participant information and consent form. Outcomes {12} Primary outcome The primary outcome will be the side-to-side difference in anterior tibial translation as measured on the GNRB arthrometer at 2 years post-operatively, between groups. Maximum anteroposterior tibial translation at 134 N will be recorded on both the operative and non-operative knees pre-operatively and post-operatively at 3, 12 and 24 months. The goals of ACL reconstruction are to restore sagittal and rotatory stability to the knee and to prevent secondary injuries such as meniscal tears, and progression to osteoarthritis . Common means of assessing this is by comparison of pre- and post-operative subjective Lachman manoeuvre and anterior drawer; however, these are often imprecise and subjective . Consequently, devices to accurately measure anterior tibial translation have been developed, most popularly the KT-1000 ™ manual arthrometer (MEDmetric®, San Diego, USA) . However, this device has been criticised for poor inter- and intra-observer reliability, and as such, newer computerised models such as the GNRB (Genourob, Laval, France) have been developed, with improved reliability . Studies have utilised these arthrometers to compare between anterior cruciate ligament surgical techniques, such as graft types , assessing post-operative outcomes , as well as analysing residual joint laxity after ligament reconstruction . This outcome has been selected due to its ability to assess whether the addition of suture tape in ACLR results in knee stability that is more similar to the contralateral uninjured knee than standard ACLR technique. Anterior tibial translation and knee laxity may vary from individual to individual, and for this reason, side-to-side difference has been chosen as the primary outcome, allowing the contralateral knee to be a within-participant baseline measure, reducing the variability in outcome owing to between participant differences. Secondary outcomes Secondary outcomes are fourfold and will include (i) PROMs, (ii) return to sport rates, (iii) complications and (iv) examination findings. i) PROMs will be recorded preoperatively and post-operatively at 6 weeks, 3 months, 12 months and 24 months. PROMs collected will include the following: EQ5D-5L: The EuroQol EQ-5D-5L is a validated, generic, self-reported outcome measure covering five health domains that are used to facilitate the calculation of quality-adjusted life years (QALYs) in health economic evaluations. The original EQ-5D questionnaire contained three response options within each of five health domains (mobility, self-care, usual activities, pain/discomfort and anxiety/depression) . More recently, the EQ-5D-5 L has been developed to overcome problems with ceiling effects and to improve sensitivity . The 5 L version consists of the same five domains as the original but with five response options. The International Knee Documentation Committee (IKDC) score: The IKDC is a knee-specific patient outcome questionnaire often used after ACL reconstruction . It is a subjective tool that provides an overall patient score (0–100), often interpreted as a measure of function, where higher scores represent a greater function. It aims to address three categories: symptoms, activity and overall function, and has shown adequate test-retest reliability and good construct validity in patients with issues at the knee . Knee injury and Osteoarthritis Outcome Score Quality of Life sub-scale (KOOS-QOL): This subscale is a 4-item questionnaire designed to evaluate the knee-specific quality of life, has been utilised among ACL surgical trials and has been established to have validity, responsiveness and reliability in patients undergoing ACLR . ACL return to sport index (ACL-RSI): Psychological readiness to return to sport will be evaluated with the ACL-RSI. It is a 12-point questionnaire validated to evaluate a patient’s confidence and readiness to return to sport and exhibits high internal validity and structural consistency . Given the theoretical increased stability offered by suture tape augmentation, particularly in the early rehabilitative period, it follows that patients treated with this may experience improved perception of knee function and subsequent confidence in rehabilitation and sporting activity across the early postoperative period. Self-reported pain: We will record postoperative pain levels and long-term knee pain scores utilising 10-point visual analogue scale pain scores (VAS) where 0 is no pain and 10 is the worst pain imaginable. ii) Complications: We will record complications and adverse events including wound complications, infection rates, return to surgery, e.g. meniscal tears, graft failure, contralateral knee ACL rupture, residual effusion, rates of symptomatic arthrofibrosis of the knee and second look arthroscopy findings and the rates of sterile effusion. These will be recorded upon review of each patient’s medical record at 2 years. iii) Return to sport and activity: Return to sport and activity will be assessed via a self-reported questionnaire delivered concomitantly with PROMs preoperatively and post-operatively at 6 weeks, 3 months, 12 months and 24 months. Return to activity will be evaluated through the Marx activity scale, a four-item activity rating scale which has been shown to have excellent retest reliability in patients with knee injuries . Participants will be asked to record on a scale of 0–4 how often they are able to perform each of running, cutting, deceleration and pivoting in their most active state across the preceding 12 months. A maximum score is 16 points. Participants will also be asked about their dichotomous return to sport status, their expectations regarding timing of return to sport, return to preinjury level of sport and the time postoperatively to return to sporting activity. These questions, along with the full set of pre-operative and post-operative PROMS, are included in Additional file . iv) Examination findings Examination findings will be taken preoperatively at examination under anaesthesia and post-operatively at 6 weeks, 3 months and 12 months. Examination findings including the range of motion (ROM), the presence of knee effusion, Lachmans, pivot shift and anterior drawer testing will be recorded. All assessments will be undertaken by lead surgeons S.T, L.B. Participant timeline {13} Participants will be recruited from the orthopaedic outpatient clinic at our institution. Participants will be eligible for assessment if they are referred to either of the associated investigators’ clinic (L.B or S.T) and are consented for ACLR. Eligibility for trial participation based on inclusion and exclusion criteria will then be assessed. Patients who are eligible will be offered a discussion about the study and the opportunity to ask questions in line with best-practice informed consent. Data will be collected pre-operatively and post-operatively at four time points (6 weeks, 3 months, 1 year and 2 years) and will include arthrometric measure of anterior tibial translation on the GNRB arthrometer and PROM-based data, as well as return to sport data, complication rates and examination findings (Fig. ). Participants will then be asked to complete an online questionnaire about their injury, preoperative return to sport expectations and PROMs. Following this, they will be referred for routine preoperative rehabilitation and undergo preoperative GNRB assessment with a member of the research team. Demographic data will be extracted from their medical record with the participant’s consent at this time. Initial examination findings will be recorded at examination under anaesthesia preceding arthroscopy. Final evaluation of eligibility criteria will occur during arthroscopy. Those patients who have consented for involvement in the study and are deemed eligible during surgery will be randomised using the REDCap platform and allocated to one of the trial arms (ACLR or ST-ACLR) and blinded to the allocation (Fig. ). Postoperatively, all GNRB measurements will be taken at the same standard time points to routine surgical follow-up by a member of the research team, or a member of the physiotherapy unit familiar with GNRB protocols. All PROMs and return to sport data will be recorded via online surveys directly into the REDCAP database. Follow-up surveys will be conducted via automated email service at the corresponding time point. All subsequent examinations will be conducted by lead surgeons S.T and L.B with data entered directly into a study data collection report form (paper version), and then manually entered into the REDCAP database. There will be only one time point (2 years) at which patients will be required to attend follow-up (for GNRB testing) without the need for a standard clinical follow-up as per standard practice; otherwise, patients will be within the hospital for their standard clinical follow-up at the time of testing. Sample size {14} All statistical methods and power analysis were developed through consultation with an independent statistician. Power analysis was performed with G*power 3.1 Software. To achieve a minimum 80% statistical power with an alpha value of 0.05, power analysis based on a hypothesised mean residual side-to-side difference of 2 mm of anteroposterior tibial translation yielded a sample size of n =48 (24 per group) utilising an expected independent t -test [ – ]. Although previously published KT-1000 data was utilised for power analysis, it has been shown that the GNRB exhibits similar absolute measures in side-to-side anterior tibial translation to KT-1000 utilised by experienced observers, whilst having reduced scatter around the mean . This would mean that utilising KT-1000 published data to establish a power analysis would be a conservative method, and likely over-estimate the required sample size to detect significant differences. Secondary power analysis based on 2-year IKDC total scores yielded a total sample size of 48 ( n =24 in each group) to detect significant differences with 80% power and an alpha value of 0.05 with an anticipated independent sample two-tailed t -test. This is based on a previous retrospective study of suture-tape ACL reconstruction which reported a mean change from baseline IKDC of 56 ± 17 in SA-augmented ACLR, compared with 38 ± 25 in conventional ACLR . Secondly, previous studies evaluating the IKDC which reported that a change score of 12.8 points would be sufficient to detect differences between those who have and have not had improved knee outcomes. Power analysis yielded a sample size of 48. Dropout rates are expected to be 30% based on institutional experience, owing to loss to follow-up and contralateral ACL rupture and graft failure, and as such 66 patients will be recruited. This is expected to take approximately 1–2 years of recruitment based on historical institutional caseloads. Recruitment {15} Recruitment will be conducted continuously across patients eligible until the target randomised sample size is achieved. Based on institutional caseloads, it is anticipated this will take 1–2 years. A dedicated member of the research team will be tasked with recruitment on clinic days, such that all prospective participants are given appropriate opportunity to understand the study before and during involvement.
This study will be conducted across three subsidiary hospitals, within a single major metropolitan Australian academic hospital (site details available on ANZ Trial Registry). All patients will be recruited from the adult orthopaedic outpatient clinic. A total of 66 participants will be recruited and randomised to a treatment arm. Owing to the nature of the randomisation model, greater than 66 patients may be recruited to reach the eventual target randomisation number. All data will be collected within Australia.
All inclusion criteria will be assessed by the treating surgeon and associate investigators. Participants can be included if they are: Waitlisted for ACLR with either of the associated investigators (S.T or L.B). Waitlisting is based on evidence of complete ACL rupture, based on clinical assessment and MRI imaging. Patients waitlisted will also have the appropriate lifestyle indications to warrant surgical reconstruction of the ligament. Able to give informed consent and to participate fully in the interventions and follow-up procedures. Aged 18 and over. Concomitant meniscal and/or osteochondral pathology can be included. Planned for surgery using an ipsilateral hamstring tendon autograft Patients will be excluded if they: Have had a previous ACL reconstruction on the ipsilateral knee. Have had a previous ACL injury to the non-operative knee. Are of a developmental age where the presence of open physes would otherwise alter the surgical technique utilised. Have grade 2 or 3 medial collateral ligament (MCL)/lateral collateral ligament (LCL) injury, associated posterior cruciate ligament (PCL)/ posterolateral corner (PLC) injury that requires surgical intervention. Have inflammatory arthritis. Are pregnant. Have an articular cartilage defect requiring treatment that would alter the post-operative rehabilitation protocol and timelines. Have a meniscal injury requiring treatment that would alter the post-operative rehabilitation protocol and timelines (i.e. meniscal root or bucket handle tear repair). Have an ACL re-rupture risk significant enough to warrant the addition of an osteotomy or deformity corrective procedure or lateral extraarticular tenodesis.
Participants will undergo assessment and provide written informed consent for ACLR surgery with their treating surgeon or one of the orthopaedic department’s trainee surgeons in consultation with the treating surgeon. Patients waitlisted for primary ACLR with either of the Associate Investigators (ST or LB) will be approached and offered participation in the study. Patients will either be approached in person or via phone after their appointment. All consent for trial involvement will be performed by an informed member of the research team. The patient will be advised that they have the right to privacy and any information obtained in connection with this project and that could identify them will remain strictly confidential. Information will only be disclosed with the patient’s permission, except as required by law. They will be informed that at the time that in the unlikely event that a patient is deceased, any original paper records kept will be treated according to standardised hospital policies. Given the intention-to-treat basis of this trial, data already collected will be included within the trial analysis. Information may be used in a deidentified manner within presentations and publications in peer-reviewed medical journals. In accordance with relevant Australian and/or Victorian privacy and other relevant laws, the patient will have the right to access the information collected and stored by the researchers. Patients will be informed that they are free to refuse participation, and if they decide or withdraw at any time, they will not compromise their future medical care. In either case, they will be given a physical or digital copy of the participant information and consent form (PICF) and offered the opportunity to ask questions of any member of the research team. They will be then sent (or given) a digital consent form via REDCAP to sign electronically, which will be then stored in our institution’s secure REDCap platform. A copy of the signed consent form will also be sent to the participant.
Participants will undergo extended consent for use of study information in ancillary studies emanating from this trial. Participants will be advised that the results of this study may be utilised in a de-identified manner within publications or scientific research presentations.
Explanation for the choice of comparators {6b} This study will compare ACL reconstruction with hamstrings autograft with (ST-ACLR) and without (ACLR) suture tape augmentation. The suture tape will be looped through the proximal femoral button. Both procedures are standard care at our institution. The surgical technique has been decided at the discretion of the senior surgeons. Intervention description {11a} Graft fixation Fixation will be performed with an adjustable suspensory ACL TightRope® 2 RT device (Arthrex, Naples, FL, USA) on the femoral side and a PEEK interference screw (Arthrex, Naples, FL, USA) on the tibial side with a diameter the same size as the tibial tunnel. Graft harvest and preparation The semitendinosus and gracilis tendons will be harvested from the ipsilateral knee with a tendon harvester. Each tendon will be doubled over the ACL TightRope® 2 RT device (Arthrex, Naples, FL, USA) and the free ends sutured together with 2-Fibreloop suture (Arthrex, Naples, FL, USA) to create a 4-strand hamstring graft. If the graft diameter is less than 7mm in females or 7.5 mm in males, both tendons will be tripled to create a 6-strand construct. The length of the graft may be whipstitched with a 1 Vicryl suture to tubularise and compress the graft. Femoral tunnel preparation An anatomic single-bundle reconstruction will be performed. The ACL femoral footprint will be identified and a point between the AM and PL bundles will be selected, erring towards the AM bundle. The position will be confirmed from the medial portal. A Spade Tip guide wire (Arthrex, Naples, FL, USA) will be passed via the anteromedial portal with the knee in maximum flexion and the tunnel length measured off the Spade Tip drill guide. A femoral reamer corresponding to the graft femoral diameter will be drilled via the anteromedial portal over the SpadeTip wire (Arthrex, Naples, FL, USA) to a depth of 25 to 30mm. A shuttling suture will be placed. Tibia tunnel preparation The tibial footprint will be identified and any residual tibial stump will be preserved where possible. An Arthrex ACL aimer set to 55° will be passed via the medial portal to facilitate the passage of a guide wire through the mid-point of the tibial ACL attachment. The full tibial tunnel will be created with a cylindrical reamer corresponding with the graft tibial diameter . Graft passage, tensioning and fixation technique The femoral button sutures will be passed through the femoral tunnel using the shuttling suture. The cortical suspensory button will be passed and confirmed to have flipped with a toggle test and reference to a marking on the tightrope corresponding to the femoral tunnel length and a second mark 7 mm closer to the graft. The graft will then be docked into the femoral tunnel, 5 mm short of the tunnel depth by shortening the adjustable loop. Two markings on the graft corresponding to the tunnel depth and 5 mm short of the tunnel depth as measured from the femoral end of the graft will be used to assess the amount docked into the femoral tunnel. The tibial side of the graft is then cycled 15 times with maximal manual tension applied. Following this, a proprietary “fish scale” tensioning device (Arthrex Inc, Naples, FL) is used to apply tension to the graft at 80N and an interference screw (PEEK; Arthrex Inc.) is placed over a nitinol wire with the knee at 0° of flexion with a posterior force on the tibia. The knee is then cycled through range of motion to ensure full range. The graft re-tensioned from the femoral side at 30° of flexion. Suture tape augmentation technique Patients randomised to the SA arm will use the Tightrope 2 femoral cortical button, which has a single strand of Fibertape looped through the proximal Tightrope to be run alongside the graft to serve as an augment. The femoral button and graft will then be docked into the femoral tunnel, 5 mm short of the tunnel depth by shortening the adjustable loop in a similar fashion to the standard ACL reconstruction group. The tibial side of the graft is then cycled 15 times with maximal manual tension applied. Secondary tibial fixation of the suture tape will be performed before the tibial graft, with a 4.75-mm SwiveLock anchor in full extension. A guide pin was drilled 1.5 cm distal to the tibial tunnel to a depth of 20mm and overrreamed with a 3.4-mm diameter reamer and 4.75-mm tap. Ensuring the free ends of the Fibertape are separate from the tibial-sided graft sutures, a haemostat is then placed underneath the FiberTape limbs to ensure it is not tighter than the graft. The suture tape is then fixed with a 4.75-mm SwiveLock anchor with the knee in maximal knee extension. The graft will be tensioned at 0° of knee flexion with a proprietary “fish scale” tensioning device used to apply tension to the graft at 80N, and a posterior force on the tibia. The tibial side of the graft will then be secured with an interference screw (PEEK; Arthrex Inc.), with the suture tape running alongside the graft through the tibial tunnel, with tension on the graft only. The knee is then cycled through a range of motion to ensure the full range is present. In order to ensure greater tension on the graft than the tape, the femoral side of the graft is then re-tensioned to ensure fully docked within the femoral tunnel at 30° of knee flexion. Closure Closure is performed after irrigation and haemostasis in a layered fashion. Criteria for discontinuing or modifying allocated interventions {11b} Owing to the nature of the intervention, being a surgical technique and use of specific implant, and the timing following randomisation, it is not anticipated that any participants will be discontinued from allocated intervention following randomisation. If a participant suffers graft failure or contralateral ACL knee rupture, then analysis via GNRB assessment may be changed. Participants may withdraw from the study without citing a reason, but reasons may include the participant has chosen to withdraw, postoperative graft rupture or failure or the participant has experienced an adverse event. If a patient indicates that they wish to withdraw, this will be recorded in the REDCap project, and they will not be asked to complete any further research-only assessments. They will be reassured that they will continue to receive care and any subsequent management at Western Health. If a patient withdraws from the study, data collected to this point will remain within the study for analysis unless the participant specifically requests otherwise. Strategies to improve adherence to interventions {11c} All follow-up, short of one interaction, will be performed in conjunction with standard surgical care. All PROMs will be collected by easily accessible medium, including text message or email contact. This study will be co-ordinated by a dedicated research assistant in order to give participants the time to ask and have answered questions with regard to trial or concomitant care, as well as to ensure the workload of trial delivery is adequately met. The research team will remain in contact with all participants and encourage their attendance at standard care appointments where data collection will occur via SMS reminders and phone calls as required. Relevant concomitant care permitted or prohibited during the trial {11d} Post-operatively, all patients will undergo standardised medical and rehabilitative protocols according to the Fowler Kennedy Physiotherapy following ACL Reconstruction Protocol. All patients will receive day 1 physiotherapy and standard post-discharge management including wound review at 2 weeks, with clinical reviews at 6 weeks, 3 months and 12 months. Participants will be offered institutional physiotherapy or be free to pursue physiotherapy through an independent provider. Participants will not be asked to modify any of their medication or treatment relating to other medical conditions. Return to sport will not be permitted until criteria stipulated by the Fowler Kennedy protocol are met, as agreed by the senior investigators. As stipulated by the protocol, return to sport is initiated in a graduated format from the 6–9-month mark, provided the operative knee is without pain or effusion during, or after functional sports-related training drills. Lower extremity function scores (LEFS) should be 76 points or greater at this point in rehabilitation. The LEFS is a self-reported questionnaire used to evaluate the functional status of an individual with a lower extremity musculoskeletal dysfunction. The individual must also be able to demonstrate the appropriate strength and endurance needed for their specific sport.
This study will compare ACL reconstruction with hamstrings autograft with (ST-ACLR) and without (ACLR) suture tape augmentation. The suture tape will be looped through the proximal femoral button. Both procedures are standard care at our institution. The surgical technique has been decided at the discretion of the senior surgeons.
Graft fixation Fixation will be performed with an adjustable suspensory ACL TightRope® 2 RT device (Arthrex, Naples, FL, USA) on the femoral side and a PEEK interference screw (Arthrex, Naples, FL, USA) on the tibial side with a diameter the same size as the tibial tunnel. Graft harvest and preparation The semitendinosus and gracilis tendons will be harvested from the ipsilateral knee with a tendon harvester. Each tendon will be doubled over the ACL TightRope® 2 RT device (Arthrex, Naples, FL, USA) and the free ends sutured together with 2-Fibreloop suture (Arthrex, Naples, FL, USA) to create a 4-strand hamstring graft. If the graft diameter is less than 7mm in females or 7.5 mm in males, both tendons will be tripled to create a 6-strand construct. The length of the graft may be whipstitched with a 1 Vicryl suture to tubularise and compress the graft. Femoral tunnel preparation An anatomic single-bundle reconstruction will be performed. The ACL femoral footprint will be identified and a point between the AM and PL bundles will be selected, erring towards the AM bundle. The position will be confirmed from the medial portal. A Spade Tip guide wire (Arthrex, Naples, FL, USA) will be passed via the anteromedial portal with the knee in maximum flexion and the tunnel length measured off the Spade Tip drill guide. A femoral reamer corresponding to the graft femoral diameter will be drilled via the anteromedial portal over the SpadeTip wire (Arthrex, Naples, FL, USA) to a depth of 25 to 30mm. A shuttling suture will be placed. Tibia tunnel preparation The tibial footprint will be identified and any residual tibial stump will be preserved where possible. An Arthrex ACL aimer set to 55° will be passed via the medial portal to facilitate the passage of a guide wire through the mid-point of the tibial ACL attachment. The full tibial tunnel will be created with a cylindrical reamer corresponding with the graft tibial diameter . Graft passage, tensioning and fixation technique The femoral button sutures will be passed through the femoral tunnel using the shuttling suture. The cortical suspensory button will be passed and confirmed to have flipped with a toggle test and reference to a marking on the tightrope corresponding to the femoral tunnel length and a second mark 7 mm closer to the graft. The graft will then be docked into the femoral tunnel, 5 mm short of the tunnel depth by shortening the adjustable loop. Two markings on the graft corresponding to the tunnel depth and 5 mm short of the tunnel depth as measured from the femoral end of the graft will be used to assess the amount docked into the femoral tunnel. The tibial side of the graft is then cycled 15 times with maximal manual tension applied. Following this, a proprietary “fish scale” tensioning device (Arthrex Inc, Naples, FL) is used to apply tension to the graft at 80N and an interference screw (PEEK; Arthrex Inc.) is placed over a nitinol wire with the knee at 0° of flexion with a posterior force on the tibia. The knee is then cycled through range of motion to ensure full range. The graft re-tensioned from the femoral side at 30° of flexion. Suture tape augmentation technique Patients randomised to the SA arm will use the Tightrope 2 femoral cortical button, which has a single strand of Fibertape looped through the proximal Tightrope to be run alongside the graft to serve as an augment. The femoral button and graft will then be docked into the femoral tunnel, 5 mm short of the tunnel depth by shortening the adjustable loop in a similar fashion to the standard ACL reconstruction group. The tibial side of the graft is then cycled 15 times with maximal manual tension applied. Secondary tibial fixation of the suture tape will be performed before the tibial graft, with a 4.75-mm SwiveLock anchor in full extension. A guide pin was drilled 1.5 cm distal to the tibial tunnel to a depth of 20mm and overrreamed with a 3.4-mm diameter reamer and 4.75-mm tap. Ensuring the free ends of the Fibertape are separate from the tibial-sided graft sutures, a haemostat is then placed underneath the FiberTape limbs to ensure it is not tighter than the graft. The suture tape is then fixed with a 4.75-mm SwiveLock anchor with the knee in maximal knee extension. The graft will be tensioned at 0° of knee flexion with a proprietary “fish scale” tensioning device used to apply tension to the graft at 80N, and a posterior force on the tibia. The tibial side of the graft will then be secured with an interference screw (PEEK; Arthrex Inc.), with the suture tape running alongside the graft through the tibial tunnel, with tension on the graft only. The knee is then cycled through a range of motion to ensure the full range is present. In order to ensure greater tension on the graft than the tape, the femoral side of the graft is then re-tensioned to ensure fully docked within the femoral tunnel at 30° of knee flexion. Closure Closure is performed after irrigation and haemostasis in a layered fashion.
Fixation will be performed with an adjustable suspensory ACL TightRope® 2 RT device (Arthrex, Naples, FL, USA) on the femoral side and a PEEK interference screw (Arthrex, Naples, FL, USA) on the tibial side with a diameter the same size as the tibial tunnel.
The semitendinosus and gracilis tendons will be harvested from the ipsilateral knee with a tendon harvester. Each tendon will be doubled over the ACL TightRope® 2 RT device (Arthrex, Naples, FL, USA) and the free ends sutured together with 2-Fibreloop suture (Arthrex, Naples, FL, USA) to create a 4-strand hamstring graft. If the graft diameter is less than 7mm in females or 7.5 mm in males, both tendons will be tripled to create a 6-strand construct. The length of the graft may be whipstitched with a 1 Vicryl suture to tubularise and compress the graft.
An anatomic single-bundle reconstruction will be performed. The ACL femoral footprint will be identified and a point between the AM and PL bundles will be selected, erring towards the AM bundle. The position will be confirmed from the medial portal. A Spade Tip guide wire (Arthrex, Naples, FL, USA) will be passed via the anteromedial portal with the knee in maximum flexion and the tunnel length measured off the Spade Tip drill guide. A femoral reamer corresponding to the graft femoral diameter will be drilled via the anteromedial portal over the SpadeTip wire (Arthrex, Naples, FL, USA) to a depth of 25 to 30mm. A shuttling suture will be placed.
The tibial footprint will be identified and any residual tibial stump will be preserved where possible. An Arthrex ACL aimer set to 55° will be passed via the medial portal to facilitate the passage of a guide wire through the mid-point of the tibial ACL attachment. The full tibial tunnel will be created with a cylindrical reamer corresponding with the graft tibial diameter .
The femoral button sutures will be passed through the femoral tunnel using the shuttling suture. The cortical suspensory button will be passed and confirmed to have flipped with a toggle test and reference to a marking on the tightrope corresponding to the femoral tunnel length and a second mark 7 mm closer to the graft. The graft will then be docked into the femoral tunnel, 5 mm short of the tunnel depth by shortening the adjustable loop. Two markings on the graft corresponding to the tunnel depth and 5 mm short of the tunnel depth as measured from the femoral end of the graft will be used to assess the amount docked into the femoral tunnel. The tibial side of the graft is then cycled 15 times with maximal manual tension applied. Following this, a proprietary “fish scale” tensioning device (Arthrex Inc, Naples, FL) is used to apply tension to the graft at 80N and an interference screw (PEEK; Arthrex Inc.) is placed over a nitinol wire with the knee at 0° of flexion with a posterior force on the tibia. The knee is then cycled through range of motion to ensure full range. The graft re-tensioned from the femoral side at 30° of flexion.
Patients randomised to the SA arm will use the Tightrope 2 femoral cortical button, which has a single strand of Fibertape looped through the proximal Tightrope to be run alongside the graft to serve as an augment. The femoral button and graft will then be docked into the femoral tunnel, 5 mm short of the tunnel depth by shortening the adjustable loop in a similar fashion to the standard ACL reconstruction group. The tibial side of the graft is then cycled 15 times with maximal manual tension applied. Secondary tibial fixation of the suture tape will be performed before the tibial graft, with a 4.75-mm SwiveLock anchor in full extension. A guide pin was drilled 1.5 cm distal to the tibial tunnel to a depth of 20mm and overrreamed with a 3.4-mm diameter reamer and 4.75-mm tap. Ensuring the free ends of the Fibertape are separate from the tibial-sided graft sutures, a haemostat is then placed underneath the FiberTape limbs to ensure it is not tighter than the graft. The suture tape is then fixed with a 4.75-mm SwiveLock anchor with the knee in maximal knee extension. The graft will be tensioned at 0° of knee flexion with a proprietary “fish scale” tensioning device used to apply tension to the graft at 80N, and a posterior force on the tibia. The tibial side of the graft will then be secured with an interference screw (PEEK; Arthrex Inc.), with the suture tape running alongside the graft through the tibial tunnel, with tension on the graft only. The knee is then cycled through a range of motion to ensure the full range is present. In order to ensure greater tension on the graft than the tape, the femoral side of the graft is then re-tensioned to ensure fully docked within the femoral tunnel at 30° of knee flexion.
Closure is performed after irrigation and haemostasis in a layered fashion.
Owing to the nature of the intervention, being a surgical technique and use of specific implant, and the timing following randomisation, it is not anticipated that any participants will be discontinued from allocated intervention following randomisation. If a participant suffers graft failure or contralateral ACL knee rupture, then analysis via GNRB assessment may be changed. Participants may withdraw from the study without citing a reason, but reasons may include the participant has chosen to withdraw, postoperative graft rupture or failure or the participant has experienced an adverse event. If a patient indicates that they wish to withdraw, this will be recorded in the REDCap project, and they will not be asked to complete any further research-only assessments. They will be reassured that they will continue to receive care and any subsequent management at Western Health. If a patient withdraws from the study, data collected to this point will remain within the study for analysis unless the participant specifically requests otherwise.
All follow-up, short of one interaction, will be performed in conjunction with standard surgical care. All PROMs will be collected by easily accessible medium, including text message or email contact. This study will be co-ordinated by a dedicated research assistant in order to give participants the time to ask and have answered questions with regard to trial or concomitant care, as well as to ensure the workload of trial delivery is adequately met. The research team will remain in contact with all participants and encourage their attendance at standard care appointments where data collection will occur via SMS reminders and phone calls as required.
Post-operatively, all patients will undergo standardised medical and rehabilitative protocols according to the Fowler Kennedy Physiotherapy following ACL Reconstruction Protocol. All patients will receive day 1 physiotherapy and standard post-discharge management including wound review at 2 weeks, with clinical reviews at 6 weeks, 3 months and 12 months. Participants will be offered institutional physiotherapy or be free to pursue physiotherapy through an independent provider. Participants will not be asked to modify any of their medication or treatment relating to other medical conditions. Return to sport will not be permitted until criteria stipulated by the Fowler Kennedy protocol are met, as agreed by the senior investigators. As stipulated by the protocol, return to sport is initiated in a graduated format from the 6–9-month mark, provided the operative knee is without pain or effusion during, or after functional sports-related training drills. Lower extremity function scores (LEFS) should be 76 points or greater at this point in rehabilitation. The LEFS is a self-reported questionnaire used to evaluate the functional status of an individual with a lower extremity musculoskeletal dysfunction. The individual must also be able to demonstrate the appropriate strength and endurance needed for their specific sport.
All participants will be offered routine medical and surgical post-trial care commensurate with their condition at the discretion of the treating surgeon. There will be no compensation offered to participants either for their involvement in the study. This is outlined in the participant information and consent form.
Primary outcome The primary outcome will be the side-to-side difference in anterior tibial translation as measured on the GNRB arthrometer at 2 years post-operatively, between groups. Maximum anteroposterior tibial translation at 134 N will be recorded on both the operative and non-operative knees pre-operatively and post-operatively at 3, 12 and 24 months. The goals of ACL reconstruction are to restore sagittal and rotatory stability to the knee and to prevent secondary injuries such as meniscal tears, and progression to osteoarthritis . Common means of assessing this is by comparison of pre- and post-operative subjective Lachman manoeuvre and anterior drawer; however, these are often imprecise and subjective . Consequently, devices to accurately measure anterior tibial translation have been developed, most popularly the KT-1000 ™ manual arthrometer (MEDmetric®, San Diego, USA) . However, this device has been criticised for poor inter- and intra-observer reliability, and as such, newer computerised models such as the GNRB (Genourob, Laval, France) have been developed, with improved reliability . Studies have utilised these arthrometers to compare between anterior cruciate ligament surgical techniques, such as graft types , assessing post-operative outcomes , as well as analysing residual joint laxity after ligament reconstruction . This outcome has been selected due to its ability to assess whether the addition of suture tape in ACLR results in knee stability that is more similar to the contralateral uninjured knee than standard ACLR technique. Anterior tibial translation and knee laxity may vary from individual to individual, and for this reason, side-to-side difference has been chosen as the primary outcome, allowing the contralateral knee to be a within-participant baseline measure, reducing the variability in outcome owing to between participant differences. Secondary outcomes Secondary outcomes are fourfold and will include (i) PROMs, (ii) return to sport rates, (iii) complications and (iv) examination findings. i) PROMs will be recorded preoperatively and post-operatively at 6 weeks, 3 months, 12 months and 24 months. PROMs collected will include the following: EQ5D-5L: The EuroQol EQ-5D-5L is a validated, generic, self-reported outcome measure covering five health domains that are used to facilitate the calculation of quality-adjusted life years (QALYs) in health economic evaluations. The original EQ-5D questionnaire contained three response options within each of five health domains (mobility, self-care, usual activities, pain/discomfort and anxiety/depression) . More recently, the EQ-5D-5 L has been developed to overcome problems with ceiling effects and to improve sensitivity . The 5 L version consists of the same five domains as the original but with five response options. The International Knee Documentation Committee (IKDC) score: The IKDC is a knee-specific patient outcome questionnaire often used after ACL reconstruction . It is a subjective tool that provides an overall patient score (0–100), often interpreted as a measure of function, where higher scores represent a greater function. It aims to address three categories: symptoms, activity and overall function, and has shown adequate test-retest reliability and good construct validity in patients with issues at the knee . Knee injury and Osteoarthritis Outcome Score Quality of Life sub-scale (KOOS-QOL): This subscale is a 4-item questionnaire designed to evaluate the knee-specific quality of life, has been utilised among ACL surgical trials and has been established to have validity, responsiveness and reliability in patients undergoing ACLR . ACL return to sport index (ACL-RSI): Psychological readiness to return to sport will be evaluated with the ACL-RSI. It is a 12-point questionnaire validated to evaluate a patient’s confidence and readiness to return to sport and exhibits high internal validity and structural consistency . Given the theoretical increased stability offered by suture tape augmentation, particularly in the early rehabilitative period, it follows that patients treated with this may experience improved perception of knee function and subsequent confidence in rehabilitation and sporting activity across the early postoperative period. Self-reported pain: We will record postoperative pain levels and long-term knee pain scores utilising 10-point visual analogue scale pain scores (VAS) where 0 is no pain and 10 is the worst pain imaginable. ii) Complications: We will record complications and adverse events including wound complications, infection rates, return to surgery, e.g. meniscal tears, graft failure, contralateral knee ACL rupture, residual effusion, rates of symptomatic arthrofibrosis of the knee and second look arthroscopy findings and the rates of sterile effusion. These will be recorded upon review of each patient’s medical record at 2 years. iii) Return to sport and activity: Return to sport and activity will be assessed via a self-reported questionnaire delivered concomitantly with PROMs preoperatively and post-operatively at 6 weeks, 3 months, 12 months and 24 months. Return to activity will be evaluated through the Marx activity scale, a four-item activity rating scale which has been shown to have excellent retest reliability in patients with knee injuries . Participants will be asked to record on a scale of 0–4 how often they are able to perform each of running, cutting, deceleration and pivoting in their most active state across the preceding 12 months. A maximum score is 16 points. Participants will also be asked about their dichotomous return to sport status, their expectations regarding timing of return to sport, return to preinjury level of sport and the time postoperatively to return to sporting activity. These questions, along with the full set of pre-operative and post-operative PROMS, are included in Additional file . iv) Examination findings Examination findings will be taken preoperatively at examination under anaesthesia and post-operatively at 6 weeks, 3 months and 12 months. Examination findings including the range of motion (ROM), the presence of knee effusion, Lachmans, pivot shift and anterior drawer testing will be recorded. All assessments will be undertaken by lead surgeons S.T, L.B.
The primary outcome will be the side-to-side difference in anterior tibial translation as measured on the GNRB arthrometer at 2 years post-operatively, between groups. Maximum anteroposterior tibial translation at 134 N will be recorded on both the operative and non-operative knees pre-operatively and post-operatively at 3, 12 and 24 months. The goals of ACL reconstruction are to restore sagittal and rotatory stability to the knee and to prevent secondary injuries such as meniscal tears, and progression to osteoarthritis . Common means of assessing this is by comparison of pre- and post-operative subjective Lachman manoeuvre and anterior drawer; however, these are often imprecise and subjective . Consequently, devices to accurately measure anterior tibial translation have been developed, most popularly the KT-1000 ™ manual arthrometer (MEDmetric®, San Diego, USA) . However, this device has been criticised for poor inter- and intra-observer reliability, and as such, newer computerised models such as the GNRB (Genourob, Laval, France) have been developed, with improved reliability . Studies have utilised these arthrometers to compare between anterior cruciate ligament surgical techniques, such as graft types , assessing post-operative outcomes , as well as analysing residual joint laxity after ligament reconstruction . This outcome has been selected due to its ability to assess whether the addition of suture tape in ACLR results in knee stability that is more similar to the contralateral uninjured knee than standard ACLR technique. Anterior tibial translation and knee laxity may vary from individual to individual, and for this reason, side-to-side difference has been chosen as the primary outcome, allowing the contralateral knee to be a within-participant baseline measure, reducing the variability in outcome owing to between participant differences.
Secondary outcomes are fourfold and will include (i) PROMs, (ii) return to sport rates, (iii) complications and (iv) examination findings. i) PROMs will be recorded preoperatively and post-operatively at 6 weeks, 3 months, 12 months and 24 months. PROMs collected will include the following: EQ5D-5L: The EuroQol EQ-5D-5L is a validated, generic, self-reported outcome measure covering five health domains that are used to facilitate the calculation of quality-adjusted life years (QALYs) in health economic evaluations. The original EQ-5D questionnaire contained three response options within each of five health domains (mobility, self-care, usual activities, pain/discomfort and anxiety/depression) . More recently, the EQ-5D-5 L has been developed to overcome problems with ceiling effects and to improve sensitivity . The 5 L version consists of the same five domains as the original but with five response options. The International Knee Documentation Committee (IKDC) score: The IKDC is a knee-specific patient outcome questionnaire often used after ACL reconstruction . It is a subjective tool that provides an overall patient score (0–100), often interpreted as a measure of function, where higher scores represent a greater function. It aims to address three categories: symptoms, activity and overall function, and has shown adequate test-retest reliability and good construct validity in patients with issues at the knee . Knee injury and Osteoarthritis Outcome Score Quality of Life sub-scale (KOOS-QOL): This subscale is a 4-item questionnaire designed to evaluate the knee-specific quality of life, has been utilised among ACL surgical trials and has been established to have validity, responsiveness and reliability in patients undergoing ACLR . ACL return to sport index (ACL-RSI): Psychological readiness to return to sport will be evaluated with the ACL-RSI. It is a 12-point questionnaire validated to evaluate a patient’s confidence and readiness to return to sport and exhibits high internal validity and structural consistency . Given the theoretical increased stability offered by suture tape augmentation, particularly in the early rehabilitative period, it follows that patients treated with this may experience improved perception of knee function and subsequent confidence in rehabilitation and sporting activity across the early postoperative period. Self-reported pain: We will record postoperative pain levels and long-term knee pain scores utilising 10-point visual analogue scale pain scores (VAS) where 0 is no pain and 10 is the worst pain imaginable. ii) Complications: We will record complications and adverse events including wound complications, infection rates, return to surgery, e.g. meniscal tears, graft failure, contralateral knee ACL rupture, residual effusion, rates of symptomatic arthrofibrosis of the knee and second look arthroscopy findings and the rates of sterile effusion. These will be recorded upon review of each patient’s medical record at 2 years. iii) Return to sport and activity: Return to sport and activity will be assessed via a self-reported questionnaire delivered concomitantly with PROMs preoperatively and post-operatively at 6 weeks, 3 months, 12 months and 24 months. Return to activity will be evaluated through the Marx activity scale, a four-item activity rating scale which has been shown to have excellent retest reliability in patients with knee injuries . Participants will be asked to record on a scale of 0–4 how often they are able to perform each of running, cutting, deceleration and pivoting in their most active state across the preceding 12 months. A maximum score is 16 points. Participants will also be asked about their dichotomous return to sport status, their expectations regarding timing of return to sport, return to preinjury level of sport and the time postoperatively to return to sporting activity. These questions, along with the full set of pre-operative and post-operative PROMS, are included in Additional file . iv) Examination findings Examination findings will be taken preoperatively at examination under anaesthesia and post-operatively at 6 weeks, 3 months and 12 months. Examination findings including the range of motion (ROM), the presence of knee effusion, Lachmans, pivot shift and anterior drawer testing will be recorded. All assessments will be undertaken by lead surgeons S.T, L.B.
Participants will be recruited from the orthopaedic outpatient clinic at our institution. Participants will be eligible for assessment if they are referred to either of the associated investigators’ clinic (L.B or S.T) and are consented for ACLR. Eligibility for trial participation based on inclusion and exclusion criteria will then be assessed. Patients who are eligible will be offered a discussion about the study and the opportunity to ask questions in line with best-practice informed consent. Data will be collected pre-operatively and post-operatively at four time points (6 weeks, 3 months, 1 year and 2 years) and will include arthrometric measure of anterior tibial translation on the GNRB arthrometer and PROM-based data, as well as return to sport data, complication rates and examination findings (Fig. ). Participants will then be asked to complete an online questionnaire about their injury, preoperative return to sport expectations and PROMs. Following this, they will be referred for routine preoperative rehabilitation and undergo preoperative GNRB assessment with a member of the research team. Demographic data will be extracted from their medical record with the participant’s consent at this time. Initial examination findings will be recorded at examination under anaesthesia preceding arthroscopy. Final evaluation of eligibility criteria will occur during arthroscopy. Those patients who have consented for involvement in the study and are deemed eligible during surgery will be randomised using the REDCap platform and allocated to one of the trial arms (ACLR or ST-ACLR) and blinded to the allocation (Fig. ). Postoperatively, all GNRB measurements will be taken at the same standard time points to routine surgical follow-up by a member of the research team, or a member of the physiotherapy unit familiar with GNRB protocols. All PROMs and return to sport data will be recorded via online surveys directly into the REDCAP database. Follow-up surveys will be conducted via automated email service at the corresponding time point. All subsequent examinations will be conducted by lead surgeons S.T and L.B with data entered directly into a study data collection report form (paper version), and then manually entered into the REDCAP database. There will be only one time point (2 years) at which patients will be required to attend follow-up (for GNRB testing) without the need for a standard clinical follow-up as per standard practice; otherwise, patients will be within the hospital for their standard clinical follow-up at the time of testing.
All statistical methods and power analysis were developed through consultation with an independent statistician. Power analysis was performed with G*power 3.1 Software. To achieve a minimum 80% statistical power with an alpha value of 0.05, power analysis based on a hypothesised mean residual side-to-side difference of 2 mm of anteroposterior tibial translation yielded a sample size of n =48 (24 per group) utilising an expected independent t -test [ – ]. Although previously published KT-1000 data was utilised for power analysis, it has been shown that the GNRB exhibits similar absolute measures in side-to-side anterior tibial translation to KT-1000 utilised by experienced observers, whilst having reduced scatter around the mean . This would mean that utilising KT-1000 published data to establish a power analysis would be a conservative method, and likely over-estimate the required sample size to detect significant differences. Secondary power analysis based on 2-year IKDC total scores yielded a total sample size of 48 ( n =24 in each group) to detect significant differences with 80% power and an alpha value of 0.05 with an anticipated independent sample two-tailed t -test. This is based on a previous retrospective study of suture-tape ACL reconstruction which reported a mean change from baseline IKDC of 56 ± 17 in SA-augmented ACLR, compared with 38 ± 25 in conventional ACLR . Secondly, previous studies evaluating the IKDC which reported that a change score of 12.8 points would be sufficient to detect differences between those who have and have not had improved knee outcomes. Power analysis yielded a sample size of 48. Dropout rates are expected to be 30% based on institutional experience, owing to loss to follow-up and contralateral ACL rupture and graft failure, and as such 66 patients will be recruited. This is expected to take approximately 1–2 years of recruitment based on historical institutional caseloads.
Recruitment will be conducted continuously across patients eligible until the target randomised sample size is achieved. Based on institutional caseloads, it is anticipated this will take 1–2 years. A dedicated member of the research team will be tasked with recruitment on clinic days, such that all prospective participants are given appropriate opportunity to understand the study before and during involvement.
Sequence generation {16a} The randomisation design is a computer-generated permuted single-blind block randomisation. The randomisation sequence will be developed by an independent statistician, with all other research team members blinded to the randomisation sequence. Participants will be randomised using permuted block randomisation which allows for a better guarantee of equal-sized treatment groups whilst protecting against prediction of allocation towards the end of the recruitment period. Randomisation will be allocated in a 1:1 fashion, whereby patients will undergo either standard ACLR or suture tape-augmented ACLR. Randomisation will be stratified according to age greater than 25 years. Concealment mechanism {16b} The centrally managed, blinded randomisation model will ensure allocation concealment and prevent selection bias. The allocation sequence will be stored within the REDCap database and inaccessible to all but the statistician, ensuring concealment. Following randomisation, the allocation details will be displayed on the web-based system for each participant, and an automated email will be sent to members of the research team to ensure appropriate documentation in the medical record. Implementation {16c} The allocation sequence will be developed by an independent statistician. It will be integrated into the REDCap automated randomisation platform and will be blinded to all other investigators. Enrolment of participants will be overseen by the lead surgeons, and associate investigators ST and LB, and will be carried out by associate investigators LH and EN. Randomisation and assignment to the intervention arm will be performed intraoperatively, following assessment and management of concomitant cartilage and meniscal injuries by the associate investigators LB and ST.
The randomisation design is a computer-generated permuted single-blind block randomisation. The randomisation sequence will be developed by an independent statistician, with all other research team members blinded to the randomisation sequence. Participants will be randomised using permuted block randomisation which allows for a better guarantee of equal-sized treatment groups whilst protecting against prediction of allocation towards the end of the recruitment period. Randomisation will be allocated in a 1:1 fashion, whereby patients will undergo either standard ACLR or suture tape-augmented ACLR. Randomisation will be stratified according to age greater than 25 years.
The centrally managed, blinded randomisation model will ensure allocation concealment and prevent selection bias. The allocation sequence will be stored within the REDCap database and inaccessible to all but the statistician, ensuring concealment. Following randomisation, the allocation details will be displayed on the web-based system for each participant, and an automated email will be sent to members of the research team to ensure appropriate documentation in the medical record.
The allocation sequence will be developed by an independent statistician. It will be integrated into the REDCap automated randomisation platform and will be blinded to all other investigators. Enrolment of participants will be overseen by the lead surgeons, and associate investigators ST and LB, and will be carried out by associate investigators LH and EN. Randomisation and assignment to the intervention arm will be performed intraoperatively, following assessment and management of concomitant cartilage and meniscal injuries by the associate investigators LB and ST.
Who will be blinded {17a} Trial participants will be blinded to the allocation arm for the duration of the study. GNRB arthrometric measurements are inherently low in bias, secondary to the computer-generated application of force to the posterior tibia; however, technicians taking the measures will be blinded to allocation arm. Owing to the nature of examination assessments, the best technician to carry these out is the same as the treating surgeon and thus blinding is not appropriate in this setting. The questionnaire-based outcomes are patient-reported and therefore also assessor-blinded. Procedure for unblinding if needed {17b} As the surgeon and only outcome assessors are not blinded, unblinding will not be required.
Trial participants will be blinded to the allocation arm for the duration of the study. GNRB arthrometric measurements are inherently low in bias, secondary to the computer-generated application of force to the posterior tibia; however, technicians taking the measures will be blinded to allocation arm. Owing to the nature of examination assessments, the best technician to carry these out is the same as the treating surgeon and thus blinding is not appropriate in this setting. The questionnaire-based outcomes are patient-reported and therefore also assessor-blinded.
As the surgeon and only outcome assessors are not blinded, unblinding will not be required.
Plans for assessment and collection of outcomes {18a} All data will be entered into a custom web-based REDCap database accessible by study research staff only. All GNRB data will be transposed from automated Genuroub software outputs to custom-designed data entry forms in the preoperative and postoperative setting by a member of the research team specifically trained in both software. PROMs and return to sport data will be collected by a single questionnaire and entered directly via online surveys by the participant. All validated PROM scores are cited above and can be found in the original citation; all non-validated questionnaires are included in the Additional file of this document. Examination findings will be entered specifically into the medical record of each participant, and a study-specific data collection form and transposed into the web-based database. Plans to promote participant retention and complete follow-up {18b} PROMs and return to sport data will be collected by online survey, deliverable by email or text message with the aims to increase follow-up rates among a technologically inclined patient cohort. All follow-up examinations and GNRB assessments will be performed concurrently with follow-up surgical appointments until the 12-month mark in order to reduce the study burden on the individual. If a participant withdraws from the study, or suffers graft failure, no further data will be collected; however, data collected up until this point will be included in the study and absent data will be analyzed according to intention to treat principles. Data management {19} All data will be stored on a secure institutional REDCap server in a database custom-built for this study. Validation of data fields will be built into this platform where relevant, as well as the use of “required” fields to minimise missing data. Double data entry will not be possible. Confidentiality {27} The research information will be re-identifiable. All participants will be assigned a study ID. A data re-identification key file will be stored as an encrypted file separate to the file containing the data. This will be a password-protected file stored on the hospital server. Only the research team can match the participant’s name to their code number, if it is necessary to do so. The Principal Investigator will be responsible for the secure storage of the data collected in this project. Any hard copy data will be identified by a study ID number only and kept secure in a locked filing cabinet within a locked office. Only named researchers will have access to all data collected in this project. Electronic datasets will be stored securely within the institutional server as a password-protected excel file/on the institutional REDCap platform. Plans for collection, laboratory evaluation and storage of biological specimens for genetic or molecular analysis in this trial/future use {33} No biological specimens will be collected as part of this study.
All data will be entered into a custom web-based REDCap database accessible by study research staff only. All GNRB data will be transposed from automated Genuroub software outputs to custom-designed data entry forms in the preoperative and postoperative setting by a member of the research team specifically trained in both software. PROMs and return to sport data will be collected by a single questionnaire and entered directly via online surveys by the participant. All validated PROM scores are cited above and can be found in the original citation; all non-validated questionnaires are included in the Additional file of this document. Examination findings will be entered specifically into the medical record of each participant, and a study-specific data collection form and transposed into the web-based database.
PROMs and return to sport data will be collected by online survey, deliverable by email or text message with the aims to increase follow-up rates among a technologically inclined patient cohort. All follow-up examinations and GNRB assessments will be performed concurrently with follow-up surgical appointments until the 12-month mark in order to reduce the study burden on the individual. If a participant withdraws from the study, or suffers graft failure, no further data will be collected; however, data collected up until this point will be included in the study and absent data will be analyzed according to intention to treat principles.
All data will be stored on a secure institutional REDCap server in a database custom-built for this study. Validation of data fields will be built into this platform where relevant, as well as the use of “required” fields to minimise missing data. Double data entry will not be possible.
The research information will be re-identifiable. All participants will be assigned a study ID. A data re-identification key file will be stored as an encrypted file separate to the file containing the data. This will be a password-protected file stored on the hospital server. Only the research team can match the participant’s name to their code number, if it is necessary to do so. The Principal Investigator will be responsible for the secure storage of the data collected in this project. Any hard copy data will be identified by a study ID number only and kept secure in a locked filing cabinet within a locked office. Only named researchers will have access to all data collected in this project. Electronic datasets will be stored securely within the institutional server as a password-protected excel file/on the institutional REDCap platform.
No biological specimens will be collected as part of this study.
Statistical methods for primary and secondary outcomes {20a} Side-to-side difference in anterior tibial translation (the primary outcome), overall ACL-RSI, EQ5D-5L, Marx activity scales, IKDC and KOOS-QOL will be calculated as means, with measures of dispersion reported as standard deviation. Differences between groups will be reported as mean differences, with dispersion reported with confidence intervals. Data normality will be assessed with the Shapiro-Wilk test for continuous parametric assessment. Between-group comparisons, such as the primary outcome of mean difference in GNRB anteroposterior laxity at 2 years, will be assessed with independent samples t -tests. If data is found to be non-parametric in distribution, the Wilcoxon rank sum test will be utilised. Comparison of two dichotomous variables, such as return to sport rates and complication rates, will be performed with Fischer’s exact test. Repeated measures such as GNRB measures and mean PROMs will be assessed using mixed effects linear analysis. All tests are two-sided. A p value of <0.05 will be considered the cut-off for statistical significance. Interim analyses {21b} Interim analysis will be performed 1 year after the surgery of the final randomised participant. There will be no formal criteria for trial termination; however, if it is observed that the intervention is associated with significant harm to subsequently enrolled participants, then consideration of trial termination will be conducted. Methods for additional analyses (e.g. subgroup analyses) {20b} Subgroup analysis will be explored to identify possible treatment effect modifying baseline factors such as age, sex, return to sport expectations and pre-operative ipsilateral knee laxity as measured on the GNRB arthrometer. Methods in analysis to handle protocol non-adherence and any statistical methods to handle missing data {20c} All principal analyses will be based on the intention-to-treat principle, analysing participants in the groups to which they are randomised. Missing data will be quantified and if possible multiple imputation will be used; otherwise, simple imputation will be used. Owing to the nature of enrolment and randomisation in this study, it is unlikely that a participant will be randomised and subsequently not receive the intervention, and thus adherence does not apply directly to the intervention arm. Plans to give access to the full protocol, participant-level data and statistical code {31c} The protocol will be registered at the Australia New Zealand Clinical Trials Registry (ANZCTR), where specifics not present in the current publication may be reviewed.
Side-to-side difference in anterior tibial translation (the primary outcome), overall ACL-RSI, EQ5D-5L, Marx activity scales, IKDC and KOOS-QOL will be calculated as means, with measures of dispersion reported as standard deviation. Differences between groups will be reported as mean differences, with dispersion reported with confidence intervals. Data normality will be assessed with the Shapiro-Wilk test for continuous parametric assessment. Between-group comparisons, such as the primary outcome of mean difference in GNRB anteroposterior laxity at 2 years, will be assessed with independent samples t -tests. If data is found to be non-parametric in distribution, the Wilcoxon rank sum test will be utilised. Comparison of two dichotomous variables, such as return to sport rates and complication rates, will be performed with Fischer’s exact test. Repeated measures such as GNRB measures and mean PROMs will be assessed using mixed effects linear analysis. All tests are two-sided. A p value of <0.05 will be considered the cut-off for statistical significance.
Interim analysis will be performed 1 year after the surgery of the final randomised participant. There will be no formal criteria for trial termination; however, if it is observed that the intervention is associated with significant harm to subsequently enrolled participants, then consideration of trial termination will be conducted.
Subgroup analysis will be explored to identify possible treatment effect modifying baseline factors such as age, sex, return to sport expectations and pre-operative ipsilateral knee laxity as measured on the GNRB arthrometer.
All principal analyses will be based on the intention-to-treat principle, analysing participants in the groups to which they are randomised. Missing data will be quantified and if possible multiple imputation will be used; otherwise, simple imputation will be used. Owing to the nature of enrolment and randomisation in this study, it is unlikely that a participant will be randomised and subsequently not receive the intervention, and thus adherence does not apply directly to the intervention arm.
The protocol will be registered at the Australia New Zealand Clinical Trials Registry (ANZCTR), where specifics not present in the current publication may be reviewed.
Composition of the coordinating centre and trial steering committee {5d} The Principal Investigator, the research assistant/co-ordinator and at least one other Investigator (Internal Trial Monitoring Committee) will meet at least monthly to monitor the progress of the trial to discuss study progress and procedures, adverse events and any other issues, Composition of the data monitoring committee, its role and reporting structure {21a} The Principal Investigator will act as the data manager for the trial. The research team will meet as above in lieu of an official Data Monitoring Committee. Direct access to the data will be granted to authorised representatives from the sponsor, host institution or ethics board for monitoring and/or audit to ensure compliance with regulations. Adverse event reporting and harms {22} Participant safety will be ensured via standard institutional protocols. The use of the GNRB is established to be safe in pre-operative and early post-operative patients , with no adverse events reported, and this will be expressed verbally to patients prior to participation. Adverse events are defined as medical occurrences that may or may not have a causal relationship with the surgical treatment administered. Any adverse events identified will be recorded and reviewed regularly by the Internal Trial Monitoring Committee. The Principal Investigator will determine causality. The number and type of adverse events (AEs) will be recorded up to 24 months. Any AEs will be noted on the REDCap database. These events will also be reviewed by the participants’ medical and surgical teams (who are not a part of the investigative team) as part of standard care. Serious adverse events (SAE) are defined as any untoward and unexpected medical occurrence that results in death, is life-threatening, requires hospitalisation or prolongation of existing inpatients’ hospitalisation, results in persistent or significant disability or incapacity, is a congenital anomaly or birth defect, or is any other important medical condition which, although not included in the above, may require medical or surgical intervention to prevent one of the outcomes listed. All SAEs will be reported to the approving ethics board within 72 h of the investigators becoming aware of them. SAEs that may be expected as part of the surgical interventions and that do not need to be reported to the HREC are complications of anaesthesia or surgery (wound infection, bleeding or damage to adjacent structures such as nerves, tendons and blood vessels, delayed wound healing, and thromboembolic events, femoral or tibial fracture). Frequency and plans for auditing trial conduct {23} Auditing of trial conduct will be performed at a frequency and depth as determined by the local ethics board, independent of investigators and research team. Plans for communicating important protocol amendments to relevant parties (e.g. trial participants, ethical committees) {25} All modifications to study protocols, following approval with the local ethics board, will be documented as amendments in the ANZ clinical trial registry. All participants having previously signed consent forms prior to change in protocol will be notified by email including the detail of the change and its impact on them and offered an opportunity to review a new written consent. If the change will not impact the participants’ involvement in the study, a change in the statistical methods, or a reduction in the frequency of recording of SAEs, or a change in study personnel, as an example, would not impact the participants’ involvement in the trial and thus they will not be notified of these changes. Dissemination plans {31a} By signing the consent form, the participants give their permission to allow the de-identified data generated by this research to be shared/discussed with the local institutional orthopaedic unit and those working within it. Information may be used in publication in peer-reviewed medical journals. The results may also be presented at relevant national or international meetings and conferences. The information from this study will be disseminated within presentations or publications related to the present study, or within research related to this study.
The Principal Investigator, the research assistant/co-ordinator and at least one other Investigator (Internal Trial Monitoring Committee) will meet at least monthly to monitor the progress of the trial to discuss study progress and procedures, adverse events and any other issues,
The Principal Investigator will act as the data manager for the trial. The research team will meet as above in lieu of an official Data Monitoring Committee. Direct access to the data will be granted to authorised representatives from the sponsor, host institution or ethics board for monitoring and/or audit to ensure compliance with regulations.
Participant safety will be ensured via standard institutional protocols. The use of the GNRB is established to be safe in pre-operative and early post-operative patients , with no adverse events reported, and this will be expressed verbally to patients prior to participation. Adverse events are defined as medical occurrences that may or may not have a causal relationship with the surgical treatment administered. Any adverse events identified will be recorded and reviewed regularly by the Internal Trial Monitoring Committee. The Principal Investigator will determine causality. The number and type of adverse events (AEs) will be recorded up to 24 months. Any AEs will be noted on the REDCap database. These events will also be reviewed by the participants’ medical and surgical teams (who are not a part of the investigative team) as part of standard care. Serious adverse events (SAE) are defined as any untoward and unexpected medical occurrence that results in death, is life-threatening, requires hospitalisation or prolongation of existing inpatients’ hospitalisation, results in persistent or significant disability or incapacity, is a congenital anomaly or birth defect, or is any other important medical condition which, although not included in the above, may require medical or surgical intervention to prevent one of the outcomes listed. All SAEs will be reported to the approving ethics board within 72 h of the investigators becoming aware of them. SAEs that may be expected as part of the surgical interventions and that do not need to be reported to the HREC are complications of anaesthesia or surgery (wound infection, bleeding or damage to adjacent structures such as nerves, tendons and blood vessels, delayed wound healing, and thromboembolic events, femoral or tibial fracture).
Auditing of trial conduct will be performed at a frequency and depth as determined by the local ethics board, independent of investigators and research team.
All modifications to study protocols, following approval with the local ethics board, will be documented as amendments in the ANZ clinical trial registry. All participants having previously signed consent forms prior to change in protocol will be notified by email including the detail of the change and its impact on them and offered an opportunity to review a new written consent. If the change will not impact the participants’ involvement in the study, a change in the statistical methods, or a reduction in the frequency of recording of SAEs, or a change in study personnel, as an example, would not impact the participants’ involvement in the trial and thus they will not be notified of these changes.
By signing the consent form, the participants give their permission to allow the de-identified data generated by this research to be shared/discussed with the local institutional orthopaedic unit and those working within it. Information may be used in publication in peer-reviewed medical journals. The results may also be presented at relevant national or international meetings and conferences. The information from this study will be disseminated within presentations or publications related to the present study, or within research related to this study.
This manuscript reports on the methodological design of the STACLR trial (Suture Tape Augmentation of Anterior Cruciate Ligament Reconstruction), the first randomised design prospective trial comparing suture tape-augmented ACL reconstruction to standard ACL reconstruction in adult patients. In this study, 2-year objective knee laxity, subjective patient-reported outcomes, complications and return to sport rates will be compared in patient treated with tape-augmented ACL reconstruction to those without. The hypothesis of the present study is that suture tape augmentation of ACL reconstruction will result in reduced side-to-side anterior tibial laxity at 2 years post-operative. Secondary hypotheses include an improvement in PROMs and equivalent complication profiles in tape-augmented grafts. There are several biomechanical studies comparing tape-augmented reconstruction to standard technique, whilst several animal studies and a scattering of retrospective nonrandomised clinical studies exist, all of which provide an established concept and lay the platform for future research. The most devastating complication of ACL reconstruction may be graft failure; however, owing to the sample size demanded, the present study will not feasibly investigate this, and it is important to note this when interpreting results. In a retrospective cohort study of tape augmentation in ACL reconstruction, Parkes et al. in a post hoc analysis reported a sample size of 1290 to adequately power for graft failure. Nevertheless, on principle, the action of tape augmentation will be to increase the biomechanical strength of the graft, a property that has been established by author groups including Noonan et al. and Lai et al. , where tape-augmented grafts have higher load to failure and reduced elongation under cyclic loading. It follows that this may result in reduced laxity in post-operative knees compared with the standard ACL technique. Therefore, the role of tape augmentation in increasing biomechanical strength is to protect the graft during maturation and protect those grafts at higher risk of failure. Observed reduced knee laxity post-operatively is a well-utilised measure in evaluating the return of anteroposterior stability after ACL reconstruction [ – ]. Notably, early laxity has been associated with poor outcomes including increased risk of graft failure, reduced length of athletic career, permanently increased knee laxity and lower return to function scores . Therefore, the use of an arthrometric measure, namely the GNRB anteroposterior laxity (selected on the basis of its purported superiority over the KT-1000), is an efficient method of assessing the efficacy of tape-augmentation in achieving its biomechanical goal in an achievable sample size. Given the previous pitfalls of synthetic graft augmentation , skeptics of the technique may cite stress shielding or synovial complications such as sterile effusion as drawbacks to the approach. Animal studies, however, have evidence that despite the biomechanical protection of cyclic elongation, histological evidence of graft remodeling occurs , signifying that with the appropriate tensioning of the tape in full extension, stress shielding may be avoided. Similarly, Smith et al. 29 have reported equivalent rates of synovitis and ongoing effusion in tape-augmented and non-augmented canine knees at 6 months. These early findings purporting reduced complication profiles observed in other synthetic graft models establish the need for further, prospective clinical research. This study aims to address some of the above outstanding questions remaining regarding suture tape augmentation of ACL reconstruction.
The current protocol is Version 1.3, dated 30 August 2021. Recruitment commenced on 03 March 2022. Recruitment is anticipated to finish on 30 June 2024.
Additional file 1.
|
Assessment of the underlying causes of adult deaths using a short version of verbal autopsy in Xaiyabouli Province, Lao People’s Democratic Republic | 2dd1c5fe-4802-4587-8848-1bcfec690ebd | 10037893 | Forensic Medicine[mh] | Information of deaths, such as the number of deaths, sex and age of the deceased persons, and causes of death, are essential information for policymakers and government officers to develop health policies and to implement appropriate interventions for improving public health in each country [ – ]. In most developed countries, the underlying cause of death (UCOD) is certified by medical doctors and reported to the government using the International Statistical Classification of Diseases and Related Health Problems (ICD), which is the standard medical classification list of the World Health Organization (WHO) . However, many developing countries have no or limited mortality statistics because the systems of vital registration and death certificate by doctors are weak and many people die at home with no access to healthcare services. When a death occurs outside health facilities, the identification of UCOD is more difficult compared to a death at a health facility . A verbal autopsy (VA) is a useful tool to determine the possible cause of death by interviewing family members or care givers regarding signs, symptoms, available medical history, treatment, and circumstances before a death [ , , ]. The WHO started encouraging the use of VA in developing countries and developed reporting forms in the 1970s . The first WHO VA standard tools were developed in 2007, including a questionnaire for the three age groups, cause of death certification, and coding resources consistent with ICD-10, and they were revised in 2014, 2016, and 2022 [ , , , , ]. There are four methods for VA, namely physician review without algorithmic diagnosis criteria, physician review using an algorithms, computer algorithms, and probabilistic approaches , but there is no method with universal advantages because the disease frequency, available personnel, and/or the social and cultural system are different in each country or region [ – ]. However, previous studies suggested that a VA method can provide a rough estimation of UCOD distribution close to that obtained at health facilities if the categories are not so detailed . Lao People’s Democratic Republic (Lao PDR) is a lower-middle income country in Southeast Asia with a population of 7,379,358 in 2021 . It is suggested that deaths due to noncommunicable disease (NCD), such as ischemic heart disease, chronic kidney diseases, diabetes, and chronic respiratory diseases, are increasing in Lao PDR [ – ]. However, previous studies on causes of death in Lao PDR analyzed only deaths at hospitals [ – ]. There have been no government reports on UCOD because the civil registration and death notification systems are weak and lack medical certification. When someone dies at a health facility, healthcare workers issue a death notification that includes the cause of death and a representative of his/her household submits it to the village office and the district home affairs office . When someone dies outside health facilities, his/her family member receives a death notification without the cause of death from the village office and they submit it to the district home affairs office . The information of all deaths is sent from the district home affairs office to the Ministry of Home Affairs. There is a reporting system of the number of hospital deaths from public health facilities to the Ministry of Health, but it includes only maternal deaths and under-five-year-old child deaths. In Lao PDR, approximately 93% of all deaths occur outside of health facilities and 63.9% of all deaths were not registered by civil registration in 2018 . There is a surveillance system of the causes of death but it is also only for maternal deaths and under-five-year-old child deaths at health facilities. Therefore, collecting information of the cause of death in the adult population, especially deaths outside health facilities, is a challenge for the Lao government. The WHO VA instrument for the adult population includes many questions to identify the cause of death . To investigate the causes of death in Lao PDR, especially deaths outside health facilities, a shorter version of the VA instrument is necessary to reduce the investigation time and the budget. Therefore, this study aimed to develop a short version of the VA instrument and identify the causes of death in the adult population in Xaiyabouli Province, Lao PDR.
Development questions to identify UCOD In this study, a short version of the VA instrument was developed to identify the cause of death of a person who was 15 years or older that occurred outside of health facilities but not maternal death during pregnancy and delivery or within 28 days after childbirth. First, questions for the short VA were listed in the order of clearness in features indicating a category of UCOD; accident/injury, sudden death, stroke (apoplexy), tumor, diarrhea, respiratory disease, tetanus, meningitis, liver failure, heart/renal disease, and senility (natural death). Accident/injury was categorized into suicide/homicide (S/H), bite/sting or food poisoning (B/S), traffic accident (TA), and other injury (OI) (Table ). Suicide and homicide are recorded in the village office in Lao PDR. Sudden death meant a death within 24 h from onset and classified into myocardial infarction (MI) and arachnoid hemorrhage (AH). When the deceased did not have a chest pain or severe headache, UCOD was other sudden death (OS). When the death occurred within a year after the onset of palsy, UCOD was stroke (ST). When tumors were recognized in the breast, neck, head, abdomen, or other parts in the last month before the death, UCOD was tumor (TU) and the part of the tumor was recorded. When severe diarrhea was found in the last week before the death, UCOD was bloody diarrhea disease (BD) or non-bloody diarrhea disease (ND) according to the characteristics of diarrhea. Respiratory disease was recognized by cough, sputum, and dyspnea and categorized into pneumonia (PN), asthma (AS), and other respiratory disease (OR). Tetanus (TE), meningitis (ME), and liver failure (LF) were decided by typical symptoms and signs. When the deceased had dyspnea when he/she worked, walked, or lied down, UCOD was heart/renal failure (H/R) and other heart disease (OH) according to having edema in the face, legs, ankles, or feet. When the deceased had symptoms other than those included in these questions, the symptoms were recorded. When the deceased had no symptoms or no information from medical records, UCOD was senility (SE) when the age was 70 or older and other disease (OD) when the age was 69 years or younger. UCODs for the short VA instrument were decided considering the situation of Lao PDR and the main causes of deaths in the province from 2016 to 2018. The interview for UCOD started from Question 0 (Q0) “Do you have information on the situation or condition of the deceased?” If an interviewee had no information on the situation or condition of the deceased, the case was excluded from the subjects of the VA. When an interviewee answered yes to Q0, he/she was asked sequentially from Q1 to Q12 (Additional file ) until UCOD was identified. Q2 was “Did the deceased visit a health facility within a year before the death?” When the answer was yes, the interviewer asked “Were you or a family member explained about the diagnosis possibly being related to the cause of death?” When a death occurred within a year after the last visit to a health facility and the medical record was available, the information from the record was taken into account for determination of UCOD. Except for the homicide/suicide and injury, the categories of UCOD with the frequency intuitively more than 1% were included in the short VA form. A short version of the VA instrument The short VA instrument in this study included four parts: Part 1 was for collecting socio-demographic data of the deceased, namely name, sex, birthday, village name, ethnic group, the educational level, occupation, number of family members, and the date of the death. Part 2 was for collecting the data (name and birthday) of the interviewee, who was a family member of the deceased. Part 3 was a structured questionnaire for assigning the UCOD. Part 4 included the name of the interviewer, the date of interview, and time of interview. The time was measured from Part 1 to the last question of Part 4. The short VA was made to be used by trained healthcare workers. The standard interview time was set to be less than 20 min. The priority was high comparability among different areas in Lao PDR and it was assumed that interviewees would understand the standard Lao language. The draft of the VA short version in Lao language was tested on 19 deaths at Xaiyabouli Provincial Hospital and the final version was made after revision. Preliminary verification of the short VA form using hospital deaths To estimate rough validity of this short VA form, UCOD of people who died at the Xaiyabouli Provincial Hospital between January and September in 2020 were investigated by an interview with a family member using the short VA form and compared to the UCOD in the medical records of the patients. There were 179 deaths at the provincial hospital during the period and 100 deceased patients whose houses were in 19 villages around the hospital were selected. Three nurses of the provincial health department were trained for short VA and all clinical information of the deceased patients were masked to the nurses. The nurse visited the home of the deceased patient 15–30 days after the death and interviewed a family member of the patient who was 18 years old or older. Q2 was excluded from the short VA form to avoid the statement by interviewees on possible UCOD. The completed short VA forms were submitted to the principal researcher to make the final decision to decide the category of the UCOD. Two doctors, other than the principal researcher, reviewed the medical records and decided the UCOD of each patient. The doctors had training about identifying UCOD and UCOD decided by short VA was masked to them. UCODs of 97 deaths (18 females and 79 males) by short VA and in the medical records were compared, because family members of three patients were not at home when the survey team visited. Written informed consent was obtained from a family member of each patient. The data of deaths outside medical facilities in Xaiyabouli Province in 2020 The short VA was applied to the deceased people who were 15 years or older and died outside health facilities in Xaiyabouli Province from January to December 2020. The interviewers were 37 nurses and 50 assistant doctors of all district health offices ( n = 11) and health centers ( n = 76) in the province. All 87 interviewers attended a training session using the interviewer’s manual. When a death occurs outside health facilities, the death is reported from the head of the village or village health volunteers to the corresponding district health office or health center. The interviewers closely worked with the head of the village or village health volunteers in each village of Xaiyabouli Province. The interviewers visited the household of the deceased 15–30 days after the death. A face-to-face interview was conducted with a family member who was 18 years or older and lived with the deceased in the same household until the day of the death. Informed consent was obtained from the family member before the interview. The time (minutes) of the interview was recorded. In 2020, 1,031 deaths of people aged 15 years or older were reported in the province and 1,012 deaths were investigated using the short VA form because families of 19 deceased people declined to be interviewed. All completed short VA forms were submitted to supervisors to be examined to check that there was no missing information and then submitted to the principal researcher to make the final decision to assign the UCOD. The data of deaths at health facilities in Xaiyabouli Province in 2020 In 2020, 223 patients aged 15 years or older died at health facilities in Xaiyabouli Province. The information of the dead patients (hospital deaths) were taken from their medical records including sex, age, ethnic group, occupation, name of the health facility, and cause of death. Comparison of categories of UCODs UCODs were categorized into nine groups, such as accident/injury (suicide/homicide, bite/sting/food poisoning, traffic accident, drowning, and other injury), heart/renal disease (myocardial infarction, heart/renal failure with edema, heart failure, renal failure, other heart disease), respiratory disease (pneumonia, asthma, and other respiratory disease), brain disease (arachnoid hemorrhage and stroke), liver/gastro-intestine disease (liver cirrhosis, liver failure, hepatitis, bloody or non-bloody diarrhea, and gastro-intestine bleeding), infection (tuberculosis, tetanus, meningitis, sepsis, and rickettsia), tumor, senility, and other (unknown and other disease). All 1,012 deaths outside of health facilities were home deaths and all 223 deaths at health facilities were hospital deaths. The percentages of UCOD categories were compared between home deaths ( n = 1,012) and hospital deaths ( n = 223), between the age of 15–59 years old ( n = 530) and age ≥ 60 years old ( n = 705), and between males ( n = 695) and females ( n = 540). Statistical analysis Data were analyzed using SPSS version 21.0 (IBM SPSS Inc, Amonk, NY, USA). In the verification study, the sensitivity, the specificity, and the positive predictive value of the short VA were calculated based on the hospital diagnosis as a gold standard. Cohen’s kappa was used to estimate the agreement between the short VA and hospital diagnosis and κ > 0.75 was considered excellent agreement . Fisher’s exact test was used for comparing UCOD categories between two groups. A P -value < 0.05 was considered statistically significant. Ethical issues This study was approved by the Research Ethical Review Committee of the National Institute of Public Health, Lao PDR (approval number: 049 /NECHR). All methods were performed in accordance with the relevant guidelines and regulations.
In this study, a short version of the VA instrument was developed to identify the cause of death of a person who was 15 years or older that occurred outside of health facilities but not maternal death during pregnancy and delivery or within 28 days after childbirth. First, questions for the short VA were listed in the order of clearness in features indicating a category of UCOD; accident/injury, sudden death, stroke (apoplexy), tumor, diarrhea, respiratory disease, tetanus, meningitis, liver failure, heart/renal disease, and senility (natural death). Accident/injury was categorized into suicide/homicide (S/H), bite/sting or food poisoning (B/S), traffic accident (TA), and other injury (OI) (Table ). Suicide and homicide are recorded in the village office in Lao PDR. Sudden death meant a death within 24 h from onset and classified into myocardial infarction (MI) and arachnoid hemorrhage (AH). When the deceased did not have a chest pain or severe headache, UCOD was other sudden death (OS). When the death occurred within a year after the onset of palsy, UCOD was stroke (ST). When tumors were recognized in the breast, neck, head, abdomen, or other parts in the last month before the death, UCOD was tumor (TU) and the part of the tumor was recorded. When severe diarrhea was found in the last week before the death, UCOD was bloody diarrhea disease (BD) or non-bloody diarrhea disease (ND) according to the characteristics of diarrhea. Respiratory disease was recognized by cough, sputum, and dyspnea and categorized into pneumonia (PN), asthma (AS), and other respiratory disease (OR). Tetanus (TE), meningitis (ME), and liver failure (LF) were decided by typical symptoms and signs. When the deceased had dyspnea when he/she worked, walked, or lied down, UCOD was heart/renal failure (H/R) and other heart disease (OH) according to having edema in the face, legs, ankles, or feet. When the deceased had symptoms other than those included in these questions, the symptoms were recorded. When the deceased had no symptoms or no information from medical records, UCOD was senility (SE) when the age was 70 or older and other disease (OD) when the age was 69 years or younger. UCODs for the short VA instrument were decided considering the situation of Lao PDR and the main causes of deaths in the province from 2016 to 2018. The interview for UCOD started from Question 0 (Q0) “Do you have information on the situation or condition of the deceased?” If an interviewee had no information on the situation or condition of the deceased, the case was excluded from the subjects of the VA. When an interviewee answered yes to Q0, he/she was asked sequentially from Q1 to Q12 (Additional file ) until UCOD was identified. Q2 was “Did the deceased visit a health facility within a year before the death?” When the answer was yes, the interviewer asked “Were you or a family member explained about the diagnosis possibly being related to the cause of death?” When a death occurred within a year after the last visit to a health facility and the medical record was available, the information from the record was taken into account for determination of UCOD. Except for the homicide/suicide and injury, the categories of UCOD with the frequency intuitively more than 1% were included in the short VA form.
The short VA instrument in this study included four parts: Part 1 was for collecting socio-demographic data of the deceased, namely name, sex, birthday, village name, ethnic group, the educational level, occupation, number of family members, and the date of the death. Part 2 was for collecting the data (name and birthday) of the interviewee, who was a family member of the deceased. Part 3 was a structured questionnaire for assigning the UCOD. Part 4 included the name of the interviewer, the date of interview, and time of interview. The time was measured from Part 1 to the last question of Part 4. The short VA was made to be used by trained healthcare workers. The standard interview time was set to be less than 20 min. The priority was high comparability among different areas in Lao PDR and it was assumed that interviewees would understand the standard Lao language. The draft of the VA short version in Lao language was tested on 19 deaths at Xaiyabouli Provincial Hospital and the final version was made after revision.
To estimate rough validity of this short VA form, UCOD of people who died at the Xaiyabouli Provincial Hospital between January and September in 2020 were investigated by an interview with a family member using the short VA form and compared to the UCOD in the medical records of the patients. There were 179 deaths at the provincial hospital during the period and 100 deceased patients whose houses were in 19 villages around the hospital were selected. Three nurses of the provincial health department were trained for short VA and all clinical information of the deceased patients were masked to the nurses. The nurse visited the home of the deceased patient 15–30 days after the death and interviewed a family member of the patient who was 18 years old or older. Q2 was excluded from the short VA form to avoid the statement by interviewees on possible UCOD. The completed short VA forms were submitted to the principal researcher to make the final decision to decide the category of the UCOD. Two doctors, other than the principal researcher, reviewed the medical records and decided the UCOD of each patient. The doctors had training about identifying UCOD and UCOD decided by short VA was masked to them. UCODs of 97 deaths (18 females and 79 males) by short VA and in the medical records were compared, because family members of three patients were not at home when the survey team visited. Written informed consent was obtained from a family member of each patient.
The short VA was applied to the deceased people who were 15 years or older and died outside health facilities in Xaiyabouli Province from January to December 2020. The interviewers were 37 nurses and 50 assistant doctors of all district health offices ( n = 11) and health centers ( n = 76) in the province. All 87 interviewers attended a training session using the interviewer’s manual. When a death occurs outside health facilities, the death is reported from the head of the village or village health volunteers to the corresponding district health office or health center. The interviewers closely worked with the head of the village or village health volunteers in each village of Xaiyabouli Province. The interviewers visited the household of the deceased 15–30 days after the death. A face-to-face interview was conducted with a family member who was 18 years or older and lived with the deceased in the same household until the day of the death. Informed consent was obtained from the family member before the interview. The time (minutes) of the interview was recorded. In 2020, 1,031 deaths of people aged 15 years or older were reported in the province and 1,012 deaths were investigated using the short VA form because families of 19 deceased people declined to be interviewed. All completed short VA forms were submitted to supervisors to be examined to check that there was no missing information and then submitted to the principal researcher to make the final decision to assign the UCOD.
In 2020, 223 patients aged 15 years or older died at health facilities in Xaiyabouli Province. The information of the dead patients (hospital deaths) were taken from their medical records including sex, age, ethnic group, occupation, name of the health facility, and cause of death.
UCODs were categorized into nine groups, such as accident/injury (suicide/homicide, bite/sting/food poisoning, traffic accident, drowning, and other injury), heart/renal disease (myocardial infarction, heart/renal failure with edema, heart failure, renal failure, other heart disease), respiratory disease (pneumonia, asthma, and other respiratory disease), brain disease (arachnoid hemorrhage and stroke), liver/gastro-intestine disease (liver cirrhosis, liver failure, hepatitis, bloody or non-bloody diarrhea, and gastro-intestine bleeding), infection (tuberculosis, tetanus, meningitis, sepsis, and rickettsia), tumor, senility, and other (unknown and other disease). All 1,012 deaths outside of health facilities were home deaths and all 223 deaths at health facilities were hospital deaths. The percentages of UCOD categories were compared between home deaths ( n = 1,012) and hospital deaths ( n = 223), between the age of 15–59 years old ( n = 530) and age ≥ 60 years old ( n = 705), and between males ( n = 695) and females ( n = 540).
Data were analyzed using SPSS version 21.0 (IBM SPSS Inc, Amonk, NY, USA). In the verification study, the sensitivity, the specificity, and the positive predictive value of the short VA were calculated based on the hospital diagnosis as a gold standard. Cohen’s kappa was used to estimate the agreement between the short VA and hospital diagnosis and κ > 0.75 was considered excellent agreement . Fisher’s exact test was used for comparing UCOD categories between two groups. A P -value < 0.05 was considered statistically significant.
This study was approved by the Research Ethical Review Committee of the National Institute of Public Health, Lao PDR (approval number: 049 /NECHR). All methods were performed in accordance with the relevant guidelines and regulations.
Development and verification of the short VA form In this study, the short VA form was developed including 22 UCODs considering the situation of Lao PDR. The UCOD was coded by ICD-10 and corresponding category of WHO VA (Table ) . The short VA form was validated by comparing UCOD of 97 patients who died at Xaiyabouli Provincial Hospital that were recorded in the medical records and those decided by the short VA. The sensitivity of the short VA was 85.7% (12/14) for myocardial infarction, 80.0% (4/5) for asthma, and 84.6% (11/13) for heart/renal failure with face edema (Table ). The sensitivity was 90.7% (88/97, 95% confidence interval 83.1–95.7%). The specificity and the positive predictive value of the short VA were 98.8% and 92.3% for myocardial infarction, 97.6% and 84.6% for heart/renal failure with face edema, 98.9% and 88.9% for pneumonia, 98.9% and 87.5% for stroke, respectively. Kappa statistics showed that κ was 0.896, which means a perfect agreement between UCOD by the short VA and hospital diagnoses. Characteristics of deaths in Xaiyabouli Province in 2020 In 2020, a total of 1,254 deaths in Xaiyabouli Province were reported and the information of 1,235 deaths were collected including 1,012 deaths outside health facilities and 223 deaths at health facilities. All 1,012 deaths outside health facilities were at home (81.9%) and all 223 deaths at hospitals (18.1%) but not health centers (Table ). The information of 1,012 home deaths was collected using the short VA form (Fig. ). Of all the 1,235 deaths in 2020, most deaths were males ( n = 695, 56.3%), 60–74 years old ( n = 352, 28.5%) and 75 years or older ( n = 353, 28.6%), Lao-Tai ethnic group ( n = 1,021, 82.7%), and farmers ( n = 957, 77.5%) (Table ). Compared to hospital deaths, home deaths had more women (45.1% vs. 37.7%, P = 0.044), the age group ≥ 75 years old (31.8% vs. 13.9%), and Lao-Tai ethnic group (84.8% vs 73.0%). In terms of occupation, farmers were more likely to die at hospitals (87.0%) compared to at home (75.4%). Most interviewees of short VA were daughters ( n = 409, 40.4%), wives ( n = 275, 27.2%), and husbands ( n = 195, 19.3%). The average interview time was 12.4 min (range, 3–19 min) and most interviews were 10–19 min ( n = 770, 76.1%). UCOD of all deaths in Xaiyabouli Province UCOD of the 1,235 deaths were identified using the short VA or from medical records at hospitals (Fig. ). The most common cause was senility ( n = 164, 13.3%), followed by heart/renal failure ( n = 130, 10.5%), pneumonia ( n = 119, 9.6%), traffic accident ( n = 88, 7.1%), myocardial infarction ( n = 79, 6.4%), and stroke ( n = 78, 6.3%) (Table ). Heart failure or renal failure could not be separately identified as UCOD of home deaths ( n = 73, 7.2%), although heart failure ( n = 26) and renal failure ( n = 31) were diagnosed for hospital deaths. Of all deaths, 50 deaths (4.0%) were due to suicide/homicide and 52 deaths (4.2%) due to tumors, and causes of 183 deaths (14.8%) were categorized into others because UCOD was not identified. The major UCOD category was heart/renal disease ( n = 225, 18.2%) followed by accident/injury ( n = 180, 14.6%), respiratory disease ( n = 165, 13.4%), senile ( n = 164, 13.3%), and brain disease ( n = 129, 10.4%). Comparison of UCOD between home deaths and hospital deaths, between age groups, and between males and females When the UCOD categories were compared between home deaths and hospital deaths, home deaths had more accident/injury (16.0% vs. 8.1%, P = 0.002) and tumor (4.7% vs. 1.8%, P < 0.001) (Table ). All deaths due to senility were home deaths. Deaths caused by heart/renal disease (15.1% vs. 32.3%, P < 0.001), respiratory disease (12.2% vs. 18.8%, P = 0.012), liver/gastro-intestine disease (5.3% vs. 9.0%, P = 0.043), and infection (3.1% vs. 14.3%, P < 0.001) were less likely to have occurred at home compared to at hospitals. All deaths were divided into the age group of 15–59 years old and the age group of 60 years old or older. The age group of 15–59 years had more deaths of the categories of accident/injury (28.1% vs. 4.4%, P < 0.001), liver/gastro-intestine disease (8.1% vs. 4.4%, P = 0.008), infection (7.2% vs. 3.5%, P = 0.006), and tumor (6.0% vs. 2.8%, P < 0.001) (Table ). Comparison of UCOD between males and females showed that males had significantly fewer natural deaths (11.2% vs. 15.9%, P = 0.018) and more deaths due to tumor (5.2% vs. 3.0%, P = 0.018) than females.
In this study, the short VA form was developed including 22 UCODs considering the situation of Lao PDR. The UCOD was coded by ICD-10 and corresponding category of WHO VA (Table ) . The short VA form was validated by comparing UCOD of 97 patients who died at Xaiyabouli Provincial Hospital that were recorded in the medical records and those decided by the short VA. The sensitivity of the short VA was 85.7% (12/14) for myocardial infarction, 80.0% (4/5) for asthma, and 84.6% (11/13) for heart/renal failure with face edema (Table ). The sensitivity was 90.7% (88/97, 95% confidence interval 83.1–95.7%). The specificity and the positive predictive value of the short VA were 98.8% and 92.3% for myocardial infarction, 97.6% and 84.6% for heart/renal failure with face edema, 98.9% and 88.9% for pneumonia, 98.9% and 87.5% for stroke, respectively. Kappa statistics showed that κ was 0.896, which means a perfect agreement between UCOD by the short VA and hospital diagnoses.
In 2020, a total of 1,254 deaths in Xaiyabouli Province were reported and the information of 1,235 deaths were collected including 1,012 deaths outside health facilities and 223 deaths at health facilities. All 1,012 deaths outside health facilities were at home (81.9%) and all 223 deaths at hospitals (18.1%) but not health centers (Table ). The information of 1,012 home deaths was collected using the short VA form (Fig. ). Of all the 1,235 deaths in 2020, most deaths were males ( n = 695, 56.3%), 60–74 years old ( n = 352, 28.5%) and 75 years or older ( n = 353, 28.6%), Lao-Tai ethnic group ( n = 1,021, 82.7%), and farmers ( n = 957, 77.5%) (Table ). Compared to hospital deaths, home deaths had more women (45.1% vs. 37.7%, P = 0.044), the age group ≥ 75 years old (31.8% vs. 13.9%), and Lao-Tai ethnic group (84.8% vs 73.0%). In terms of occupation, farmers were more likely to die at hospitals (87.0%) compared to at home (75.4%). Most interviewees of short VA were daughters ( n = 409, 40.4%), wives ( n = 275, 27.2%), and husbands ( n = 195, 19.3%). The average interview time was 12.4 min (range, 3–19 min) and most interviews were 10–19 min ( n = 770, 76.1%).
UCOD of the 1,235 deaths were identified using the short VA or from medical records at hospitals (Fig. ). The most common cause was senility ( n = 164, 13.3%), followed by heart/renal failure ( n = 130, 10.5%), pneumonia ( n = 119, 9.6%), traffic accident ( n = 88, 7.1%), myocardial infarction ( n = 79, 6.4%), and stroke ( n = 78, 6.3%) (Table ). Heart failure or renal failure could not be separately identified as UCOD of home deaths ( n = 73, 7.2%), although heart failure ( n = 26) and renal failure ( n = 31) were diagnosed for hospital deaths. Of all deaths, 50 deaths (4.0%) were due to suicide/homicide and 52 deaths (4.2%) due to tumors, and causes of 183 deaths (14.8%) were categorized into others because UCOD was not identified. The major UCOD category was heart/renal disease ( n = 225, 18.2%) followed by accident/injury ( n = 180, 14.6%), respiratory disease ( n = 165, 13.4%), senile ( n = 164, 13.3%), and brain disease ( n = 129, 10.4%).
When the UCOD categories were compared between home deaths and hospital deaths, home deaths had more accident/injury (16.0% vs. 8.1%, P = 0.002) and tumor (4.7% vs. 1.8%, P < 0.001) (Table ). All deaths due to senility were home deaths. Deaths caused by heart/renal disease (15.1% vs. 32.3%, P < 0.001), respiratory disease (12.2% vs. 18.8%, P = 0.012), liver/gastro-intestine disease (5.3% vs. 9.0%, P = 0.043), and infection (3.1% vs. 14.3%, P < 0.001) were less likely to have occurred at home compared to at hospitals. All deaths were divided into the age group of 15–59 years old and the age group of 60 years old or older. The age group of 15–59 years had more deaths of the categories of accident/injury (28.1% vs. 4.4%, P < 0.001), liver/gastro-intestine disease (8.1% vs. 4.4%, P = 0.008), infection (7.2% vs. 3.5%, P = 0.006), and tumor (6.0% vs. 2.8%, P < 0.001) (Table ). Comparison of UCOD between males and females showed that males had significantly fewer natural deaths (11.2% vs. 15.9%, P = 0.018) and more deaths due to tumor (5.2% vs. 3.0%, P = 0.018) than females.
In this study, the short VA was developed to identify UCODs of deaths outside health facilities in Xaiyabouli Province. The preliminary validation of the short VA form by hospital diagnosis showed high sensitivity and specificity. However, UCOD of 17.2% of home deaths could not be identified by the short VA and heart failure and renal failure could not be distinguished. Symptoms of chronic heart failure and end-stage renal failure are similar and the two diseases are major progressive factors of each other . To identify more kinds of UCOD, being diagnosed and having laboratory examinations at health facilities before deaths is needed. The WHO VA instrument includes questions to ask about the medical history, results of laboratory exams, and diagnosis by healthcare professional . However, access to healthcare service is poor in rural areas and provinces and basic laboratory tests for high-income countries, such as biochemical tests and pathological tests, are not regularly performed or unavailable even at provincial hospitals [ , – ]. Therefore, the short VA form was useful to understand the causes of most home deaths by interviewing family members for a short time but the form should be improved based on a more accurate larger-size validation study. To the best of our knowledge, this is the first study to show the UCODs in the adult population in Lao PDR, including deaths outside hospitals. In this study, heart/renal disease was the major UCOD category among all deaths as well as hospital deaths and one of the main categories among home deaths. Deaths due to arachnoid hemorrhage and stroke accounted for 10.4% of all deaths and 69 deaths categorized into other UCOD had a history of hypertension or diabetes mellites. The results of this study suggest that 46.2% of all deaths were due to NCD (heart/renal disease, respiratory disease, brain disease, and tumor), which was lower than that estimated by the WHO in 2016 . In the WHO estimation, 60% of all deaths were due to NCD, including cardiovascular disease (27%), cancer (12%), chronic respiratory disease (5%), diabetes (4%), and others (12%). In 2014, the Lao government implemented a multisectoral action plan to prevent NCD by reducing risks (tobacco use, harmful alcohol consumption, unhealthy diet, and physical inactivity) and promoting treatments (cardiovascular disease, diabetes mellitus, and cancer) . Not only the implementation of cost-effective interventions but also the monitoring of the morbidity and mortality of NCD are essential to help reduce NCD in the whole country. Characteristics of home deaths were different from hospital deaths. Home deaths had more deceased who were females and 75 years old or older. It may be because all natural deaths were home deaths and the life expectancy of women (70 years old) is longer than that of men (66 years old) . The result that the Lao-Tai ethnic group had more home deaths suggests that beliefs are different among ethnic groups. It is reported that people of Lao ethnic groups (Lao Loum, Lao Theung, and Lao Sung) believe that the spirits of people are not reincarnated but become malevolent ones when the people die by accident or during childbirth . The educational level cannot be compared between home deaths and hospital deaths. When the educational levels of people who died at home were compared with the average educational levels in the general population, home deaths in this study had more people who had no education (31.2% vs. 16.0%) and fewer people who had education at university or higher (2.7% vs. 14.3%) . People with a higher educational level may have treatment at health facilities more than those with a lower educational level. In this study, the percentages of UCOD categories were different between home deaths and hospital deaths. Home deaths had more deaths due to accident/injury, tumor, and senility than hospital deaths. On the other hand, hospital deaths had more deaths due to acute illness, such as heart or renal failure, pneumonia, gastro-intestine bleeding, and sepsis. In Lao PDR, most people focus on acute illness but not on disease prevention or health promotion and people often come to health facilities after diseases are advanced. Lao people prefer to die at home because they believe that the soul of a person may wander and not be reincarnated when he or she dies outside the home . Therefore, when patients are diagnosed as having uncurable diseases, they prefer going to traditional healers rather than visiting health facilities for treatment. When a patient at a hospital is going to die, their family hopes that the patient will return to their home before dying. Senility was the major cause of home deaths and all deaths due to senility were at home. If an elderly person who died at home due to senility visited a health facility, they might be diagnosed with heart disease, cerebral disease, or pneumonia as the cause of death . Comparing the UCODs between males and females, deaths due to senility were significantly higher among females than males. This is because of longer life expectancy among females than males. The average life-year of females was also older (62.6 years) than that of males (59.7 years) in this study. Deaths due to tumor, especially abdominal tumor, were more likely to occur in males compared to females. In a previous study on cancer mortality at hospitals in 2007–2008, the major cancer for death was liver cancer among both males and females, but of all cancer deaths the percentage of liver cancer death was higher among males (52.2%) than females (28.4%) . Another study including hospital deaths in Vientiane Capital in 2013–2015 showed that liver cancer accounted for 1.9% of all deaths among males, which was higher than among females (0.8%) . According to the WHO data on mortality in Lao PDR, the major cancer among males was liver cancer but that among females was breast cancer . Abdominal tumor in this study may suggest ascites due to cancer, especially liver cancer. Liver cancer may be the major cancer in Lao PDR due to the high prevalence of HBV and HCV and poor access to health service for receiving appropriate treatment . Compared to the age group of 60 years or older, the age group of younger than 60 years had higher mortality due to accident/injury, liver/gastro-intestine disease, infection, and tumor, especially traffic accident, suicide/homicide, and liver failure. These results were consistent with those of previous studies in Lao PDR and other developing countries [ , , – ]. In this study, all deaths due to drowning occurred in the younger age group. It may be because Xaiyabouli Province is mountainous and has a hydroelectric dam using the Mekong River. It is reported that most people with liver disease die aged between 18 to 65 years and that mortality due to cancer in the young generation is higher in low- and middle-income countries compared to high-income countries . To reduce preventable deaths in the younger generation, interventions for preventing traffic accidents, promoting mental health service, and establishing a safe environment for people are needed in the province. This study has some limitations. First, UCOD of 17.2% of home deaths could not be identified using the short VA form. Heart failure and renal failure could not be distinguished and causes of diarrhea or liver failure could not be found. Second, this study may not include all deaths that occurred in the province because 63.9% of all deaths in the whole country were not registered using civil registration in 2018 . Third, the results of this study cannot be representative of the data of Lao PDR because this study included only deaths in a province for a year. The data of both hospital deaths and home deaths should be collected in all provinces and analyzed to identify the causes of death of the adult population. To collect UCOD of home deaths, the adult mortality surveillance by healthcare workers using a short VA tool should be routine work for district health offices and health centers .
This study showed that the major UCOD category was heart/renal disease in the adult generation of Xaiyabouli Province in 2020 and that 81.9% of all deaths were home deaths. Compared to hospital deaths, home deaths had more deceased who were women, who were ≥ 75 years old, and who were Lao-Tai ethnic group. Home deaths had more deaths due to accident/injury, tumor, and senility but fewer deaths due to heart/renal disease, respiratory disease, liver/gastro-intestine disease, and infection. The age group of 15–59 years had more deaths of the categories of accident/injury, liver/gastro-intestine disease, infection, and tumor. Males had more deaths due to tumor and fewer natural deaths than females. Cost-effective interventions based on the multisectoral NCD prevention plan should be appropriately implemented. To understand UCOD of all deaths in the country, mortality surveillance using the short VA tool should be conducted for all home deaths, although the short VA needs to be revised based on a large country-wide validation study.
Additional file 1. Short verbal autopsy form for adult deaths outside health facilities in Lao PDR.
|
Medical students’ perceptions towards distance e-Learning in gynecology ward during the COVID-19 pandemic | 7251ce39-76d3-46fe-854d-a287130c0907 | 10038303 | Gynaecology[mh] | The coronavirus disease 2019 (COVID-19) pandemic has caused serious public health problems and compromised the health of individuals and communities . The spread of the COVID-19 virus across the world has affected the educational systems worldwide, including in Iran. The closure of educational institutions have created numerous challenges for medical students and teachers . In response to the COVID-19 situation, the Association of American Medical Colleges made an announcement on March 17, 2020, and issued guidelines for medical schools to pause clinical rotations . The National COVID-19 Administration in Iran also announced an emergency state . Hence, all medical schools were forced to search for creative techniques to foster clinical medical education. Distance e-Learning (DEL) emerged as the method of teaching to maintain the continuity of education during the pandemic . DEL is defined as “using computer technology to deliver training, including technology-supported learning online, offline, or both” . This shift was considered the best solution to clinical education problems in the era of COVID-19. The effectiveness of DEL amidst the COVID-19 crisis has been investigated in several studies . A survey by Al-Balas et al. showed that distance learning was the best solution to maintain clinical learning processes in emergency situations such as the COVID-19 pandemic . In Iran’s medical education field, eLearning was not considered a new phenomenon in teaching medical students. The pandemic forced the use of virtual teaching and learning processes to complete the syllabus within the time frame. The Universities in Iran has implemented several technological tools to secure the remote clinical teaching process. Training sessions took place both for medical students and clinical teachers to learn how to interact via specific platforms, such as Adobe Connect, Skyroom, and Big Blue Button software. Nevertheless, the characterization of this format of teaching is perceived to be more difficult compared to face-to-face teaching for medical students . In this regard, it is crucial to evaluate the perception of medical students on DEL approaches in the COVID-19 pandemic. This research focuses on the experience of DEL in clinical teaching at the Iran University of Medical Sciences from medical students’ viewpoints. This study aims to explore medical students’ perspectives on DEL in the Gynecology Ward during the COVID-19 pandemic. This cross-sectional study was conducted at the Iran University of Medical Sciences from September 2020 to September 2021. The study sample included 130 medical students, in the fifth and sixth year of Medicine who participated in distance training courses in the gynecology ward during the COVID-19 pandemic. All medical students were included for the study. A self-administered questionnaire was used in this study. Questionnaires were developed through literature review and consultation with gynecology and eLearning experts. Face and content validity was measured using qualitative methods. To measure face validity, 5 expert were selected and asked to check the questionnaire items respecting simplicity and clarity. The same experts are asked to examine the content validity of items with respect to relevance and representativeness. In this process, some items were added, deleted, and revised. Internal consistency was assessed with Cronbach’s alpha (0.83). The Thirty-item questionnaire consisted of six factors: (Demographic data, preferred eLearning tools and location for use, theoretical education, practical skills education, and instructor and infrastructure using a five-point Likert scale ("strongly agree," "agree," "no idea," "disagree," and "strongly disagree"). The research objectives were explained to the participants before the study. In addition, written informed consent was obtained. An online questionnaire was provided using Porsline software in the Persian language. Microsoft office excel software used for the data analysis. The result of the study was analyzed by using descriptive statistics such as frequency distribution and percentage. The link to the questionnaire was forwarded to the students by email. The Broadband internet connections were available to a much wider group of medical student who has participated in this study. Internet access for those who lack home broadband internet, were mobile access via their smartphones. Some medical students turn to public internet use (hospital or library) to participate in online class. Broadband internet connections were available to the instructors also (gynecology faculties). They broadcasted from their homes or the hospital. Remote online classes were held synchronously via Adobe Connect, Skyroom, and Big Blue Button software and recordings were accessible afterwards. Online teaching was planned according to the educational calendar. We used some strategies for engaging medical student during online classes such as using online group debate that promotes individual and group reflections in class, watching interactive videos, using a quiz during a synchronous learning activity, setting collaborative group tasks and opening student and teacher webcams. The ethics committee approved this study at the Iran University of Medical Sciences (code: IR.IUMS.REC.1400.208). The questionnaire was sent to 170 medical students. Of the 130 respondents 65% were female and 35% were male. There were 57 (43.8%) externs and 73 (56.2) interns. Preferred devices and locations for participation in Distance E-learning Courses is shown in . shows the frequency distribution of respondents regarding medical students’ perceptions of DEL in the Gynecology ward during the Covid-19 Pandemic. The results of the online survey suggest that medical students have found both positive and negative aspects of clinical learning by DEL format in Gynecology ward. The results displayed in , are consistent with those of other studies and suggest that m-learning is an effective method for digital learning through the current and future crises, same as the COVID-19 pandemic. Therefore, the preparation for mobile learning would be important to manage future global threats . This finding is in agreement with Bacolod’s (2022) findings which showed mobile learning as a crucial learning tool and agreed about its importance during pandemic despite some difficulties in implementation . These findings further support the idea that m-learning is very helpful in improving the study gap during COVID-19 pandemic and education policymakers must incorporate mobile learning technology in educational systems . Additionally, in a study by Ronnie et al. from the Philippines, 93% the students had a smartphone, and 83% had a laptop or computer. The results displayed in , support the previous research that medical students had reported difficulties in clinical learning related to online teaching strategies during the pandemic . A study by Dost et al. in 2020 showed that 75.99% of the medical students reported that online teaching had not positively replaced the clinical teaching that they experienced via bedside teaching . This finding is also different from the result of the study by Hasan et al (2020) that 21.1% of medical students agreed that e-learning could be used effectively for clinical skill education . A result of the study by Daffalla-Awadalla Gismalla (2021) showed that 64% of medical students perceived that online learning is the greatest strategy during COVID-19 lockdown . In this regard, a study by Elzainy et al. in Saudi Arabia showed that 78% of students agreed or strongly agreed that e-learning had compensated for the suspension of in-person education due to the COVID-19 pandemic . A result a study by Rafay et al. (2022) showed that 65% considered online education to be an unproductive teaching technique. Only 25% of the students required to be examined online however 34.5% were of the view that online oral examination was not good technique of examination . Medical students were asked about their attitudes towards theoretical skills learning by DEL format. The findings of the current study are consistent with those of Tayem et al (2022) who found that 73.3% of medical students preferred distance learning for theoretical parts and students described that distance learning, improved interaction with teachers and peers. (45.6% and 48.9%, respectively). 60.1% of medical students were comfortable with online assessment . In the study by Elzainy et al. (2020), 60% of medical students strongly agreed or agreed with the effectiveness of online assessments for determining their knowledge level, and 80% strongly agreed or agreed with online teaching of some theoretical courses . Medical students were asked to report on their perceived level of instructor competencies and infrastructure support (See ). In a study by Mortagy et al. (2022), many medical students (44.1%) said that their instructors were not well ready for online education. A significant percentage of medical students (45.2%) stated that having internet. difficulties such as internet connection problems and or internet speed . In study of by Saurabh et al. (2021), about 62.7% of medical students had internet access. 67% of students were prepared to actively communicate with their peers and teachers electronically. 20.5% of medical students had linked online learning from home to conventional lectures . However, with a small sample size, caution must be applied, as the findings might not be transferable to other contexts. Although the current study is a descriptive study that uses the perceptions of medical students in Iran University the findings must be inferred with caution. Medical students had different opinion about importance and application of the distance eLearning in clinical education. Some of them believed that this method was a good tool that helped them in their coverage of most of the practical topics required in the Gynecology ward and their teachers had enough skills and experience to teach theoretical subjects virtually and this method provided an opportunity for discussion, teamwork and reflection for them. On the one hand, participants were not believed Distance E-Learning was an appropriate method for learning basic skills and practicing clinical skills effectively. Faculty development of medical teachers in online teaching methods also need to be considered. The results of this study have implications for policy and practice with regard to distance eLearning clinical education training in Iran. S1 File (SAV) Click here for additional data file. |
Improved rock phosphate dissolution from organic acids is driven by nitrate assimilation of bacteria isolated from nitrate and CaCO | 2a8073e4-2b5b-404d-a977-6b087bb5b404 | 10038309 | Microbiology[mh] | Phosphorus (P) is essential for life and is one of the key macro-nutrients for plant growth and development, including photosynthesis, energy transfers and nutrient uptake in plants [ – ]. Although organic and inorganic P forms are abundant in soils, they are highly insoluble (95–99%) with little mobility and low availability in most soil conditions . Therefore, P availability is often limiting plant growth due to its poor use efficiency (5–30%). Unfortunately, even when using high rates of chemical P-fertilizers application a considerable amount of the applied P-based chemical fertilizer is rapidly immobilized and precipitated with cations such as Ca 2+ , Al 3+ and Fe 3+ , which lead to the formation of poorly available P for plants . Besides, excessive application of P causes environmental and economic problems due to soil erosion and runoff water containing large amount of soluble P . The application of raw forms of P such as rock phosphate (RP), which is the source of P fertilizers, has been extensively studied as a method to overcome the problems of P limitation in agronomic crops . Rock phosphates are less expensive than purified and formulated P fertilizers, but are insoluble especially in soils with high pH and low organic matter, limiting its direct use as soil amendment . Faced with this situation, the idea arises to implement low technological cost alternatives to improve the acquisition of P by plants from RP; these include the use of rhizosphere and endophyte microorganisms that promote the solubilization of P from chemically forms such as RP [ – ]. Phosphate-solubilizing bacteria (PSB) play a fundamental role in biogeochemical P cycling in agricultural ecosystems, as they are involved in the transformation of insoluble forms of P into monobasic and dibasic phosphate (HPO 4 2- , H 2 PO 4 - ) available to plants . The better understanding of the mechanisms responsible of RP solubilization takes a crucial role about how to improve the efficiency of PSB . Although several mechanisms may be responsible of P solubilization, the main one is through the production of organic acids which involves either: (i) acidification by lowering pH, (ii) chelation of cations (mainly Ca 2+ ) bound to phosphate through their hydroxyl and carboxyl groups, (iii) exchange reactions with phosphate for adsorption sites or (iv) formation of soluble complexes with metals [ – ]. Proton-excretion accompanying NH 4 + assimilation is thought to be an alternative mechanism of P solubilization which has been found in some bacterial species, i.e. Pseudomonas fluorescens RAF15 or Bacillus marisflavi FA7 . At present, a key knowledge gap is the role that N supply plays in the mechanisms of RP solubilization since PSB need N and organic C to produce organic acids that solubilize the P. Besides, the production of organic acids requires N for transcription and translation . Since N addition is required for the production of solubilizing compounds, the N-source could control the extent of acid production by PSB which would be crucial to evaluate the suitability of PSB as effective RP solubilizer in the plant rhizosphere or for their use in the biotechnological production of RP-based fertilizers. In this work, we firstly hypothesized that P solubilization of RP depends on the release of organic acids by PSB strains and is greatly influenced by the N source (NO 3 - and NH 4 + ) supplied to the bacteria. We secondly hypothesized that the physical-chemical properties of soils such as texture, pH, available-P, N-NH 4 + , N-NO 3 - and CaCO 3 might also affect the bacterial features in terms of RP solubilization. To test the first and second hypotheses, we used a collection of two hundred and forty-three strains isolated from three contrasted soils. The isolates were at first screened using the National Botanical Research Institute´s phosphate NBRIP agarose media with NH 4 + or NO 3 - as the sole N-source. Then, their capacity to release soluble-P from RP was evaluated to select the strains with the highest percentage of RP-solubilization. The effect of the two N-sources on the bacterial growth, pH changes and organic acid production in the culture medium associated with the capacity to solubilize the RP was also reported. Thirdly, the isolates with the highest capacity to solubilize the RP were identified based on sequencing and phylogenetic analysis of the 16S rRNA gene.
Soil sampling Three sites of sampling, differing in their physico-chemicals properties, were selected to isolate rock phosphate solubilizing bacteria. Soils were identified as “A”, “B”, and “C” ( ) and were collected from the top soil (0–20 cm depth) of cropped fields. The soil A was collected from the DIASCOPE INRAE experimental station located in the South of France (Mauguio) (43.612°N; 3.976°E). It was classified as Skeletic Rhodic Luvisols , and was characterized by its neutral to alkaline pH and high stones content. The soil B, was sampled from the experimental site of Restinclières located at Prades-le-lez, 15km North of Montpellier (43°42′15 N, 3°51′41 E) and was classified as a deep Fluvisol (WRB, 2007) with high level of CaCO3 leading to alkaline pH . The soil C, was sampled from the “Faidherbi-Flux” collaborative observatory for greenhouse gas balance and ecosystem services ( https://lped.info/wikiObsSN/?Faidherbia-Flux ). It is located in the natural agro-silvo-pastoral parkland of Sob (14◦29′45N, 16◦27′13W), 135 km East of Dakar, West Senegal . It is classified as an Arenosol with very low level of C and P. All soil samples were collected from three pseudo replicates then mixed and immediately homogenized, sieved through a 2-mm diameter mesh, and stored at 4 ºC until analysis. Rock phosphate The rock phosphate (RP) used in this study was supplied by the OCP (Office Chérifien des Phosphates), Ben Guerir, Morocco. The analysis for the total contents of the elements on its composition was performed through inductively coupled plasma (ICP) at the laboratory of Arras (France) ( ). Isolation of bacterial strains For bacterial isolation, 1 g of soil was dispersed in 100 mL of 1% NaCl in 250 mL Erlenmeyer flasks by shaking in an orbital shaker. Subsequently, several serial dilutions (10 −2 –10 −7 ) were prepared in NaCl and 0.1 mL of these dilutions were spread with sterile glass rods in Petri dishes filled with solidified CASO agar (Sigma) medium (containing 15 g L -1 casein peptone, 5 g L -1 soy peptone and 15 g L -1 agar). The Petri plates were incubated for 4 weeks at 18 ºC in the dark. After incubation, the plates with the lowest dilutions and the least numbers of colonies were chosen for bacterial isolation. Subsequently, starting from the most dilute plates, the first 200 colonies encountered were transferred to new Petri dishes (60 mm) with CASO agar media and were incubated in the dark at 18 ºC for three weeks. From each soil sample, unpurified and non-growing colonies were discarded. Afterwards, bacterial colonies of each soil sample (A, B and C) were chosen at random. Each strain was numbered, and 93, 100, and 50 numbers were randomly selected for soil samples A, B, and C, respectively. In total, 243 isolates were further evaluated on their abilities to solubilize the rock phosphate. Screening of RP-solubilization bacteria with different N-sources A «spot test» was used to screen rapidly the ability of bacterial strains previously isolated to solubilize the rock phosphate according to the N-source, that was either NH 4 + or NO 3 - . The screening was run on agarose square Petri dishes (12 × 12 cm) which were filled with 70 mL of National Botanical Research Institute´s phosphate (NBRIP) medium with some modifications and containing, per liter: 10 g glucose, 5 g MgCl 6 *6 H 2 O, 0.25 g MgSO 4 *7H 2 O, 0.2 g KCl, 2.5 g rock phosphate, 0.1 g (NH 4 ) 2 SO 4 or KNO 3 , 1 mL of a vitamin solution (containing, per liter: 5 g panthotenate, 20 g inositol, 2 g nicotinic acid, 0.25 g pyridoxal hydrochloride, 0.25 g thiamine hydrochloride, and 0.01 g D-biotine), 0.006 g bromophenol blue (pH indicator dye), and 18 g agarose. A pre-inoculum of each bacterial strain was prepared in a 96-well plate containing LB medium and incubated for 3 days at 28 ºC. After that, 10 μL of each bacterial suspension was inoculated on NBRIP agarose medium with RP and NH 4 + or NO 3 - as the sole source of P and N, respectively. The plates were then incubated at 28 ºC till the occurrence of a halo of solubilization (yellow area around the colony) appearing generally after 7 days of incubation. Hence, all colonies were checked at 7 days, and only those producing halos were selected. Bacterial growth, RP solubilisation capacity and production of organic acids in liquid NBRIP media with different N-sources Three experiments were conducted to test the capacity of the isolates to solubilize the RP by organic acids production with different N-sources. The first experiment was conducted to measure quantitatively the percentage of the RP solubilization (% RP-solubilization value) on the previous selected colonies to identify the isolates with the highest capacities to release the soluble-P for further experiments. To do this, 200 μL of each bacterial culture previously grown in LB media for 72 h at 28°C (exponential phase) were used to inoculate 50 mL of NBRIP liquid media with the same composition as above-mentioned. The flask containing the inoculated media was incubated for 7 days at 28 ºC. A non-inoculated treatment was also set up. At 7 days, 1 mL of the NBRIP liquid medium was collected and centrifuged for 10 min at 3,100 rcf to obtain a free-bacterial NBRIP medium to measure the soluble-P content using the malachite green method . The % RP-solubilization value of each bacterial isolate was then calculated as follow: (soluble-P isolate /P total), with soluble-P isolate as the soluble-P content (μg mL -1 ) release from the RP into the NBRIP liquid medium, and P total indicated the total P added in the NBRIP liquid media as RP (2.5 g L -1 ). Since a residual soluble-P content release from the RP was expected as the result of the incubation, the % RP-solubilization value was also calculated for the non-inoculated treatments. The experiment was run in triplicate. The second experiment was set up to evaluate the bacterial growth and RP-solubilization capacity on the above-selected isolates after 0, 3, 7, 14 and 21 days of culture. 1 mL of the bacterial suspensions, produced as before, was used to measure the optical density (OD) at 600 nm as an indirect measurement of bacterial growth using a UV-Visible Spectrometer (Spectronis Helios α UV-Vis). Afterwards, the bacterial suspension was subjected to a centrifugation for 10 min at 3,100 rcf to obtain a free-bacterial NBRIP medium which was used to measure the pH with a pH microelectrode (Fisherbrand ™ , diameter of 3 mm), and the soluble-P content using the malachite green method. The third experiment was performed to identify and to quantify the production of organic acids in the free-bacterial NBRIP liquid medium of each bacterial strain after 21 days of incubation. The free-bacterial NBRIP liquid medium with NH 4 + as the sole N-source contained salts such as MgCl 2 *6H 2 O, KCl, (NH 4 ) 2 SO 4 and MgSO 4 *7H 2 O, resulting in high amounts of Cl - and SO 4 = ions incompatible with HPIC analysis. To remove these ions, the free-bacterial NBRIP liquid media were filtrated using cartridges in series containing Ag (Dionex OnGuard II-AG cartridge, Thermo Scientific) and Ba (Dionex OnGuard II-Ba cartridge, Thermo Scientific), trapping Cl - and SO 4 = , respectively. Meanwhile, the free-bacterial NBRIP liquid media with NO 3 - as the sole N-source contained high amounts of salts such as MgCl 2 *6H 2 O and KCl, resulting in elevated Cl - concentrations. In this case, Cl - ions were eliminated using only cartridges containing Ag (Dionex OnGuard II-AG cartridge, Thermo Scientific). The cartridges were rinsed with 10 ml ultrapure water before filtering 1 ml of each free-bacterial NBRIP liquid medium through the cartridge. The two free-bacterial NBRIP liquid media with NH 4 + or NO 3 - were also passed through a 0.45 μm filter and diluted with ultrapure water before to be analyzed in an HPIC system (Dionex IonPac) equipped with an AS11-HC-11 μm column (4 × 250 mm) and a Dionex AERS 500 suppressor. All separations were performed at a flow rate of 1 ml min −1 . We tested various KOH gradients in the mobile phase to improve the separation of organic anions from mineral ones. Finally, the gradient composition we used for our analyses is given in . The column was at room temperature (25 ± 2 °C). The injection volume was 25 μl taken from the sample prepared in a 1 ml vial. Solutions of 15 organic anions were first prepared from the synthetic compounds listed in . For each compound, different solutions were prepared in ultrapure water that were (1) a stock solution, kept at 4°C; (2) a solution for determining the retention time (RT) of the organic anion and (3) solutions with 7 final concentrations for building a calibration curve ( ). Each sample of free-bacterial NBRIP medium was injected 3 times (100 μL per injection). Peaks on each chromatogram were compared with the chromatogram obtained with standard solution to identify each organic acid, and concentrations were extrapolated from the standard curve. 16S rRNA gene sequencing from bacterial isolates The taxonomic identities of bacterial strains showing the highest ability to solubilize the RP with NH 4 + or NO 3 - were assigned by 16S rRNA gene sequence analysis. Bacterial genomic DNA was extracted by suspending a few colonies from each strain in 1 mL of MilliQ water in a microcentrifuge tube. For nearly full-length amplification of the 16S rRNA, the primer pair FD1 ( 5ʹ -AGAGTTTGATCCTGGCTCAG- 3ʹ ) and RP2 ( 5ʹ- ACGGCTACCTTGTTACGACTT-3ʹ ) was used . PCR mixtures were composed of 5 μL PCR buffer (10x), 4 μL MgCl 2 (final concentration 2 mM), 1 μL dNTPs (10 mM each), 1 μL each forward and reverse primers (10 mM), 0.3 μL Taq polymerase (5 U μL -1 Qbiogene) and 5 μL DNA solution in a total volume of 50 μL. The thermal cycling program was: i) 5 min at 95 ºC, ii) 35 cycles: 15 s at 94 ºC, 30 s at 60 ºC, and 90 min at 72 ºC, iii) 10 min at 72 ºC. PCR products were purified and sequenced using the internal primers 926F ( 5ʹ -AAACTYAAAKGAATTGACGG-3 ʹ ) and 1100R ( 5ʹ -GGGTTGCGCTCGTTG- 3ʹ ) at Genewiz (Leipzig, Germany). The sequence obtained for each isolate was compared for similarity level with the reference strains from genomic database banks using Rdp taxonomy tool available at the https://www.rdp.cme.msu.edu web site. The sequences were registered in NCBI under the number given in Table 5. The phylogenetic tree was built using phylogenia.fr [ – ]. Data analysis Unless otherwise stated, the results are given as mean ± standard error (n = 3). Statistical analysis was carried out using Statistica 13 software. The normality was firstly tested using a Levene test. When normality was reached, the ability of isolates to solubilize the RP (soluble-P content and % RP-solubilization ability) with NH 4 + or NO 3 - as the sole N-source after 7 days of culture was analyzed using one-way ANOVA to select the most effectives isolates. The same was performed on the organic acids produced by the bacterial strains after 21 days of incubation. Comparison of means was performed by Tukey´s HSD post hoc at three levels of significance: *p<0.05; **p<0.01; ***p<0.001. Two-way ANOVA with repeated measures was run to evaluate significant differences among bacterial strains and time of incubation on bacterial growth (OD 600 nm), soluble-P content and pH. To gain a better understanding about the distribution of the physical-chemical properties (clay, fine silt, coarse silt, fine sand, coarse sand, CaCO 3 , total C, total N, N-NO 3 - , N-NH 4 + , Olsen P and pH) in the three soil samples collected (A, B and C), a principal component analysis (PCA) was done using the CANOCO 5.0 for windows. To gain a new knowledge about the influence of the N source in the production of organic acids and thus, in the RP-solubilization, an integrated multivariate analysis was also run, using the same software, to determinate a new set of correlations among: i) the RP-solubilization isolates from soil A, B and C with NH 4 + or NO 3 - as the sole N-source, ii) the soluble-P content and the organic acids: gluconic, lactic, glycolic, acetic, butyric, formic, propionic, pyruvic, malic, maleic, oxalic and citric acids measured in the NBRIP liquid medium after 21 days of culture, and iii) the physical-chemical soil properties from soil A (Olsen P, N-NH 4 + content and coarse silt), B (N-NO3- and CaCO 3 content) and C (coarse and fine sand) selected through the first PCA analysis. These statistical analyses were run on reduced and centered variables.
Three sites of sampling, differing in their physico-chemicals properties, were selected to isolate rock phosphate solubilizing bacteria. Soils were identified as “A”, “B”, and “C” ( ) and were collected from the top soil (0–20 cm depth) of cropped fields. The soil A was collected from the DIASCOPE INRAE experimental station located in the South of France (Mauguio) (43.612°N; 3.976°E). It was classified as Skeletic Rhodic Luvisols , and was characterized by its neutral to alkaline pH and high stones content. The soil B, was sampled from the experimental site of Restinclières located at Prades-le-lez, 15km North of Montpellier (43°42′15 N, 3°51′41 E) and was classified as a deep Fluvisol (WRB, 2007) with high level of CaCO3 leading to alkaline pH . The soil C, was sampled from the “Faidherbi-Flux” collaborative observatory for greenhouse gas balance and ecosystem services ( https://lped.info/wikiObsSN/?Faidherbia-Flux ). It is located in the natural agro-silvo-pastoral parkland of Sob (14◦29′45N, 16◦27′13W), 135 km East of Dakar, West Senegal . It is classified as an Arenosol with very low level of C and P. All soil samples were collected from three pseudo replicates then mixed and immediately homogenized, sieved through a 2-mm diameter mesh, and stored at 4 ºC until analysis.
The rock phosphate (RP) used in this study was supplied by the OCP (Office Chérifien des Phosphates), Ben Guerir, Morocco. The analysis for the total contents of the elements on its composition was performed through inductively coupled plasma (ICP) at the laboratory of Arras (France) ( ).
For bacterial isolation, 1 g of soil was dispersed in 100 mL of 1% NaCl in 250 mL Erlenmeyer flasks by shaking in an orbital shaker. Subsequently, several serial dilutions (10 −2 –10 −7 ) were prepared in NaCl and 0.1 mL of these dilutions were spread with sterile glass rods in Petri dishes filled with solidified CASO agar (Sigma) medium (containing 15 g L -1 casein peptone, 5 g L -1 soy peptone and 15 g L -1 agar). The Petri plates were incubated for 4 weeks at 18 ºC in the dark. After incubation, the plates with the lowest dilutions and the least numbers of colonies were chosen for bacterial isolation. Subsequently, starting from the most dilute plates, the first 200 colonies encountered were transferred to new Petri dishes (60 mm) with CASO agar media and were incubated in the dark at 18 ºC for three weeks. From each soil sample, unpurified and non-growing colonies were discarded. Afterwards, bacterial colonies of each soil sample (A, B and C) were chosen at random. Each strain was numbered, and 93, 100, and 50 numbers were randomly selected for soil samples A, B, and C, respectively. In total, 243 isolates were further evaluated on their abilities to solubilize the rock phosphate.
A «spot test» was used to screen rapidly the ability of bacterial strains previously isolated to solubilize the rock phosphate according to the N-source, that was either NH 4 + or NO 3 - . The screening was run on agarose square Petri dishes (12 × 12 cm) which were filled with 70 mL of National Botanical Research Institute´s phosphate (NBRIP) medium with some modifications and containing, per liter: 10 g glucose, 5 g MgCl 6 *6 H 2 O, 0.25 g MgSO 4 *7H 2 O, 0.2 g KCl, 2.5 g rock phosphate, 0.1 g (NH 4 ) 2 SO 4 or KNO 3 , 1 mL of a vitamin solution (containing, per liter: 5 g panthotenate, 20 g inositol, 2 g nicotinic acid, 0.25 g pyridoxal hydrochloride, 0.25 g thiamine hydrochloride, and 0.01 g D-biotine), 0.006 g bromophenol blue (pH indicator dye), and 18 g agarose. A pre-inoculum of each bacterial strain was prepared in a 96-well plate containing LB medium and incubated for 3 days at 28 ºC. After that, 10 μL of each bacterial suspension was inoculated on NBRIP agarose medium with RP and NH 4 + or NO 3 - as the sole source of P and N, respectively. The plates were then incubated at 28 ºC till the occurrence of a halo of solubilization (yellow area around the colony) appearing generally after 7 days of incubation. Hence, all colonies were checked at 7 days, and only those producing halos were selected.
Three experiments were conducted to test the capacity of the isolates to solubilize the RP by organic acids production with different N-sources. The first experiment was conducted to measure quantitatively the percentage of the RP solubilization (% RP-solubilization value) on the previous selected colonies to identify the isolates with the highest capacities to release the soluble-P for further experiments. To do this, 200 μL of each bacterial culture previously grown in LB media for 72 h at 28°C (exponential phase) were used to inoculate 50 mL of NBRIP liquid media with the same composition as above-mentioned. The flask containing the inoculated media was incubated for 7 days at 28 ºC. A non-inoculated treatment was also set up. At 7 days, 1 mL of the NBRIP liquid medium was collected and centrifuged for 10 min at 3,100 rcf to obtain a free-bacterial NBRIP medium to measure the soluble-P content using the malachite green method . The % RP-solubilization value of each bacterial isolate was then calculated as follow: (soluble-P isolate /P total), with soluble-P isolate as the soluble-P content (μg mL -1 ) release from the RP into the NBRIP liquid medium, and P total indicated the total P added in the NBRIP liquid media as RP (2.5 g L -1 ). Since a residual soluble-P content release from the RP was expected as the result of the incubation, the % RP-solubilization value was also calculated for the non-inoculated treatments. The experiment was run in triplicate. The second experiment was set up to evaluate the bacterial growth and RP-solubilization capacity on the above-selected isolates after 0, 3, 7, 14 and 21 days of culture. 1 mL of the bacterial suspensions, produced as before, was used to measure the optical density (OD) at 600 nm as an indirect measurement of bacterial growth using a UV-Visible Spectrometer (Spectronis Helios α UV-Vis). Afterwards, the bacterial suspension was subjected to a centrifugation for 10 min at 3,100 rcf to obtain a free-bacterial NBRIP medium which was used to measure the pH with a pH microelectrode (Fisherbrand ™ , diameter of 3 mm), and the soluble-P content using the malachite green method. The third experiment was performed to identify and to quantify the production of organic acids in the free-bacterial NBRIP liquid medium of each bacterial strain after 21 days of incubation. The free-bacterial NBRIP liquid medium with NH 4 + as the sole N-source contained salts such as MgCl 2 *6H 2 O, KCl, (NH 4 ) 2 SO 4 and MgSO 4 *7H 2 O, resulting in high amounts of Cl - and SO 4 = ions incompatible with HPIC analysis. To remove these ions, the free-bacterial NBRIP liquid media were filtrated using cartridges in series containing Ag (Dionex OnGuard II-AG cartridge, Thermo Scientific) and Ba (Dionex OnGuard II-Ba cartridge, Thermo Scientific), trapping Cl - and SO 4 = , respectively. Meanwhile, the free-bacterial NBRIP liquid media with NO 3 - as the sole N-source contained high amounts of salts such as MgCl 2 *6H 2 O and KCl, resulting in elevated Cl - concentrations. In this case, Cl - ions were eliminated using only cartridges containing Ag (Dionex OnGuard II-AG cartridge, Thermo Scientific). The cartridges were rinsed with 10 ml ultrapure water before filtering 1 ml of each free-bacterial NBRIP liquid medium through the cartridge. The two free-bacterial NBRIP liquid media with NH 4 + or NO 3 - were also passed through a 0.45 μm filter and diluted with ultrapure water before to be analyzed in an HPIC system (Dionex IonPac) equipped with an AS11-HC-11 μm column (4 × 250 mm) and a Dionex AERS 500 suppressor. All separations were performed at a flow rate of 1 ml min −1 . We tested various KOH gradients in the mobile phase to improve the separation of organic anions from mineral ones. Finally, the gradient composition we used for our analyses is given in . The column was at room temperature (25 ± 2 °C). The injection volume was 25 μl taken from the sample prepared in a 1 ml vial. Solutions of 15 organic anions were first prepared from the synthetic compounds listed in . For each compound, different solutions were prepared in ultrapure water that were (1) a stock solution, kept at 4°C; (2) a solution for determining the retention time (RT) of the organic anion and (3) solutions with 7 final concentrations for building a calibration curve ( ). Each sample of free-bacterial NBRIP medium was injected 3 times (100 μL per injection). Peaks on each chromatogram were compared with the chromatogram obtained with standard solution to identify each organic acid, and concentrations were extrapolated from the standard curve.
The taxonomic identities of bacterial strains showing the highest ability to solubilize the RP with NH 4 + or NO 3 - were assigned by 16S rRNA gene sequence analysis. Bacterial genomic DNA was extracted by suspending a few colonies from each strain in 1 mL of MilliQ water in a microcentrifuge tube. For nearly full-length amplification of the 16S rRNA, the primer pair FD1 ( 5ʹ -AGAGTTTGATCCTGGCTCAG- 3ʹ ) and RP2 ( 5ʹ- ACGGCTACCTTGTTACGACTT-3ʹ ) was used . PCR mixtures were composed of 5 μL PCR buffer (10x), 4 μL MgCl 2 (final concentration 2 mM), 1 μL dNTPs (10 mM each), 1 μL each forward and reverse primers (10 mM), 0.3 μL Taq polymerase (5 U μL -1 Qbiogene) and 5 μL DNA solution in a total volume of 50 μL. The thermal cycling program was: i) 5 min at 95 ºC, ii) 35 cycles: 15 s at 94 ºC, 30 s at 60 ºC, and 90 min at 72 ºC, iii) 10 min at 72 ºC. PCR products were purified and sequenced using the internal primers 926F ( 5ʹ -AAACTYAAAKGAATTGACGG-3 ʹ ) and 1100R ( 5ʹ -GGGTTGCGCTCGTTG- 3ʹ ) at Genewiz (Leipzig, Germany). The sequence obtained for each isolate was compared for similarity level with the reference strains from genomic database banks using Rdp taxonomy tool available at the https://www.rdp.cme.msu.edu web site. The sequences were registered in NCBI under the number given in Table 5. The phylogenetic tree was built using phylogenia.fr [ – ].
Unless otherwise stated, the results are given as mean ± standard error (n = 3). Statistical analysis was carried out using Statistica 13 software. The normality was firstly tested using a Levene test. When normality was reached, the ability of isolates to solubilize the RP (soluble-P content and % RP-solubilization ability) with NH 4 + or NO 3 - as the sole N-source after 7 days of culture was analyzed using one-way ANOVA to select the most effectives isolates. The same was performed on the organic acids produced by the bacterial strains after 21 days of incubation. Comparison of means was performed by Tukey´s HSD post hoc at three levels of significance: *p<0.05; **p<0.01; ***p<0.001. Two-way ANOVA with repeated measures was run to evaluate significant differences among bacterial strains and time of incubation on bacterial growth (OD 600 nm), soluble-P content and pH. To gain a better understanding about the distribution of the physical-chemical properties (clay, fine silt, coarse silt, fine sand, coarse sand, CaCO 3 , total C, total N, N-NO 3 - , N-NH 4 + , Olsen P and pH) in the three soil samples collected (A, B and C), a principal component analysis (PCA) was done using the CANOCO 5.0 for windows. To gain a new knowledge about the influence of the N source in the production of organic acids and thus, in the RP-solubilization, an integrated multivariate analysis was also run, using the same software, to determinate a new set of correlations among: i) the RP-solubilization isolates from soil A, B and C with NH 4 + or NO 3 - as the sole N-source, ii) the soluble-P content and the organic acids: gluconic, lactic, glycolic, acetic, butyric, formic, propionic, pyruvic, malic, maleic, oxalic and citric acids measured in the NBRIP liquid medium after 21 days of culture, and iii) the physical-chemical soil properties from soil A (Olsen P, N-NH 4 + content and coarse silt), B (N-NO3- and CaCO 3 content) and C (coarse and fine sand) selected through the first PCA analysis. These statistical analyses were run on reduced and centered variables.
Screening of RP-solubilization bacteria under different N sources A total of two hundred and forty-three bacterial colonies, isolated from soils A, B and C, were screened on their abilities to solubilize the RP using the «spot test» on NBRIP agarose media with NH 4 + or NO 3 - as the sole N-source. In total, forty isolates were able to solubilize the RP due to the presence of halo with NH 4 + or NO 3 - . Twenty-three isolates out of these forty strains had the capabilities to solubilize the RP (presence of solubilization halo) with NH 4 + , while seventeen isolates out of forty strains produced clear halos in NBRIP agarose media with NO 3 - . Remarkably, none of the isolate had same ability to solubilize the RP with either NH 4 + or NO 3 - as N-source, except one isolate displaying a halo on both N-sources. Bacterial growth and RP solubilization capacity under different N-sources A first set of experiments was established to evaluate the capacity of the bacterial strains, producing a halo of solubilization, to release the soluble-P content from RP and thus, measure the % RP-solubilization value with NH 4 + or NO 3 - as the sole N-source, after 7 days of culture ( ). A residual amount of soluble-P was detected in each non-inoculated NBRIP liquid media (NI) with NH 4 + or NO 3 - at 7 days, reaching values between 8.2–9.6 μg mL -1 , respectively. When NH 4 + was supplied as the sole N-source, only nine isolates, identified as 59B, 24A, 4A, 9C, 87B, 12A, 47A, 15A and 46B out of twenty-three strains significantly increased the % RP-solubilization values ranging from 3.5 to 9.2% after 7 days of incubation. Meanwhile, with NO 3 - as the sole N-source, seven strains, labeled as 23B, 48B, 41C, 39B, 59B, 32A and 6C out of seventeen isolates strongly release the soluble-P from the RP with % RP-solubilization values ranging between 2.4–36% ( ). The isolate 59B, which was able to solubilize the RP with NH 4 + or NO 3 - , showed higher % RP-solubilization with NH 4 + (9%) compared to NO 3 - (3%). The isolates with the highest % RP-solubilization were then selected for the experiment described below. The second set of experiments was set up to evaluate the bacterial growth (OD 600 nm), soluble-P content and pH values of the above-selected isolates inoculated in the NBRIP liquid media with RP and NH 4 + or NO 3 - as the sole source of P and N after 3, 7, 14 and 21 days of culture (Figs and ). When NH 4 + was added as the sole N-source, the OD detected in the non-inoculated NBRIP liquid media reached values up to 0.3 as the result of the slow release of soluble-P content from RP after 21 days of incubation ( ). However, all isolates were found to increase gradually the OD over the time of culture, becoming higher than the non-inoculated NBRIP liquid media ( ). The highest bacterial growth was found for 12A, 15A, 4A and 46B isolates which significantly differ compared to the other strains at the end of the experiment ( ) (Bs, p≤0.000; T, p≤0.000). The residual soluble-P content detected in the non-inoculated NBRIP liquid media reached values up to 20 μg mL -1 after 21 days of incubation with steady pH values during the experiment ( ). However, all bacterial strains were capable to release the soluble-P from RP over the time of incubation with a concomitant pH drop after 3 days of culture (soluble-P content, Bs, p≤0.000; T, p≤0.000; pH, Bs, p≤0.000; T, p≤0.000) ( ). The isolates could be separated in two groups, the first one including three bacterial strains (24A, 87B and 59B) able to release soluble-P amounts higher than 50 μg mL -1 and to decrease concomitantly the pH value of the medium below 5. Among these strains, the 87B one was the most efficient to release soluble-P, with values ranging from 124 (day 14) to 50 (day 21) μg mL -1 and a final pH medium of 4.4 ( ). The two other isolates (24A and 59B) released also soluble-P in their medium with concentrations increasing regularly with the culture time, reaching 50 μg mL -1 after 21 days. Both isolates acidified their culture medium, with pH values about 4.8 ( ). The second group includes the 6 other isolates tested, namely 12A, 15A, 4A, 46B, 47A and 9C. The isolate 47A was less efficient than the other ones to release soluble-P amounting to 24 μg mL -1 , despite its ability to decrease the medium pH down to 5.5 ( ). All the other isolates were able to release similar amounts of soluble-P, with values ranging from 24 to 36 μg mL -1 ( ). As previously found in NH 4 + medium, the OD detected in the non-inoculated NBRIP liquid medium with NO 3 - as the sole N-source reached values up to 0.2 ( ). Even though all isolates increased gradually their growth over the time, the highest OD values were found for 39B and 6C isolates, in contrast with the lowest growth, which was found for 48B strain after 21 days of culture ( ) (Bs, p≤0.000; T, p≤0.000). A residual amount of soluble-P was detected in the non-inoculated NBRIP liquid medium reaching values up to 25 μg mL -1 after 21 days of incubation with no changes in the pH values during the experiment ( ). Overall, all isolates tested significantly increased the release of the soluble-P from RP over the time of experiment with a concomitant pH drop recorded at 3 days, which in some cases, then slowly declined until 21 days of culture (soluble-P content, Bs, p≤0.000; T, p≤0.000; pH, p≤0.000; T, p≤0.000) ( ). Among the seven strains tested, five of them were able to release high amounts of soluble-P with different kinetics. For example, isolate 23B was detected as the fastest isolate to release soluble-P with values up to 97 μg P mL -1 at 3 days, which were also related to a pH drop (4.1) ( ). In contrast, at 21 days, 59B, 41C, 32A, and 39B isolates were more efficient in releasing soluble-P from RP reaching values up to 159, 186, 129, and 156 μg P mL -1 , respectively, with pH values ranging from 4 to 4.8 ( ). Finally, the two remaining strains (48B and 6C) were found to be the less efficient ones in solubilizing the RP over the time of experiment although both strains were also able to reduce significantly the pH of the medium ( ). Identification and quantification of organic acids The third set of experiments was run to identify and to quantify the organic acids produced by the selected bacterial strains with RP and N-sources (NH 4 + versus NO 3 - ) as the sole source of P and N after 21 days of culture (Tables and ). The HPIC analysis revealed that with RP and NH 4 + as the sole source of P and N, the isolates were able to release eight organic acids (gluconic, lactic, glycolic, acetic, propionic, formic, pyruvic and maleic acids) after 21 days of incubation ( ). Among these organic acids, only three of them (gluconic, lactic and pyruvic acids) were released by all strains. Gluconic and lactic acids were produced at the highest rates, with gluconic acid concentrations reaching up to 9 and 11 mM released by isolates 12A and 47A, respectively, and lactic acid up to 7 mM released by isolates 47A and 87B. Pyruvic acid was released at rates ranging between 0.7 (isolate 87B) up to 5.7 mM (isolate 12A). Secretion of formic acid was released by all strains but 15A and 4A and was highly heterogeneous, with concentrations varying from 0.09 (isolate 12A) up to 8 mM (isolate 87B). Only three strains were able to produce propionic acid (12A, 47A and 9C) or maleic acid (59B, 4A and 9C). The strain 9C released the highest concentrations of propionic (1.6 mM) and maleic (2.6 mM) acids. Glycolic acid was found in the culture medium of five isolates at low concentrations, ranging from 0.004 (isolate 87B) to 0.2 mM (isolate 4A). Finally, acetic acid was only produced by 24A with values up to 3.3 mM. With NO 3 - as the sole N-source, ten organic acids were identified in the culture medium after 21 days of experiment ( ). Half of the organic acids (gluconic, lactic, glycolic, acetic and formic acids) displayed high concentrations in the culture medium, with values generally greater than 1 mM. The gluconic acid was produced by all strains except 23B, 39B and 59B, with the greatest concentration measured for 41C (5.6 mM). Conversely, lactic acid was secreted by all isolates with the highest concentration for 23B and 39B strains reaching values up to 61 and 52 mM, respectively. The majority of the isolates, except 23B, 59B and 6C, were able to secrete glycolic acid with the highest level for 39B (7.2 mM). The 23B, 48B and 39B strains did not secrete acetic acid, but the most efficient isolate was 32A reaching values up to 30 mM. Formic acid was produced by all isolates with the greatest values for 41C (19 mM). The five remaining organic acids (butyric, pyruvic, malic, oxalic and citric acids) were found at much lower concentrations than the previous ones, with values generally in the mM range. The 6C strain was found the only isolate with abilities to secrete butyric acid, but at very low concentration (0.04 mM). All bacterial strains were able to produce pyruvic acid except the 23B isolate. The greatest concentrations were found for 32A with values up to 13 mM. Malic acid was produced by 41C and 32A at greater rates than 6C, reaching values between 0.8 and 0.6 mM, respectively. Oxalic acid was only produced by 39B with concentrations up to 0.50 mM. Citric acid was secreted by all strains, except 23B and 59B. The greatest concentrations were found for 41C reaching values up to 1 mM. Identification of the bacterial isolates A fragment of about 1.5 Kb was obtained after amplification of the 16S rRNA performed on genomic DNA extracted from the fifteen isolated strains with abilities to solubilize the RP with NH 4 + or NO 3 - as the sole N-source. The comparison of sequences with data available from Rpd Taxonomy tool enabled us to identify the isolates at generic or species level with similarities between 100 and 93% ( ). The selected isolates were closely related to four specific genera, Rhizobium (46B), Pseudomonas (23B), Paenibacillus (32A, 24A and 6C) and Bacillus (12A, 15A, 4A, 59B, 87B, 47A, 9C, 39B, 41C and 48B). All sequences were submitted to NCBI Genbank and accession numbers are given in . The phylogenetic analysis based on the 16S rRNA gene sequences of the selected isolates and representative species of closely related taxa formed seven clearly distinguishable clusters (C 1 , C 2 , C 3 , C 4 , C 5 , C 6 and C 7 ) ( ). The first cluster (C 1 ) was formed by the strains 48B and 41C with a close relationship with genus Bacillus sp . The second cluster (C 2 ) was also related to Bacillus sp . and included five strains (15A, 4A, 47A, 9C, 12A) although the strains 9C and 12A were slightly more distant than the others. The third cluster (C 3 ), composed by 46B and 23B, was connected with Gram negative bacteria such as Sinorhizobium sp. and Pseudomonas fluorescens , respectively. The 32A isolate, grouped in the fourth cluster (C 4 ), had notably relationship with Paenibacillus massiliensis . The fifth cluster (C 5 ), including 24A and 6C strains, was remarkably related with Paenibacillus sp . and Paenibacillus polymyxa . 39B and 59B isolates were grouped into the sixth cluster (C 6 ) with closely relationship with Bacillus sp . and Bacillus megaterium . Surprisingly, the strain 87B was individually grouped forming the cluster (C 7 ) which was related with any species, even if a 94% of identity was found with Bacillus sp . ( ). PCA analysis A PCA bi-plot was run to analyzed the distribution of the soil physical-chemical properties of the three sites sampled (A, B and C) ( ). The analysis revealed that soil A was mostly associated with a high content in Olsen P, N-NH 4 + and coarse silt whilst soil B was rather related to a high CaCO 3 and N-NO 3 - content and soil C was characterized by the presence of coarse and silt sand fractions. Even if soils A and B showed an alkaline pH which differed from soil C (neutral pH values), the PCA analysis did not show any clear relation between pH and type of soil, as similarly was found for the total C and N content. Neither the clay texture nor fine silt was found as properties that intimately characterized the soil A or B. A set of correlations was performed to evaluate the distribution of bacterial strains according to their abilities to solubilize the rock phosphate under different N sources (NH 4 + or NO 3 - ) based on the variables measured on the NBRIP liquid media after 21 days of incubation (soluble-P content and organic acids: gluconic, lactic, glycolic, acetic, butyric, formic, propionic, pyruvic, malic, maleic, oxalic and citric acids) and the physical-chemical properties of the three soil sampled (Olsen P, N-NH 4 + , N-NO 3 - , CaCO 3 , coarse and fine sand fraction). This analysis indicated that 81% of the variability of the data was explained by the two first axes (54% and 27%, respectively) ( ). The first axis was mainly explained by the N source used to grow the bacteria to assess RP solubilisation (NH 4 + versus NO 3 - ) and the second axis by the soil texture, especially fine and coarse sand and the CaCO 3 and N-NO 3 - content. Interestingly, the bacterial strains able to solubilize RP either on NH 4 + or NO 3 - were clearly separated on the bi-plot, clustering with the dominant N source assayed in soil. In addition to the N source, the bacterial strains clustered also along axis 2, forming four groups. For instance, the first group located in the right-upper quadrant was formed by bacterial strains isolated from soil A and C (32A, 41C and 6C) grown with NO 3 - as the sole source of N in the NBRIP liquid media. The strains 32A and 41C were the most efficient in releasing the soluble-P content from RP as the result of the production of organic acids such as: malic, acetic, citric, formic, glycolic, pyruvic and oxalic whereas 6C was the only strains in producing butyric acid. The second group, found in the right-lower quadrant, was only composed by bacterial strains isolated from soil B (59B, 39B, 48B and 23B) still grown with NO 3 - as the sole source of N in the NBRIP liquid media. These bacteria were associated with soil N-NO 3 - and CaCO 3 contents and the production of gluconic, lactic, propionic and maleic acids. The third group, located in the left-upper quadrant, was composed by strains isolated from soil A (12A, 15A, 4A, 47A, 24A) that were grown on NH 4 + as the sole source of N in the NBRIP liquid media. This group was associated with the Olsen P and N-NH 4 + content of soil A. Finally, in the left-down quadrant was found the fourth cluster composed by bacterial strains isolated from soil B (59B, 87B and 46B) able to solubilize RP with NH4+ and which were associated with CaCO 3 and NO 3 - contents.
A total of two hundred and forty-three bacterial colonies, isolated from soils A, B and C, were screened on their abilities to solubilize the RP using the «spot test» on NBRIP agarose media with NH 4 + or NO 3 - as the sole N-source. In total, forty isolates were able to solubilize the RP due to the presence of halo with NH 4 + or NO 3 - . Twenty-three isolates out of these forty strains had the capabilities to solubilize the RP (presence of solubilization halo) with NH 4 + , while seventeen isolates out of forty strains produced clear halos in NBRIP agarose media with NO 3 - . Remarkably, none of the isolate had same ability to solubilize the RP with either NH 4 + or NO 3 - as N-source, except one isolate displaying a halo on both N-sources.
A first set of experiments was established to evaluate the capacity of the bacterial strains, producing a halo of solubilization, to release the soluble-P content from RP and thus, measure the % RP-solubilization value with NH 4 + or NO 3 - as the sole N-source, after 7 days of culture ( ). A residual amount of soluble-P was detected in each non-inoculated NBRIP liquid media (NI) with NH 4 + or NO 3 - at 7 days, reaching values between 8.2–9.6 μg mL -1 , respectively. When NH 4 + was supplied as the sole N-source, only nine isolates, identified as 59B, 24A, 4A, 9C, 87B, 12A, 47A, 15A and 46B out of twenty-three strains significantly increased the % RP-solubilization values ranging from 3.5 to 9.2% after 7 days of incubation. Meanwhile, with NO 3 - as the sole N-source, seven strains, labeled as 23B, 48B, 41C, 39B, 59B, 32A and 6C out of seventeen isolates strongly release the soluble-P from the RP with % RP-solubilization values ranging between 2.4–36% ( ). The isolate 59B, which was able to solubilize the RP with NH 4 + or NO 3 - , showed higher % RP-solubilization with NH 4 + (9%) compared to NO 3 - (3%). The isolates with the highest % RP-solubilization were then selected for the experiment described below. The second set of experiments was set up to evaluate the bacterial growth (OD 600 nm), soluble-P content and pH values of the above-selected isolates inoculated in the NBRIP liquid media with RP and NH 4 + or NO 3 - as the sole source of P and N after 3, 7, 14 and 21 days of culture (Figs and ). When NH 4 + was added as the sole N-source, the OD detected in the non-inoculated NBRIP liquid media reached values up to 0.3 as the result of the slow release of soluble-P content from RP after 21 days of incubation ( ). However, all isolates were found to increase gradually the OD over the time of culture, becoming higher than the non-inoculated NBRIP liquid media ( ). The highest bacterial growth was found for 12A, 15A, 4A and 46B isolates which significantly differ compared to the other strains at the end of the experiment ( ) (Bs, p≤0.000; T, p≤0.000). The residual soluble-P content detected in the non-inoculated NBRIP liquid media reached values up to 20 μg mL -1 after 21 days of incubation with steady pH values during the experiment ( ). However, all bacterial strains were capable to release the soluble-P from RP over the time of incubation with a concomitant pH drop after 3 days of culture (soluble-P content, Bs, p≤0.000; T, p≤0.000; pH, Bs, p≤0.000; T, p≤0.000) ( ). The isolates could be separated in two groups, the first one including three bacterial strains (24A, 87B and 59B) able to release soluble-P amounts higher than 50 μg mL -1 and to decrease concomitantly the pH value of the medium below 5. Among these strains, the 87B one was the most efficient to release soluble-P, with values ranging from 124 (day 14) to 50 (day 21) μg mL -1 and a final pH medium of 4.4 ( ). The two other isolates (24A and 59B) released also soluble-P in their medium with concentrations increasing regularly with the culture time, reaching 50 μg mL -1 after 21 days. Both isolates acidified their culture medium, with pH values about 4.8 ( ). The second group includes the 6 other isolates tested, namely 12A, 15A, 4A, 46B, 47A and 9C. The isolate 47A was less efficient than the other ones to release soluble-P amounting to 24 μg mL -1 , despite its ability to decrease the medium pH down to 5.5 ( ). All the other isolates were able to release similar amounts of soluble-P, with values ranging from 24 to 36 μg mL -1 ( ). As previously found in NH 4 + medium, the OD detected in the non-inoculated NBRIP liquid medium with NO 3 - as the sole N-source reached values up to 0.2 ( ). Even though all isolates increased gradually their growth over the time, the highest OD values were found for 39B and 6C isolates, in contrast with the lowest growth, which was found for 48B strain after 21 days of culture ( ) (Bs, p≤0.000; T, p≤0.000). A residual amount of soluble-P was detected in the non-inoculated NBRIP liquid medium reaching values up to 25 μg mL -1 after 21 days of incubation with no changes in the pH values during the experiment ( ). Overall, all isolates tested significantly increased the release of the soluble-P from RP over the time of experiment with a concomitant pH drop recorded at 3 days, which in some cases, then slowly declined until 21 days of culture (soluble-P content, Bs, p≤0.000; T, p≤0.000; pH, p≤0.000; T, p≤0.000) ( ). Among the seven strains tested, five of them were able to release high amounts of soluble-P with different kinetics. For example, isolate 23B was detected as the fastest isolate to release soluble-P with values up to 97 μg P mL -1 at 3 days, which were also related to a pH drop (4.1) ( ). In contrast, at 21 days, 59B, 41C, 32A, and 39B isolates were more efficient in releasing soluble-P from RP reaching values up to 159, 186, 129, and 156 μg P mL -1 , respectively, with pH values ranging from 4 to 4.8 ( ). Finally, the two remaining strains (48B and 6C) were found to be the less efficient ones in solubilizing the RP over the time of experiment although both strains were also able to reduce significantly the pH of the medium ( ).
The third set of experiments was run to identify and to quantify the organic acids produced by the selected bacterial strains with RP and N-sources (NH 4 + versus NO 3 - ) as the sole source of P and N after 21 days of culture (Tables and ). The HPIC analysis revealed that with RP and NH 4 + as the sole source of P and N, the isolates were able to release eight organic acids (gluconic, lactic, glycolic, acetic, propionic, formic, pyruvic and maleic acids) after 21 days of incubation ( ). Among these organic acids, only three of them (gluconic, lactic and pyruvic acids) were released by all strains. Gluconic and lactic acids were produced at the highest rates, with gluconic acid concentrations reaching up to 9 and 11 mM released by isolates 12A and 47A, respectively, and lactic acid up to 7 mM released by isolates 47A and 87B. Pyruvic acid was released at rates ranging between 0.7 (isolate 87B) up to 5.7 mM (isolate 12A). Secretion of formic acid was released by all strains but 15A and 4A and was highly heterogeneous, with concentrations varying from 0.09 (isolate 12A) up to 8 mM (isolate 87B). Only three strains were able to produce propionic acid (12A, 47A and 9C) or maleic acid (59B, 4A and 9C). The strain 9C released the highest concentrations of propionic (1.6 mM) and maleic (2.6 mM) acids. Glycolic acid was found in the culture medium of five isolates at low concentrations, ranging from 0.004 (isolate 87B) to 0.2 mM (isolate 4A). Finally, acetic acid was only produced by 24A with values up to 3.3 mM. With NO 3 - as the sole N-source, ten organic acids were identified in the culture medium after 21 days of experiment ( ). Half of the organic acids (gluconic, lactic, glycolic, acetic and formic acids) displayed high concentrations in the culture medium, with values generally greater than 1 mM. The gluconic acid was produced by all strains except 23B, 39B and 59B, with the greatest concentration measured for 41C (5.6 mM). Conversely, lactic acid was secreted by all isolates with the highest concentration for 23B and 39B strains reaching values up to 61 and 52 mM, respectively. The majority of the isolates, except 23B, 59B and 6C, were able to secrete glycolic acid with the highest level for 39B (7.2 mM). The 23B, 48B and 39B strains did not secrete acetic acid, but the most efficient isolate was 32A reaching values up to 30 mM. Formic acid was produced by all isolates with the greatest values for 41C (19 mM). The five remaining organic acids (butyric, pyruvic, malic, oxalic and citric acids) were found at much lower concentrations than the previous ones, with values generally in the mM range. The 6C strain was found the only isolate with abilities to secrete butyric acid, but at very low concentration (0.04 mM). All bacterial strains were able to produce pyruvic acid except the 23B isolate. The greatest concentrations were found for 32A with values up to 13 mM. Malic acid was produced by 41C and 32A at greater rates than 6C, reaching values between 0.8 and 0.6 mM, respectively. Oxalic acid was only produced by 39B with concentrations up to 0.50 mM. Citric acid was secreted by all strains, except 23B and 59B. The greatest concentrations were found for 41C reaching values up to 1 mM.
A fragment of about 1.5 Kb was obtained after amplification of the 16S rRNA performed on genomic DNA extracted from the fifteen isolated strains with abilities to solubilize the RP with NH 4 + or NO 3 - as the sole N-source. The comparison of sequences with data available from Rpd Taxonomy tool enabled us to identify the isolates at generic or species level with similarities between 100 and 93% ( ). The selected isolates were closely related to four specific genera, Rhizobium (46B), Pseudomonas (23B), Paenibacillus (32A, 24A and 6C) and Bacillus (12A, 15A, 4A, 59B, 87B, 47A, 9C, 39B, 41C and 48B). All sequences were submitted to NCBI Genbank and accession numbers are given in . The phylogenetic analysis based on the 16S rRNA gene sequences of the selected isolates and representative species of closely related taxa formed seven clearly distinguishable clusters (C 1 , C 2 , C 3 , C 4 , C 5 , C 6 and C 7 ) ( ). The first cluster (C 1 ) was formed by the strains 48B and 41C with a close relationship with genus Bacillus sp . The second cluster (C 2 ) was also related to Bacillus sp . and included five strains (15A, 4A, 47A, 9C, 12A) although the strains 9C and 12A were slightly more distant than the others. The third cluster (C 3 ), composed by 46B and 23B, was connected with Gram negative bacteria such as Sinorhizobium sp. and Pseudomonas fluorescens , respectively. The 32A isolate, grouped in the fourth cluster (C 4 ), had notably relationship with Paenibacillus massiliensis . The fifth cluster (C 5 ), including 24A and 6C strains, was remarkably related with Paenibacillus sp . and Paenibacillus polymyxa . 39B and 59B isolates were grouped into the sixth cluster (C 6 ) with closely relationship with Bacillus sp . and Bacillus megaterium . Surprisingly, the strain 87B was individually grouped forming the cluster (C 7 ) which was related with any species, even if a 94% of identity was found with Bacillus sp . ( ).
A PCA bi-plot was run to analyzed the distribution of the soil physical-chemical properties of the three sites sampled (A, B and C) ( ). The analysis revealed that soil A was mostly associated with a high content in Olsen P, N-NH 4 + and coarse silt whilst soil B was rather related to a high CaCO 3 and N-NO 3 - content and soil C was characterized by the presence of coarse and silt sand fractions. Even if soils A and B showed an alkaline pH which differed from soil C (neutral pH values), the PCA analysis did not show any clear relation between pH and type of soil, as similarly was found for the total C and N content. Neither the clay texture nor fine silt was found as properties that intimately characterized the soil A or B. A set of correlations was performed to evaluate the distribution of bacterial strains according to their abilities to solubilize the rock phosphate under different N sources (NH 4 + or NO 3 - ) based on the variables measured on the NBRIP liquid media after 21 days of incubation (soluble-P content and organic acids: gluconic, lactic, glycolic, acetic, butyric, formic, propionic, pyruvic, malic, maleic, oxalic and citric acids) and the physical-chemical properties of the three soil sampled (Olsen P, N-NH 4 + , N-NO 3 - , CaCO 3 , coarse and fine sand fraction). This analysis indicated that 81% of the variability of the data was explained by the two first axes (54% and 27%, respectively) ( ). The first axis was mainly explained by the N source used to grow the bacteria to assess RP solubilisation (NH 4 + versus NO 3 - ) and the second axis by the soil texture, especially fine and coarse sand and the CaCO 3 and N-NO 3 - content. Interestingly, the bacterial strains able to solubilize RP either on NH 4 + or NO 3 - were clearly separated on the bi-plot, clustering with the dominant N source assayed in soil. In addition to the N source, the bacterial strains clustered also along axis 2, forming four groups. For instance, the first group located in the right-upper quadrant was formed by bacterial strains isolated from soil A and C (32A, 41C and 6C) grown with NO 3 - as the sole source of N in the NBRIP liquid media. The strains 32A and 41C were the most efficient in releasing the soluble-P content from RP as the result of the production of organic acids such as: malic, acetic, citric, formic, glycolic, pyruvic and oxalic whereas 6C was the only strains in producing butyric acid. The second group, found in the right-lower quadrant, was only composed by bacterial strains isolated from soil B (59B, 39B, 48B and 23B) still grown with NO 3 - as the sole source of N in the NBRIP liquid media. These bacteria were associated with soil N-NO 3 - and CaCO 3 contents and the production of gluconic, lactic, propionic and maleic acids. The third group, located in the left-upper quadrant, was composed by strains isolated from soil A (12A, 15A, 4A, 47A, 24A) that were grown on NH 4 + as the sole source of N in the NBRIP liquid media. This group was associated with the Olsen P and N-NH 4 + content of soil A. Finally, in the left-down quadrant was found the fourth cluster composed by bacterial strains isolated from soil B (59B, 87B and 46B) able to solubilize RP with NH4+ and which were associated with CaCO 3 and NO 3 - contents.
The RP solubilization through PSB has been recently reported as a suitable strategy to cope with the needs of the P requirements of crops . In our study, two hundred and forty-three bacterial strains isolated from three agricultural fields were screened on their abilities to solubilize the RP using the selective NBRIP agarose medium through the «spot test» . Our results indicated that only 16% of isolates (forty strains out of two hundred and forty-three) produced a halo of solubilization surrounding the colonies, regardless the N source. This finding was not surprising due to the low solubility and complex chemical structure found in RP compared to other P-sources such as tricalcium phosphate as recently reported , which in turn resulted in a less efficient bacterial capacity for dissolving this P-source. Here, we also found that the N-source led to significant differences in the ability to dissolve the RP in the agarose medium with twenty-three isolates showing a halo of degradation with NH 4 + and seventeen positives strains with NO 3 - as N-source. However, except the 59B isolate, none of the strains showed the ability to solubilize the RP with either NH 4 + or NO 3 - . It has been reported that the formation of a visible halo/zone on NBRIP agarose medium is a not infallible method to select efficiently the P-solubilizers since they could solubilize important quantities of P in broth despite they did not show a halo zone in NBRIP plate . In fact, in our study, the % of RP-solubilization quantified in liquid media revealed that only nine isolates out of twenty-three (46B, 12A, 15A, 4A, 59B, 24A, 87B, 47A and 9C) were selected as the best RP-solubilizers with NH 4 + as N-source. Meanwhile, seven isolates out of seventeen (6C, 32A, 59B, 39B, 41C, 48B, 23B) showed the greatest % of RP-solubilization with NO 3 - . This finding could be explained by the fact that bacterial strains can lose their solubilization phenotype upon repeated sub-culturing, as previously found by . Our results also evidenced that except for 59B isolate, the % of RP-solubilization was higher with NO 3 - than with NH 4 + at 7 days. This observation was surprising since NH 4 + assimilation is usually accompanied of H + extrusion with the subsequent medium acidification, and thus favoring the release of soluble-P from insoluble P-sources . Our second experiment revealed that the highest biomass production found for 46B, 12A, 15A and 4A isolates with NH 4 + as the sole N-source, were not directly related with the highest soluble-P values after 21 days of experiment. proposed that the very effective P-uptake systems of microorganisms would enable the assimilation of P from the solution disturbing the equilibrium between insoluble/soluble-P, thus insoluble-P would be indirectly dissolved by continuously removing of P from the solution. However, a further experiment would be required to verify the amount of soluble-P converted into bacterial biomass. Conversely, the lowest biomass found for 87B isolate resulted in the highest values in soluble-P content, which confirmed a previous evidence showing that bacterial strains were more efficient in dissolving P-sources instead of spending bacterial sources to biomass production . However, the pH drop found for all isolates after 3 days of culture would negatively affect the bacterial growth. In any case, in our study, the pH decline was clearly associated with the RP dissolution since the 87B, 24A and 59B isolates with the lowest pH values, were found to be the most efficient strains increasing the release of the soluble-P from RP between 14 and 21 days of experiment. Meanwhile, the highest pH values detected for 47A isolate were associated with its lesser capacity in dissolving the RP. This was in line with inverse relationship between pH decline of liquid medium and “in vitro” RP-solubilization previously reported by , indicating that pH plays a major role in the solubilization of the inorganic phosphate. At 21 days, only 39B isolate was found to enhance significantly its growth with a concomitant increase in the soluble-P content with NO 3 - as the sole N-source. This result would be in line with the previous evidences reported by who found that mechanisms of P-solubilization depend on processes associated with microbial biomass production (i.e. NH 4 + -assimilation or respiration). Similarly, as it was previously found, the pH drop seemed to be the main responsible of the highest release of soluble-P from RP for 59B, 41C, 32A and 39B isolates at the end of the culture. The use of NH 4 + as a N-source produces acid by either a proton exchange mechanism and/or organic acid secretion with a better P-solubilization . However, our study showed that the assimilation of NO 3 - by bacterial strains resulted in better capabilities to dissolve the RP rather than NH 4 + at 21 days, as the second experiment evidenced but at 7 days. Additionally, the highest levels of soluble-P found for 23B isolate at 3 days suggested that depending on the strain, the mechanisms of RP-solubilization might be also more effective in the short term with NO 3 - as the sole source of N. The production and secretion of organic acids has been recognized as a major mechanism responsible for releasing soluble-P from RP . Gluconic, formic, citric, oxalic, lactic, succinic, glycolic and acetic acids are among organic acids produced by PSB. In addition to those, pyruvic, malic or fumaric acids were also identified . In our study, regardless the N-source, gluconic, lactic, glycolic, acetic, formic and pyruvic acids were identified as the main organic acids produced by almost all PSB, with the highest concentrations for 12A, 4A, 24A, 87B, 39B, 23B, 59B, 41C, 39B and 32A isolates at 21 days. The production of organic acids during the release of soluble-P from RP, including gluconic, lactic, acetic, oxalic and citric acids among others was demonstrated as directly associated with pH decrease [ , , ]. Our results clearly demonstrated the role of lactic, acetic and formic acids in the RP dissolution for 24A, 87B and 59B isolates when NH 4 + was the sole N-source. Meanwhile, the negative relation between pH and soluble-P content found for 23B, 59B, 41C, 32A and 39B isolates was directly associated with the production of gluconic, glycolic, lactic, acetic, formic, pyruvic, maleic, oxalic and citric acids with NO 3 - as the sole N-source. Interestingly, the concentrations of lactic, acetic and formic acids produced by these isolates were almost 10-fold higher with NO 3 - than with NH 4 + , which would be explained by an inhibitory or toxic effect of NH 4 + on the enzymes responsible of organic acid secretion, or uptake of other essential nutrients or altering the electrochemical gradient, as previously suggested . Approximately 86% of the isolates belonged to Bacillus and Paenibacillus genus, as confirmed the 16S rRNA sequence analysis, which are gram-positive bacteria holding important traits related to the ability to P-solubilization, making them available to be used with agronomical purposes . The isolates 87B, 59B, 39B, 41C, 24A and 32A, which were found to be the most promising strains in dissolving the RP, were closely related to Bacillus sp ., Bacillus megaterium , Bacillus pumilus , Paenibacillus polymyxa and Paenibacillus massiliensis , except for 87B, as revealed the phylogenetic tree. In contrast, only one isolate with the highest ability to secrete lactic acid was identified as Pseudomonas fluorescens (23B). In our study, bacterial features involved in the RP dissolution and their co-occurrence in soils would be also influenced by other factors different than N-source, such as physico-chemical properties as others authors suggested . However, this study only revealed that the elevated amounts in N-NO 3 - and CaCO 3 found in the soil B would be influencing the release of organic acids such as lactic, propionic, gluconic and maleic by the strains: 59B ( Bacillus sp.), 23B ( Pseudomonas fluorescens ), 39B ( Bacillus megaterium ) and 48B ( Bacillus sp.). Meanwhile, the Olsen P, N-NH 4 + content and coarse silt in soil A or the sandy texture in soil C seemed not to influence significantly the production of organic acids by neither 12A ( Bacillus subtilis ), 15A ( Bacillus sp.), 4A ( Bacillus sp.), 47A ( Bacillus sp.), 9C ( Bacillus sp.), 59B ( Bacillus sp.), 87B ( Bacillus sp.) nor 46B ( Sinorhizobium sp.). Interestingly, the PCA analysis revealed that the strains with abilities to dissolving the RP with NH 4 + as source of N were lesser efficient in releasing organic acids, since any positive relation found. Conversely, the strains, 32A ( Paenibacillus massiliensis ), 41C ( Bacillus pumilus ) and 6C ( Paenibacillus sp .), which were found to assimilate preferentially NO 3 - , were positively related with production of butyric, malic, acetic, citric, formic, glycolic, pyruvic and oxalic acids, indicating that the N-source was apparently a key factor driving the bacteria-mediated mechanisms involved in the efficient RP dissolution.
Our results showed that only 15 strains out of 243 isolates were able to solubilize efficiently the RP with either NH 4 + or NO 3 - as N-source due to the complexity and low solubility of this P-source. Regardless the N-source, the greatest soluble-P content released from RP was detected for 87B, 24A, 59B, 23B, 41C, 32A and 39B isolates, as a consequence of the pH drop. However, 23B, 59B, 41C, 32A and 39B isolates appeared to be the most efficient in dissolving the RP. The strain 59B was the only one with abilities to solubilize the RP with either NH 4 + or NO 3 - . Gluconic acid, glycolic acid, lactic acid, acetic acid, formic acid and pyruvic acid were identified in almost all isolates with the highest values with NO 3 - rather than NH 4 + . This result evidenced the crucial role of NO 3 - in stimulating the production of organic acids, especially, butyric, malic, acetic, citric, formic, glycolic, pyruvic and oxalic acids secreted by 32A ( Paenibacillus massiliensis ), 41C ( Bacillus pumilus ) and 6C ( Paenibacillus sp .). The 86% of the isolates were identified as Gram positive bacteria belonging to Bacillus sp. and Paenibacillus sp. genera, indicating its suitability for agronomical purposes. Other factors different than N-source, but in a lesser extent, would be also involved in the secretion of organic acids such as soil CaCO 3 and N-NO 3 - content. Our results concluded that the effectiveness of the RP-solubilization would be directly associated with the organic acids production which secretion seemed to be driven by the assimilation of NO 3 - as N-source. Therefore, the N-source might be a key factor to take into consideration during the screening and selection of suitable strains involved in the P-solubilization.
S1 Table Total contents of elements present in the rock phosphate (RP). Data are given as means ± standard error (n = 3). (DOCX) Click here for additional data file. S2 Table The eluent gradient (100 mM KOH) for the program used to determinate the best anion separation. (DOCX) Click here for additional data file. S3 Table Names of the compounds and concentrations of solutions used to identify and quantify bacterial organic anion release by ion chromatography (HPIC). (DOCX) Click here for additional data file.
|
Perspectives of family medicine residents in Riyadh on leadership training: a cross-sectional study | 02002e06-b5bd-40fd-a73c-e441ff1b6229 | 10038693 | Family Medicine[mh] | As the recent coronavirus pandemic shows, health security is a critical aspect in developing countries and the world. Thus, many countries were eager to consider a substantial overhaul in the health care system. Saudi Arabia's policy has heavily emphasized primary care and called on family physicians to assume various tasks, including leadership . This is because ineffective leadership has been reported as the cause of Saudi Arabia's lack of a positive patient safety culture . Medical organizations in Saudi Arabia have further supported these efforts by including leadership concepts and skills training in medical education frameworks that guide curriculum development, such as the Saudi Board for Family Medicine 2020 (SaudiMED-FM 2020) curriculum, in addition to other postgraduate programs of the Saudi Commission for Health Specialties . The original Saudi Board for Family Medicine MED-FM curriculum was reviewed by the Saudi Commission for Health Specialties in 2019 to more accurately reflect what doctors are expected to do in today's health care system compared to the Saudi Board's previous edition . Medical educators have difficulty incorporating leadership into curricula while minimizing redundancy and guaranteeing value and relevance for all learners. Coaching is touted as a teaching strategy that can help students move beyond rote memorization of facts to develop the process skills they'll need in the future. In recent years, there has been very little published in family medicine (FM) literature about these tasks . Postgraduate and graduate leadership education programs have remained relatively limited, according to numerous studies published between 1991 and 2017, which showed that the leadership curriculum is diversified and limited in effectiveness . As for undergraduate leadership training, there has been a consensus that this training should begin in medical school, but not on what it should look like or how it should be taught or evaluated . The efficacy of an intervention depends in part on the target population's participation. Our study aims to assess the status of leadership training as perceived by family medicine residents in Riyadh to advise the development of a formal leadership training curriculum. This could address the lack of research on constructing formal leadership courses. The current project assessed how strongly residents associate family physicians with leadership, what domains of leadership residents desire more training in, and what opportunities residents identify to expand leadership training.
Setup, sampling, and process Between January and April 2022, we conducted an observational, quantitative, cross-sectional study with FM residents in Riyadh City, Saudi Arabia. A self-administered questionnaire, previously used in another study, was given to participants to complete . The questionnaire was handed out at different sites on different dates. The participants were asked to fill out the questionnaire only once to guarantee that no data was captured twice. King Saud University's College of Medicine Research Center in Riyadh, Saudi Arabia, gave its clearance for the study. There is a statement at the top saying, "Completion of this questionnaire will be treated as an indicator of your consent," as a way to get informed consent. The questionnaire consisted of three parts. The first part assessed the resident’s agreement with leadership ideals. The second part assessed the exposure to leadership domains, and the final part identified leadership training opportunities during their residency. The LEADS framework, an evidence-based, comprehensive framework for health care leaders, includes a questionnaire for testing the following leadership dimensions . The inclusion criteria were all FM residents in Riyadh while the exclusion criteria were FM residents in the training centers outside Riyadh and all other Residents in other specialties inside or outside Riyadh. Using a single-proportion sample-size calculation, n = z 2 p (1– p )/ d 2 , a 95% confidence level and a 5% margin of error were used to estimate the sample size. As a result, a sample size of 267 was required to estimate statistical significance. The questionnaire was distributed to all FM residents in all training centers during WADAs (Weekly Academic Day Activities) activity to give the same probability for each person to be included in the study. Also, we tried to reduce recall bias by testing the recall period during the pilot study and allowing participants to contact one of the authors at times as follow-up or for any queries. The survey underwent an extensive evaluation process in order to ensure its accuracy and relevance. In order to achieve this, a team of Five experts in the fields of Family Medicine and leadership were consulted. This team provided feedback on the language and content of the survey, and based on their suggestions, adjustments were made until a consensus was reached. To assess the feasibility and interpretability of the survey, a trial run was conducted with seven recently graduated FM residents. The survey was designed to be completed in under 15 min and to maintain anonymity, it was paper-based and coded. The leadership ideals of the participants were rated on a seven-point Likert scale, with 4 being considered neutral. Participants were also asked to identify their top three leadership opportunities. The results of the survey were analyzed using statistical methods, including counts, percentages, means, and standard deviations. In order to determine the reliability of the leadership ideals section of the survey, a number of measures were used. The internal consistency of this section was evaluated using the Cronbach's alpha coefficient, which equaled 0.78. Additionally, the corrected item-total correlations ranged from 0.28 to 0.69, further pointing towards the satisfactory reliability of this section of the survey. Finally, to determine any differences in leadership ideals across all resident responses and their perceived importance, a paired t test was used. The data analysis was performed using the Statistical Package for Social Studies (SPSS).
Between January and April 2022, we conducted an observational, quantitative, cross-sectional study with FM residents in Riyadh City, Saudi Arabia. A self-administered questionnaire, previously used in another study, was given to participants to complete . The questionnaire was handed out at different sites on different dates. The participants were asked to fill out the questionnaire only once to guarantee that no data was captured twice. King Saud University's College of Medicine Research Center in Riyadh, Saudi Arabia, gave its clearance for the study. There is a statement at the top saying, "Completion of this questionnaire will be treated as an indicator of your consent," as a way to get informed consent. The questionnaire consisted of three parts. The first part assessed the resident’s agreement with leadership ideals. The second part assessed the exposure to leadership domains, and the final part identified leadership training opportunities during their residency. The LEADS framework, an evidence-based, comprehensive framework for health care leaders, includes a questionnaire for testing the following leadership dimensions . The inclusion criteria were all FM residents in Riyadh while the exclusion criteria were FM residents in the training centers outside Riyadh and all other Residents in other specialties inside or outside Riyadh. Using a single-proportion sample-size calculation, n = z 2 p (1– p )/ d 2 , a 95% confidence level and a 5% margin of error were used to estimate the sample size. As a result, a sample size of 267 was required to estimate statistical significance. The questionnaire was distributed to all FM residents in all training centers during WADAs (Weekly Academic Day Activities) activity to give the same probability for each person to be included in the study. Also, we tried to reduce recall bias by testing the recall period during the pilot study and allowing participants to contact one of the authors at times as follow-up or for any queries. The survey underwent an extensive evaluation process in order to ensure its accuracy and relevance. In order to achieve this, a team of Five experts in the fields of Family Medicine and leadership were consulted. This team provided feedback on the language and content of the survey, and based on their suggestions, adjustments were made until a consensus was reached. To assess the feasibility and interpretability of the survey, a trial run was conducted with seven recently graduated FM residents. The survey was designed to be completed in under 15 min and to maintain anonymity, it was paper-based and coded. The leadership ideals of the participants were rated on a seven-point Likert scale, with 4 being considered neutral. Participants were also asked to identify their top three leadership opportunities. The results of the survey were analyzed using statistical methods, including counts, percentages, means, and standard deviations. In order to determine the reliability of the leadership ideals section of the survey, a number of measures were used. The internal consistency of this section was evaluated using the Cronbach's alpha coefficient, which equaled 0.78. Additionally, the corrected item-total correlations ranged from 0.28 to 0.69, further pointing towards the satisfactory reliability of this section of the survey. Finally, to determine any differences in leadership ideals across all resident responses and their perceived importance, a paired t test was used. The data analysis was performed using the Statistical Package for Social Studies (SPSS).
Overall, 270 residents participated in the study—134 of whom were female (49.6%). The mean age equaled 26.82, with a standard deviation of 2.63, a minimum of 18, and a maximum of 47. The participants were in their first (24.8%), second (26.3%), third (25.6%), and fourth (23.3%) year of residency. Most of the respondents did not have any leadership experiences longer than 20 h (77.0%) and did not participate in any leadership courses longer than 20 h (87.8%). The most common training places were Cluster 2 (21.1%) and King Khalid University Hospital (KKUH) (16.7%). In five out of six items regarding leadership ideals, affirmative answers were the most frequent, ranging from 70% of the agreement for 'I have had role models for effective leadership in the FM program' so far to 84.4% for 'Family physicians should take on leadership roles in their communities.' The only exception was a statement asking whether participants perceive themselves as leaders. In this case, most of the residents were neutral (62.6%); only 32.2% of the participants agreed, and 5.2% disagreed—the highest percentage for a negative answer regarding leadership ideals. Mean ratings of the agreement for each of the statements are in Fig. . A paired-samples t-test was conducted to compare the mean ratings of the statements that were rated the highest and the lowest, respectively: 'Family physicians should take on leadership roles in their communities' (M = 2.83, SD = 0.41) and 'I am a leader' (M = 2.27, SD = 0.55). A significant difference was found, t(269) = 15.03, p < 0.001. Subsequent analysis pertained to leadership domains. Table presents percentages of residents, divided by residency year, who voiced a desire for more training in certain leadership domains (i.e., they chose 'None and desired more' or 'Some but not enough' as their answer). Overall, most of the residents (50% or more) who participated in the study voiced a desire for more training in all leadership domains. Considering the highest percentages, the level of agreement between the leadership domains are conflict resolution (69%), teaching (67%), feedback (67%), and system transformation (67%). On the other hand, self-awareness was one of the least frequently chosen domains for first-year (56.7%) and third-year (49.3%) residents. Moreover, first-year (49.3%) and second-year (52.1%) participants expressed low demand for training in professionalism. Finally, effective communication training was desired by 57.7% of second-year and 52.4% of fourth-year respondents. Leadership opportunities were investigated next. As shown in Figs. and , over 50% of residents indicated that leadership electives or selected lectures, workshops, or seminars, as well as WADAs (Weekly Academic Day Activities), leadership mentors or coaches, teaching junior learners (with training), and leadership courses could be incorporated into the curriculum to foster leadership skills. Finally, less than 25% of participants referred to behavioral sciences, quality assessment projects or evidence-based medicine, online modules, resident retreats, academic projects, leadership portfolios, and leadership-specific components (with clinical and other evaluations).
In our study, we found that FM residents associate family physicians with leadership, desire more personal and system-level leadership training, and think that leadership training may be increased in the current curriculum and established in new areas. The significant difference ( p < 0.001) between the Likert scale scores for the highest-ranked leadership ideal ('Family physicians should take on leadership roles in their clinical settings') and the lowest-ranked ('I am a leader') implies that there is room for growth regarding residents' development as leaders. The results are generally higher than those of other studies that used the same scale . In resident education, current leadership curriculum guidelines emphasize the development of lower-level leadership skills and knowledge . The current curriculum has undoubtedly been significantly developed from the previous curriculum, but according to the study results, trainees still desire to learn more comprehensive leadership skills . These results show that trainees' need to improve their leadership skills was greater than in the previous study in Canada and similar previous studies conducted in many countries. This confirms trainees' need for more training in several leadership skills, giving the impression that a curriculum focusing more on leadership skills must be developed . Conflict resolution (69%), teaching (67%), feedback, and system transformation (67%) are among the more advanced concepts that residents want to learn more about, which was higher than the need for training in Canada . To account for statistical differences across domains, there was no curriculum focused specifically on leadership available at the time of this study. There was a preference for both experiential and didactic learning opportunities. Curriculum components requiring some deliverable documents, such as leadership portfolios, were considered less desirable. This may be due to the nature and complexity of the current portfolio, which caused the trainees not to prefer it as a means of developing their leadership skills. This calls for studies to evaluate its effectiveness and ways to develop it, considering trainees' experience, opinion of development, and satisfaction. Residents' level of agreement with several leadership ideals was unaffected by age, gender, or year of training. The desire for more training among third- and fourth-year residents is comparable to that of first- and second-year residents, which could be explained by a lack of substantial exposure to leadership domain training throughout their residency. Regarding leadership ideals, little variation existed between first- and fourth-year residents. A need may exist to provide residents with proper leadership training and competency, as there was no formal leadership training available when this poll was conducted. Most residents (76%) indicated that leadership electives could be incorporated into the curriculum to foster leadership skills, which was higher than the percentage in Canada . This may also be an opportunity to conduct studies to determine the importance of adding some new electives or rotations to the curriculum, the extent of the feasibility of some of the existing rotations and evaluate the possibility of modifying the rotations to be less lengthy and more numerous, especially after reducing the years of training in the new curriculum to three years. Also, most residents (65%) indicated that WADAs are a part of developing their leadership skills and a good indicator for confirming the trainees' belief in the importance of this day in developing their academic skills, allowing them to evaluate the possibility of developing this day to include other skills, including leadership skills. Teaching junior learners (with training) has been suggested by most of the trainees (59%) to develop their skills, which may be a chance to consider adopting it as part of the curriculum. Overall, the percentage of residents desiring more leadership training in Saudi Arabia is higher than in Canada in all leadership domains . Strengths and limitations A quantitative cross-sectional survey is the best way to get a wide range of views. In addition, only a modest number of people answered the survey. Only 30% of Riyadh's 900 residents responded to the questionnaire, which was sent to all the training centers in the city. Furthermore, due to the quantitative nature of this study, we were only able to collect a limited amount of information. One of the most pressing needs is to learn how residents view leadership and whether or not they believe family physicians should adopt a particular leadership style. Finally, there are no other local or regional studies to compare our findings to.
A quantitative cross-sectional survey is the best way to get a wide range of views. In addition, only a modest number of people answered the survey. Only 30% of Riyadh's 900 residents responded to the questionnaire, which was sent to all the training centers in the city. Furthermore, due to the quantitative nature of this study, we were only able to collect a limited amount of information. One of the most pressing needs is to learn how residents view leadership and whether or not they believe family physicians should adopt a particular leadership style. Finally, there are no other local or regional studies to compare our findings to.
There may be an opportunity to repeat this study from time to time to assess the extent of resident doctors' satisfaction with the development of the leadership curriculum. In addition to the possibility of applying the same study to programs of other specialties. There are also opportunities to conduct these same studies in all Kingdom programs.
Residents were enthusiastic about the idea of family physicians being leaders, which aligned with the current educational philosophy of requiring formal training. They also indicated areas where leadership training might be improved and developed in the current curriculum. The results of this poll could be used to help residents build leadership skills by incorporating them into a formal leadership curriculum.
Additional file 1.
|
A population-based nomogram to individualize treatment modality for pancreatic cancer patients underlying surgery | 7bbd155e-17aa-4f14-bec4-44dcb1ce579c | 10038997 | Internal Medicine[mh] | As one of the most aggressive and highly metastatic malignancies, pancreatic cancer ranks seventh among cancer-related deaths . Surgical resection with adjuvant/neoadjuvant therapy offers hope for long-term survival or cure in patients with non-metastatic pancreatic cancer . Local recurrence and metastasis of pancreatic cancer have brought new challenges to the treatment . However, the most effective treatment mode is uncertain, and an effective predictive model is needed to evaluate patient prognosis accurately. TNM staging focuses on the pathological tumor characteristics of patients eligible for surgery . However, even among patients with the same disease stage, the prognosis can vary widely . This population-based cancer staging method has great generality, but lacks guidance for different treatment modalities. The nomogram is a predictive model that uses a specific scoring system to assess the probability of survival of patients . Based on various combinations of serological markers, clinicopathological parameters, imaging features, genomic features or biomarkers, many predictive models have been established to predict the prognosis of pancreatic cancer at different stages – . Nevertheless, the data for the predictive model did not include patient treatment information. Therefore, there is an urgent need to develop an effective predictive model that integrates clinicopathology with therapy, especially in pancreatic cancer patients underlying surgery. In this study, post-match cohorts were created after propensity score matching (PSM) . Currently, gemcitabine-based combination therapy has improved survival in adjuvant therapy and in metastatic settings . The combination of multiple treatment options, such as FOLFIRINOX treatment and radiotherapy, brings new hope to patients. We discuss the impact of multidisciplinary treatment on patient survival and identify characteristics of patients who benefit from radiation therapy. In addition, a nomogram was constructed from the pre-competition cohort to segregate the population into different risk groups by predictors and then estimated the effect of radiotherapy on the upper group. Based on the evaluation of the patients’ prognosis, the most appropriate individualized treatment plan is provided for the patient through this nomogram.
Patient population Patient demographics, treatment information and tumor characteristics were extracted from SEER*Stat version 8.3.9. The inclusion criteria were as follows: (1) Patients with pancreatic cancer as the first and only cancer diagnosis. (2) The primary site of the tumor is limited to the International Classification of Diseases for Oncology, 3rd Edition (ICD-O-3) C25.0–C25.3 site codes, and histological codes are 8050–8089, 8500–8549, and 8140–8389 Oncological Disease Classification. (3) Patients are undergoing radical pancreatic cancer resection. (4)Patients with codes of 10–90 in RX Summ – Surg Prim Site (1998 +) and Site-specific surgery (1973–1997, with varying details by year and site) were classified into the postoperative group. Exclusion criteria were as follows: (1) Patients with second primary pancreatic cancer. (2) Patients with missing or incomplete treatment information, TNM stage (AJCC 7th Edition clinical stage), or other characteristics. (3) Patients with 0 months of follow-up (perioperative death). Study variables This study screened data from the SEER database on patients diagnosed with pancreatic cancer between 2010 and 2015, and selected patients who had undergone surgery. The variables we analyzed included age at diagnosis, sex, race, primary tumor site, tumor grade, radiotherapy, chemotherapy, surgery, marital status at diagnosis, insurance status, TNM stage (American Joint Commission on Cancer [AJCC] 7th version), pathological type, and survival information. The study endpoint was OS, the duration from diagnosis to death from any cause or last follow-up. Statistical analysis Pearson’s chi-square analysis was used to evaluate different clinical characteristics between different treatment regimens. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant. PSM was performed using the nearest neighbor matching method with a caliper of 0.001. 1:1 PSM was designed to reduce the selection bias of baseline variables between groups, including age, sex, race, marital, insurance, tumor site, histological type, grade, T, N, M stage, and treatment pattern twelve variables. After PSM, univariate and multivariate competing risks regression models were used to evaluate statistically significant variables for survival outcomes, and hazard ratios (HR) and corresponding 95% confidence intervals (CI) were calculated. The forest plot for OS Cox analyses shows the effect of different variables on survival outcomes and subgroup analysis results. In addition, Kaplan–Meier estimates were performed to show whether radiotherapy affects survival outcomes. Patients of 8026 were randomized 7:3 to a training cohort (n = 5618) and a validation cohort (n = 2408) using a random sampling function. Risk factors associated with survival were identified based on multivariate competing risk regression analysis and creation of nomograms. For the discrimination and calibration of nomograms in the training and validation cohorts, the C-index and area under the receiver operating characteristic curve (AUC) were calculated, and a calibration curve was drawn using a bootstrapping method involving 1000 resamples. Calibrations were performed at 1, 2, 3, and 5 years to compare predicted versus observed survival in pancreatic cancer patients underlying surgery. A standard curve was generated using the bootstrap method. The cohort was tested 1000 times for internal validation. For the calibration curve, the closer the curve is to the grey reference line, the closer the predicted value is to the actual situation. DCA is a method to assess the clinical utility of alternative models by quantifying the net utility of different threshold probabilities and applying them to a standard plot. Both references were patient protocol (representing higher clinical cost) and no protocol (representing no clinical benefit). In addition, the nomogram’s accuracy and clinical benefit rate were compared with the AJCC 7th edition staging system. A nomogram corresponding estimated the patient’s total score to the risk of survival for pancreatic cancer, and all patients were divided into high- and low-risk groups by the median risk score. Then, the survival analysis of patients with radiotherapy in the above groups was calculated separately. All statistical analyses were performed using R4.2.0 software (The R Project for Statistical Computing, http://www.r-project.org ). Two-sided P values were considered statistically significant if P < 0.05 . Ethics approval and consent to participate Our institutional ethics review board approved this study.
Patient demographics, treatment information and tumor characteristics were extracted from SEER*Stat version 8.3.9. The inclusion criteria were as follows: (1) Patients with pancreatic cancer as the first and only cancer diagnosis. (2) The primary site of the tumor is limited to the International Classification of Diseases for Oncology, 3rd Edition (ICD-O-3) C25.0–C25.3 site codes, and histological codes are 8050–8089, 8500–8549, and 8140–8389 Oncological Disease Classification. (3) Patients are undergoing radical pancreatic cancer resection. (4)Patients with codes of 10–90 in RX Summ – Surg Prim Site (1998 +) and Site-specific surgery (1973–1997, with varying details by year and site) were classified into the postoperative group. Exclusion criteria were as follows: (1) Patients with second primary pancreatic cancer. (2) Patients with missing or incomplete treatment information, TNM stage (AJCC 7th Edition clinical stage), or other characteristics. (3) Patients with 0 months of follow-up (perioperative death).
This study screened data from the SEER database on patients diagnosed with pancreatic cancer between 2010 and 2015, and selected patients who had undergone surgery. The variables we analyzed included age at diagnosis, sex, race, primary tumor site, tumor grade, radiotherapy, chemotherapy, surgery, marital status at diagnosis, insurance status, TNM stage (American Joint Commission on Cancer [AJCC] 7th version), pathological type, and survival information. The study endpoint was OS, the duration from diagnosis to death from any cause or last follow-up.
Pearson’s chi-square analysis was used to evaluate different clinical characteristics between different treatment regimens. All statistical tests were two-sided, and a P value < 0.05 was considered statistically significant. PSM was performed using the nearest neighbor matching method with a caliper of 0.001. 1:1 PSM was designed to reduce the selection bias of baseline variables between groups, including age, sex, race, marital, insurance, tumor site, histological type, grade, T, N, M stage, and treatment pattern twelve variables. After PSM, univariate and multivariate competing risks regression models were used to evaluate statistically significant variables for survival outcomes, and hazard ratios (HR) and corresponding 95% confidence intervals (CI) were calculated. The forest plot for OS Cox analyses shows the effect of different variables on survival outcomes and subgroup analysis results. In addition, Kaplan–Meier estimates were performed to show whether radiotherapy affects survival outcomes. Patients of 8026 were randomized 7:3 to a training cohort (n = 5618) and a validation cohort (n = 2408) using a random sampling function. Risk factors associated with survival were identified based on multivariate competing risk regression analysis and creation of nomograms. For the discrimination and calibration of nomograms in the training and validation cohorts, the C-index and area under the receiver operating characteristic curve (AUC) were calculated, and a calibration curve was drawn using a bootstrapping method involving 1000 resamples. Calibrations were performed at 1, 2, 3, and 5 years to compare predicted versus observed survival in pancreatic cancer patients underlying surgery. A standard curve was generated using the bootstrap method. The cohort was tested 1000 times for internal validation. For the calibration curve, the closer the curve is to the grey reference line, the closer the predicted value is to the actual situation. DCA is a method to assess the clinical utility of alternative models by quantifying the net utility of different threshold probabilities and applying them to a standard plot. Both references were patient protocol (representing higher clinical cost) and no protocol (representing no clinical benefit). In addition, the nomogram’s accuracy and clinical benefit rate were compared with the AJCC 7th edition staging system. A nomogram corresponding estimated the patient’s total score to the risk of survival for pancreatic cancer, and all patients were divided into high- and low-risk groups by the median risk score. Then, the survival analysis of patients with radiotherapy in the above groups was calculated separately. All statistical analyses were performed using R4.2.0 software (The R Project for Statistical Computing, http://www.r-project.org ). Two-sided P values were considered statistically significant if P < 0.05 .
Our institutional ethics review board approved this study.
Characteristics of patients and disease As Fig. shows, a total of 8026 pancreatic cancer patients were identified from the SEER database. After performing PSM, a total of 4076 pts were eliminated at match conditions of ratio (1:1) and caliper value (0.001). A total of 3950 cases, including 1975 cases with radiotherapy and 1975 cases without radiotherapy, were finally chosen for further research. The patients’ characteristics before and after matching are shown in Table . After PSM, the distributions of all variables were similar and comparable. A similar distribution was also discovered in the no- radiotherapy group of the postmatch cohort. Survival analysis Using the Kaplan–Meier survival curve, the OS of the nonradiotherapy group and the radiotherapy group before and after PSM was demonstrated by the log-rank test. Especially after PSM, OS showed that patients with radiation therapy had a better prognosis than those without radiation. The median OS was 24 months (95% CI 0.460–0.506) and 26 months (95% CI 0.467–0.513) for the non-RT group and RT group, respectively (Figure A–B). In addition, survival analysis results showed that patients with Grade II, Grade III, N1, adenocarcinoma, ductal and lobular neoplasms, T3, and T4 benefited more from adjuvant radiotherapy (Figure C–F). Independent prognostic factors in the postmatch cohort We performed univariate Cox proportional hazards regression analysis for fourteen potential factors and identified seven independent risk factors. Variables included T stage, N stage, tumor grade, histological subtype, stage, radiotherapy and chemotherapy. The multivariate Cox proportional hazards regression analysis results are clearly demonstrated in the forest plot in Fig. . In addition, we also made a forest plot to show the effect of radiotherapy in different subgroups of patients, but the results did not show that patients with the above characteristics were more likely to benefit from radiotherapy to improve their survival prognosis (Figure ). Prognostic nomogram model establishment and validation in the prematch cohort As Table shows, in the training set, the univariate analysis showed that, marital status,age, tumor site, grade, TNM stage, stage, histological type, chemotherapy and radiotherapy were significantly correlated with pancreatic cancer ( P < 0.05). Then, the multivariate Cox proportional hazards regression analysis indicated factors such as histological type, grade, T stage, N stage, stage, chemotherapy, and radiotherapy. Prognostic nomogram model establishment and validation A nomogram was established based on independent predictors obtained by multivariate Cox proportional hazards regression analysis to display the 1‐year, 2‐year, 3‐year and 5‐year prognoses for pancreatic cancer after surgery (Fig. ). The nomogram is provided through a free browser-based online calculator, and we have created a web version of the nomogram for ease of use, available at https://rockeric.shinyapps.io/DynNomapp/ . We can predict patient survival prognostic risk by inputting the patient’s pathological variables (Figure ). The bias-corrected C-index values in the training and validation cohorts were 0.721 (95% CI: 0.711–0.731) and 0.706 (95% CI: 0.690–0.722), respectively, indicating the moderate discrimination ability of the nomogram. Nomogram performance was quantified in terms of differentiation and calibration. The predictive accuracy of the nomogram was assessed by discrimination and calibration. The calibration curves of 1‐year, 2‐year, 3‐year, and 5‐year based on the training and validation cohorts are shown in Fig. (A–D). The calibration curves appeared to be very close to the ideal curve, representing good agreements between the nomogram‐predicted and the actual 1‐, 2‐, 3‐, and 5‐year prognoses for pancreatic cancer after surgery. The nomogram has good predictive accuracy and dependability in predicting pancreatic cancer after surgery. Similarly, the AUC of this study in the training cohorts and validation cohorts was also significantly higher than that of the TNM staging system (Fig. A–H). In the training cohort, DCA showed that the nomogram model is effective in clinical practice, with better clinical benefits than traditional TNM staging (Fig. A–D). Risk stratification and radiotherapy efficiency in different groups The different variable risk scores were obtained from the nomogram, and all scores for each patient were summed to calculate the total score. The patients were dichotomized into high- and low-risk groups according to the risk score of the nomogram, and the results showed that the optimal cutoff value was 122.1 by X-tile software (Figure A–B). This study further compared whether patients in different risk groups could benefit from radiotherapy. The results showed no statistically significant difference in whether patients in the low-risk group benefited from radiotherapy (HR = 1.14; 95% CI: 0.97–1.35; p = 0.17; Fig. A), while patients in the high-risk groups benefited from radiotherapy (HR = 0.78; 95% CI: 0.71–0.85; p < 0.0001; Fig. B).
As Fig. shows, a total of 8026 pancreatic cancer patients were identified from the SEER database. After performing PSM, a total of 4076 pts were eliminated at match conditions of ratio (1:1) and caliper value (0.001). A total of 3950 cases, including 1975 cases with radiotherapy and 1975 cases without radiotherapy, were finally chosen for further research. The patients’ characteristics before and after matching are shown in Table . After PSM, the distributions of all variables were similar and comparable. A similar distribution was also discovered in the no- radiotherapy group of the postmatch cohort.
Using the Kaplan–Meier survival curve, the OS of the nonradiotherapy group and the radiotherapy group before and after PSM was demonstrated by the log-rank test. Especially after PSM, OS showed that patients with radiation therapy had a better prognosis than those without radiation. The median OS was 24 months (95% CI 0.460–0.506) and 26 months (95% CI 0.467–0.513) for the non-RT group and RT group, respectively (Figure A–B). In addition, survival analysis results showed that patients with Grade II, Grade III, N1, adenocarcinoma, ductal and lobular neoplasms, T3, and T4 benefited more from adjuvant radiotherapy (Figure C–F).
We performed univariate Cox proportional hazards regression analysis for fourteen potential factors and identified seven independent risk factors. Variables included T stage, N stage, tumor grade, histological subtype, stage, radiotherapy and chemotherapy. The multivariate Cox proportional hazards regression analysis results are clearly demonstrated in the forest plot in Fig. . In addition, we also made a forest plot to show the effect of radiotherapy in different subgroups of patients, but the results did not show that patients with the above characteristics were more likely to benefit from radiotherapy to improve their survival prognosis (Figure ).
As Table shows, in the training set, the univariate analysis showed that, marital status,age, tumor site, grade, TNM stage, stage, histological type, chemotherapy and radiotherapy were significantly correlated with pancreatic cancer ( P < 0.05). Then, the multivariate Cox proportional hazards regression analysis indicated factors such as histological type, grade, T stage, N stage, stage, chemotherapy, and radiotherapy.
A nomogram was established based on independent predictors obtained by multivariate Cox proportional hazards regression analysis to display the 1‐year, 2‐year, 3‐year and 5‐year prognoses for pancreatic cancer after surgery (Fig. ). The nomogram is provided through a free browser-based online calculator, and we have created a web version of the nomogram for ease of use, available at https://rockeric.shinyapps.io/DynNomapp/ . We can predict patient survival prognostic risk by inputting the patient’s pathological variables (Figure ). The bias-corrected C-index values in the training and validation cohorts were 0.721 (95% CI: 0.711–0.731) and 0.706 (95% CI: 0.690–0.722), respectively, indicating the moderate discrimination ability of the nomogram. Nomogram performance was quantified in terms of differentiation and calibration. The predictive accuracy of the nomogram was assessed by discrimination and calibration. The calibration curves of 1‐year, 2‐year, 3‐year, and 5‐year based on the training and validation cohorts are shown in Fig. (A–D). The calibration curves appeared to be very close to the ideal curve, representing good agreements between the nomogram‐predicted and the actual 1‐, 2‐, 3‐, and 5‐year prognoses for pancreatic cancer after surgery. The nomogram has good predictive accuracy and dependability in predicting pancreatic cancer after surgery. Similarly, the AUC of this study in the training cohorts and validation cohorts was also significantly higher than that of the TNM staging system (Fig. A–H). In the training cohort, DCA showed that the nomogram model is effective in clinical practice, with better clinical benefits than traditional TNM staging (Fig. A–D).
The different variable risk scores were obtained from the nomogram, and all scores for each patient were summed to calculate the total score. The patients were dichotomized into high- and low-risk groups according to the risk score of the nomogram, and the results showed that the optimal cutoff value was 122.1 by X-tile software (Figure A–B). This study further compared whether patients in different risk groups could benefit from radiotherapy. The results showed no statistically significant difference in whether patients in the low-risk group benefited from radiotherapy (HR = 1.14; 95% CI: 0.97–1.35; p = 0.17; Fig. A), while patients in the high-risk groups benefited from radiotherapy (HR = 0.78; 95% CI: 0.71–0.85; p < 0.0001; Fig. B).
Due to the poor prognosis of pancreatic cancer, more than 90% of pancreatic cancer surgery patients will experience regional cancer recurrence, distant recurrence, or metastasis , . Multidisciplinary therapy has been developed to include some combination of systemic chemotherapy, locoregional radiation, and surgery in selected patients with pancreatic cancer. CRT with fluorouracil (FU) is considered the standard of care based on the Gastrointestinal Tumor Study Group (GITSG) trial and extensive case series analysis from Johns Hopkins and the Mayo Clinic , . However, conflicting results were obtained by the European Agency for Research and Cancer (EORTC) , and the European Pancreatic Cancer Study (ESPAC-1) . Both clinical trials showed a survival benefit with FU or gemcitabine, respectively, so chemotherapy alone as adjuvant therapy is advocated as standard in Europe. Currently, combination chemotherapy with FOLFIRINOX (leucovorin, 5-fluorouracil, irinotecan, and oxaliplatin) appears to be considered the most effective regimen for patients with pancreatic cancer. In addition, the MPACT study demonstrated improvement of overall survival in nearly one year by using nab-paclitaxel plus gemcitabine versus gemcitabine alone . These studies have clinical and therapeutic implications. Radiotherapy plays an important role in controlling the local progression of tumors. With the advancement of modern precision radiotherapy technology, precision radiotherapy techniques such as intensity-modulated radiotherapy (IMRT), stereotactic body radiotherapy (SBRT) and proton beam therapy (PBT) have been implemented – . These novel treatment methods can effectively improve the dose distribution in the target area of the radiotherapy plan, reduce the damage to normal tissues around the tumor, and achieve better local control of the tumor. The reasons why the local control benefit translates into a survival benefit may be multifactorial. The value of radiotherapy in addition to surgery for pancreatic cancer deserves further exploration. Insufficient evidence has not been found to determine the best combination of treatment options for individual patients. Therefore, there is a need to develop clinical prognostic models that can predict patient outcomes to potentially aid in developing comprehensive and patient-centered treatment regimens. According to the study, we performed a PSM analysis to minimize the effect of between-group differences. After PSM, patients who received radiation therapy had better outcomes than those who did not. These data suggest that the survival benefit of RT can be complex, especially with modern RT techniques and various doses. For pancreatic cancer patients who require surgery, individualized treatment plans and accurate assessment of patient survival and prognosis are particularly important. Therefore, we developed a reliable system to accurately predict patient survival time, taking into account multiple prognostic factors. We identified and modeled independent prognostic factors associated with survival by COX analysis. Our study shows that T stage, N stage, grade, stage and histological type are important risk factors. Radiation and chemotherapy are vital protective factors. Previous studies have attempted to model the prognosis of pancreatic cancer patients based on clinical features – . It has been evidenced that higher T and N stage, poor differentiation, stage and ductal adenocarcinoma are related to the worse prognosis. TNM staging evaluates patient prognosis by tumor size, lymph node metastasis, and distant metastasis . However, disease progression is much more complex than the classical form. It is worth noting that, overall, the contribution of T or N staging to the survival of surgical patients mainly was no more significant than that of differentiation .Our study also produced the same result. This means survival varies widely among pancreatic cancer surgery patients, even for the same TNM stage. Among other factors, the biological behavior of pancreatic cancer may be related to histological subtypes, particularly differences in survival rates for different histological types . Ductal adenocarcinoma is a factor in poor survival in patients diagnosed with pancreatic cancer compared to other histological types . The response of different histological exocrine pancreatic subtypes to tumor chemotherapy and radiotherapy remains unclear. In addition, clinical studies have shown that changes in serological indicators CA19-9, CEA, and CA125 are related to tumor progression and prognosis before and after surgery , . Vascular invasion, surgical margin status and lymph node positive rate are also important clinical indicators to predict the prognosis of patients , .Obviously, these need further research for clinical verification, so as to be better applied to clinical practice. Unlike other studies, we not only developed and validated a prognostic nomogram, but further assessed the survival benefit of radiotherapy through risk stratification. In this study, the bias-corrected C-index values in the training and validation cohorts were 0.721 (95% CI: 0.711–0.731) and 0.706 (95% CI: 0.690–0.722), respectively, indicating good discriminative power of the nomogram. Baseline nomograms using available clinicopathological variables showed good performance in distinguishing patient outcomes at 1, 2, 3, and 5 years. Based on the nomogram, a risk stratification system was formed, and all patients were clearly divided into two risk prognostic groups for further rational evaluation of comprehensive treatment. Furthermore, based on the nomogram, a risk stratification system was formed by which all patients were unambiguously divided into two risk prognostic groups. In addition, the model also screens patients for further rational comprehensive treatment through risk stratification. Some benefits of radiation therapy may be offset by side effects of diffuse radiation to nearby vital organs. This means that a subgroup should be defined to differentiate those who would benefit from radiotherapy. Through risk stratification analysis, we found that radiotherapy was beneficial to improve OS in high-risk group patients, but not in low-risk patients. Although the SEER database does not provide specific radiation information (irradiation extent and radiation dose), OS is sufficiently reliable and effective to assess the therapeutic effect of RT and could reflect the clinical effect of radiotherapy. In our risk stratification model constructed with a nomogram, the high-risk group was characterized by positive lymph nodes, moderately differentiated, poorly differentiated, adenocarcinoma, T3, and T4, and benefited from radiotherapy. The more aggressive cancer features of the high-risk group may result in residual tumor cells, but a better survival benefit with radiotherapy compared with the low-risk group. The reason may be that patients in the high-risk group achieved absolute local tumor control after radiotherapy, which translated to significant survival benefit. Therefore, for high-risk patients, radiotherapy techniques with low side effects should be used on the premise of ensuring the tolerance and effectiveness of treatment. With this nomogram, physicians can evaluate the survival prognosis of pancreatic cancer patients and preliminarily select appropriate treatment options. In addition, we have developed an easy-to-use web calculator to make it easily accessible to the general public. This study provides a new perspective for individualized treatment of pancreatic cancer patients undergoing surgery. While acknowledging some limitations, the study has several strengths. The advantages of this study are: large population base, many variables included, PSM was used to adjust confounding factors. More importantly, the study described the population-level survival benefit of patients treated with radiotherapy by OS, and advanced a pair of models for stratified analysis, with good discrimination and internal validation. Furthermore, in this retrospective study, unknown factors cannot be ruled out to have influenced the results. Therefore, more caution should be exercised when using predictive models to decide patient treatment options. With the emergence of new chemotherapy regimens and radiotherapy techniques, more factors will be included to improve the model, and prospective trials are needed to verify it. Finally, there are limitations to the study. First, although PSM addressed treatment selection bias, there is still some potential for bias in this retrospective cohort study. Second, our analysis shared the limitations of the SEER datasets. The datasets are limited by the quality and availability of the original data. The duration and regimens of chemotherapy used are not available from SEER. Radiation information, such as technology, dose and sequence, are not provided. Finally, this is a retrospective study. Third, considering the sample size, the number of covariates included in the COX regression model is large, and some of them show collinearity. The predictive model should be validated in a large, multicenter pancreatic cancer clinical trials.
In summary, it has higher accuracy, good clinical utility, and more precise prognostic prediction than traditional staging systems. Our nomogram can be used to predict survival and to assess the survival benefit of radiation therapy in different patients. Radiation therapy should be evaluated in conjunction with existing systemic chemotherapy regimens to facilitate individualized treatment.
Supplementary Figure S1. Supplementary Figure S2. Supplementary Figure S3. Supplementary Figure S4. Supplementary Table S1. Supplementary Table S2.
|
Sleep fMRI with simultaneous electrophysiology at 9.4 T in male mice | 98396461-8ba7-411b-900c-3cf4fb89775e | 10039056 | Physiology[mh] | Sleep is generally considered a tightly regulated whole-brain phenomenon, and its role for our cognition and heath is, from our own daily experience, of great significance. In recent years, studies have identified neural circuits regulating transition and stability of awake-sleep cycle , . The ascending arousal system , including monoaminergic, cholinergic, and peptidergic systems, are involved in promoting or sustaining wakefulness. Genetically defined cell populations in the preoptic area (POA), basal forebrain (BF), brainstem, and cortex, have been identified as non-rapid eye movement (NREM) promoting cells . And neurons in pedunculopontine tegmentum (PPT) and laterodorsal tegmentum (LDT) are involved in rapid eye movement (REM) sleep generation . In addition, the sleep/wake states are usually accompanied by distinct neural events which result from the synchronous activities of neural circuits, for example, spindles or slow waves in NREM sleep , and sharp wave ripples (SWRs) in both quiet wakefulness and NREM sleep. Previous studies have shown that these events play important roles in sleep architecture, synaptic plasticity and memory consolidation , , among many others. Circuitry level approaches in animals have provided detailed understanding of neural mechanisms during awake-sleep cycle, but techniques providing macroscopic view are also needed for systematical examination of sleep, a fundamentally global phenomenon. Non-invasive brain mapping tools such as functional magnetic resonance imaging (fMRI), electroencephalogram (EEG), positron emission computed tomography (PET) have provided the whole brain insights into sleep. Previous PET studies showed that the descent from wakefulness to NREM sleep was accompanied by global or regional reductions in cerebral blood flow (CBF), oxygen metabolism and glucose metabolism , . Blood-oxygen-level-dependent (BOLD) fMRI studies also demonstrated widespread hemodynamic alterations during awake-NREM cycle . Coherent patterns of slow and large-amplitude oscillating electrophysiological, hemodynamic, and cerebral-spinal fluid (CSF) dynamics were found during NREM sleep , and BOLD signals further exhibited distinct frequency patterns from wakefulness to NREM sleep . Furthermore, the progression from wakefulness to NREM sleep was accompanied by the gradual breakdown of inter-regional fMRI based functional connectivity and arousal dependent Hidden Markov Model (HMM) states . In addition, the brain state dynamics, as broadly defined, has been widely investigated and implicated in cognitive processes and brain disorders – . More broadly, fMRI based resting-state functional connectivity has been increasingly recognized to be influenced by arousal fluctuations , as human subjects often exhibit notable arousal fluctuations or even sleep during resting-state fMRI. However, the BOLD signal is a hemodynamic signal and thus is also influenced by non-neural physiological factors (e.g., respiratory and cardiac signals) that may co-vary with arousal state changes . Limited by the available neural recording and manipulation tools in humans, an animal sleep fMRI method is much needed as a platform to thoroughly disentangle arousal related neural and non-neural contributions to BOLD signals. Clearly the macroscopic observations using non-invasive imaging techniques in humans and microscopic, circuitry level knowledge in animals are difficult to integrate, due to the vast gaps of spatiotemporal scales and species. Simultaneous electrophysiology and fMRI could potentially bridge the gap. However, most simultaneous electrophysiology and fMRI sleep studies were scalp EEG-fMRI in humans , which provided limited electrophysiological information due to the intrinsic limitation of scalp EEG. Meanwhile, invasive simultaneous electrophysiology and fMRI have been developed mostly in anesthetized animals and not yet utilized in sleep research so far. Therefore, a sleep fMRI method based on simultaneously acquired invasive electrophysiology in mice could provide both local and global information and bridge the spatiotemporal and species gaps. Two major technical obstacles need to be overcome to achieve mouse sleep fMRI. First, it is difficult to perform un-anesthetized mouse fMRI because of the high stress level, due to the head restraining and loud noise of the fMRI environment, led alone to make mouse fall asleep. Second, it is well known that simultaneously electrophysiology and fMRI is highly challenging due to the mutual electromagnetic interference between electrophysiology and fMRI, which is more so in un-anesthetized animals. To this end, we developed a mouse sleep fMRI method based on simultaneously electrophysiological recording at 9.4 T. To achieve this goal, we first established a highly MR-compatible electrophysiological recording setup in un-anesthetized mice at 9.4 T. Through further optimization on awake mouse fMRI, we achieved the recording of whole awake-sleep cycle from awake to NREM and REM states during fMRI. Importantly, using MR-compatible graphene fiber (GF) electrodes, hippocampal local field potential (LFP) signal were recorded and characteristic events such as spindle and SWR were extracted during sleep fMRI. With this mouse sleep fMRI method, we revealed global patterns of NREM and REM sleep, and more importantly, the global, asymmetric and sequential dynamics of state transitions, which was also evident in their trajectories in the low-dimensional manifold. Furthermore, utilizing long short-term memory (LSTM) recurrent neural network (RNN) modeling, we found that BOLD signals could predict state transitions, up to 17.8 s, prior to electrophysiological defined transition time point. Using the neural-event-triggered (NET) fMRI approach, we found SWRs had a significantly higher BOLD responses in NREM state than in AW state, which could attribute to the co-occurrence of spindle events. In conclusion, this mouse sleep fMRI method will further advance mouse sleep research by providing both local and global view. Furthermore, it provides an accessible platform for elucidating the neural basis of arousal related BOLD signals. Combined with rich resources and tools in mice, the current method will help to establish a general multiscale framework of sleep. Sleep fMRI using MR-compatible electrophysiology recording in un-anesthetized mice To investigate how arousal fluctuation contributes to the global dynamics, we established the mouse sleep fMRI setup using the simultaneous electrophysiology and fMRI. MR-compatible electrocorticography (ECoG) and depth electrodes were custom designed and fabricated for high MR-compatibility at 9.4 T, while maintaining good electrophysiological recording quality (Fig. ). Double-sided flexible printed circuit (FPC) with polyimide film as base substrate was used to minimize the thickness and width, and to maximize the flexibility of the array. Copper layer was used for its similar magnetic susceptibility to the brain (thus minimal MRI artefact), and gold was plated over copper layer in each recording site to avoid biological toxicity of copper. The 16-Channels ECoG electrodes covered a large portion of the cerebral cortex, including retrosplenial area (RSP), motor area, somatosensory area and posterior parietal association area (Fig. and Supplementary Data ). To record subcortical activity, we further fabricated a graphene fiber based depth electrodes. Graphene fiber was shown to be highly MR-compatible in our previous deep brain stimulation (DBS) study at 9.4 T, and was repurposed for LFP recording in the hippocampus (Fig. ). The pipeline of electrode implantation surgery, habituation training and data acquisition was shown in Fig. and described in Method. Both electrodes produced minimal MRI artifacts in T2 weighted anatomical images and T2* weighted functional images (Fig. ) of the relatively small sized mouse brain at 9.4 T. With highly MR-compatible electrodes providing the feasibility of monitoring animal’s arousal states, it remained challenging to record mouse sleep in the MRI environment. Based on our extensive previous experiences on awake mouse fMRI , and systematical optimization of stress level reduction , the current un-anesthetized mouse fMRI setup allowed simultaneous electrophysiological recording while minimizing animal’s stress and thus facilitating sleep (Fig. ). In particular, the mouse’s head was tilted for 30 degree and its forelimbs were allowed to move freely, both specifically designed for minimizing stress level . It is well known that the simultaneously acquired electrophysiological signal suffers from severe MRI artifacts. Thus, we established and evaluated an off-line de-noising preprocessing pipeline (Supplementary Fig. ), which greatly suppressed MRI artifacts to the extent that they no longer affected further analysis (Fig. ). Meanwhile, we also developed an fMRI preprocessing pipeline that was optimized for 4-hour sleep fMRI data. Raw fMRI data exhibited good (temporal) signal-noise-ratio (Fig. ) with small head motion and the “6 rp + 6 Δrp + 40 PCs” regression approach was modified from our previous studies , to minimize the effects of scanner drift, motion and other non-neural physiological noises (Fig. and Supplementary Fig. ). The 40 PCs were derived from fMRI signals outside the mouse brain (Supplementary Fig. ), largely capturing the non-neuronal signals , such as head motion (Supplementary Fig. ), physiological effects (Supplementary Fig. ) and infra-slow drift (Supplementary Fig. ). Thus, using the “6 rp + 6 Δrp + 40 PCs” regression approach, arousal induced non-neuronal nuisance effects on fMRI signals were minimized. With this simultaneous electrophysiology and fMRI recording setup, a representative session of mouse sleep fMRI with awake (AW), NREM and REM sleep states was shown in Fig. with denoised ECoG, electromyography (EMG) and fMRI signals. Sleep states were conventionally classified using electrophysiological signals (see Methods for details, Supplementary Fig. ). Similar to previous optical imaging studies of mouse sleep , we utilized a 4-hour long fMRI acquisition scheme. Mice exhibited notable and increasing portion of NREM sleep as the scan progressed, and REM sleep was also observed during the 4-hour scan (Fig. ). In total, 46 4-hour sessions (184 h) simultaneous ECoG/LFP-fMRI recordings from 24 mice (27 ECoG-fMRI sessions from 14 mice and 19 LFP-fMRI sessions from 10 mice) were acquired in our dataset, including 3588 min of NREM sleep and 342 min of REM sleep (Fig. ). The above dataset with both raw and preprocessed data is openly accessible at 10.12412/BSDC.1668502646.20001. State dependent whole brain BOLD patterns and their neurophysiological correlates Human sleep is characterized by long state durations, e.g., each NREM-REM cycle is approximately 90 min , which prevents conventional general linear model (GLM) analysis due to its low frequency beyond the detection limit of fMRI. However, mouse sleep is fragmented with short state durations (Supplementary Fig. , median duration of AW: 46.0 s, NREM: 41.0 s and REM: 121.5 s in our data), thus is convenient for analyzing state dependent activations. Using the conventional GLM analysis, we mapped the brain-wide BOLD activations of NREM and REM sleep, relative to the AW state (Fig. ). For the NREM state, a large part of cerebral cortex and hippocampus were activated, while subcortical regions, such as thalamus and part of midbrain, were deactivated compared to the AW state (Fig. ). Importantly, using the simultaneously acquired ECoG or LFP signals, we explored the potential relationship between sleep state dependent BOLD and electrophysiological signals. We calculated hemodynamic response function (HRF) -convolved ECoG/LFP band-limited power and its correlation with BOLD changes (Supplementary Fig. ). The relative changes of ECoG or hippocampal LFP were in generally consistent with current understanding of NREM and REM sleep , (Fig. , left panels). Importantly, we found a significant broadband (1–100 Hz) correlation between the relative ECoG power and relative cortical BOLD signal changes (Fig. , right panel) during NREM state, which was also evident in hippocampal CA1 recordings (Fig. , right panel). For REM state, significant correlation between theta band-limited power and BOLD changes was observed (Fig. , right panels), suggesting the potential relationship of theta band and BOLD changes. Low dimensional dynamics within and between brain states The above NREM and REM activation patterns were state-dependent features. However, previous studies indicated that sleep architectures are dynamic . Taking advantages of whole-brain mouse sleep fMRI data, we investigated the low dimensional dynamics within and between brain states. First, we conducted group principal component analysis (PCA) to BOLD signals for dimensional reduction. The −100th PC accounted for at least 0.3% explained variances and accumulatively the first 100 PCs resolved greater than 78% explained variances of BOLD signals. (Fig. , Supplementary Fig. and Supplementary Data ). The first four PCs were shown in Fig. . Each PC exhibited non-stationary temporal weights from the start to the end of each state (Fig. ), suggesting dynamic involvements of different functional networks within brain states. Further dynamics were expected across brain states, thus we characterized the state transition probabilities (Fig. ). Electrophysiological and whole brain BOLD features were shown in Fig. (upper panel) and Supplementary Fig. – for four state transitions, respectively. For each state transition, we observed sequential BOLD signal fluctuations traversing the mouse brain (Supplementary Fig. and Supplementary Data ). Further global spatiotemporal patterns of all four state transitions were shown in Supplementary Fig. – . Importantly, the temporal weights of PCs (tPCs) also exhibited diverse characteristics across state transitions (Fig. , lower panel, and Supplementary Fig. ), suggesting dynamic involvements of different functional networks across brain states. Moreover, in the low dimensional spaces spanned by BOLD PCs (Fig. and Supplementary Fig. ) and electrophysiological band-limited power ratios (Fig. ), activity flows of the manifolds exhibited dynamic properties within and across brain states. Interestingly, we found the separated trajectories between “AW to NREM” and “NREM to AW” transitions (Fig. ). And we further quantified such phenomenon and the significant asymmetry were observed around state transitions in both BOLD and electrophysiological space (Fig. ), suggesting “AW to NREM” and “NREM to AW” transitions were asymmetric processes. Predictions of state transitions using the LSTM RNN model The dynamic characteristics of PCs across state transitions promoted us to investigate whether such transitions could be predicted by BOLD signals prior to electrophysiology defined transition time, and if so, which functional networks contributed to such predictions. Thus, we established a LSTM RNN model for state transition prediction, with BOLD signals (tPCs) preceding state transitions as model input and the brain state after such transition as model output (Fig. ). We systematically evaluated the model parameters, and the resulting optimized parameter set was 1 hidden LSTM layer with 50 hidden units (Fig. ). Prediction accuracy on the validation dataset was primarily related to the gap times preceding state transitions (Fig. ). Using the optimized parameter set, we found high prediction accuracy (Fig. ) of more than 85% across all seven prediction categories. The high prediction accuracy gradually decreased as the gap time increased (Fig. , upper panel), and the mean discriminate times were 10.55 s for “AW to NREM”, 3.28 s for “NREM to AW”, 17.84 s for “NREM to REM”, and 3.45 s for “REM to AW” state transitions (Fig. , middle panel). The null dataset was constructed by shuffling prediction categories. Correspondingly, sensitivity results of the LSTM RNN model revealed several regions significantly contributed to the prediction of state transitions (Fig. , lower panel, and Supplementary Fig. ), including (1) ventral thalamus (vThal), medial mammillary nucleus (MM), RSP, ventral tegmental area (VTA), median raphe nucleus (MnR), globus pallidus (external segment, GPe) and interposed cerebellar nucleus (Int) for “AW to NREM” state; (2) anterior cingulate area (ACA), agranular insular area (AI), visual areas (VIS), pontine reticular nucleus (ventral part, PnV), hippocampal formation (HPF) and primary somatosensory area (SSp) for “NREM to AW” state; (3) paragigantocellular reticular nucleus (PGRN), periaqueductal gray (PAG), HPF, VIS, medial preoptic area (MPO), supplemental somatosensory areas (SSs) and amygdala areas (AMY) for “NREM to REM” state; and (4) MM and pontine central gray (PCG) for “REM to AW” state. Interestingly, we found HPF exhibited significant contributions on the prediction of state transitions from NREM to AW and REM states. Utilizing the GF electrode placed in hippocampus CA1 region of our LFP-fMRI dataset, we compared the LFP power spectrum between “NREM to AW” and “NREM only” states, and found significantly higher power in three frequency bands, including (1) 7-37 Hz from −14 to −2 s, (2) 44-95 Hz from −12 to −1 s, and (3) 119-205 Hz from −11 to −3 s before state transitions (Fig. ). Also, we conducted same analysis between “NREM to REM” and “NREM only” states, and found significantly higher power in two frequency bands, including (1) 1-34 Hz from (beyond) −30 to −3 s, and (2) 128-141 Hz from (beyond) −30 to −13 s before state transitions (Fig. ). These “NREM only” epochs were randomly selected from the sleep fMRI dataset, and epochs of “NREM to AW”, “NREM to REM” and “NREM only” with short durations (<60 s) were excluded. Then, two sample t-test was conducted with threshold of p < 0.05. Furthermore, we isolated SWR and spindle events, among other events, based on LFP dataset (Fig. ). Altogether, SWR and spindle events showed a pronounced decrease of their occurrence probability before “NREM to AW (or REM)” state transition (Fig. ). To be note, the preceding time of SWR event changes (Fig. ) was similar to the mean discriminate time of “NREM to AW (or REM)” state transition (Fig. , middle panel) upon LSTM RNN predictions, further suggesting the validity of the above LSTM RNN prediction. State dependent global spatiotemporal pattern of SWRs Different arousal states are characterized by various neurophysiological events, such as abovementioned spindle and SWR in NREM sleep, sawtooth wave and pontine-geniculo-occipital wave in REM sleep. SWRs during AW or NREM state have been shown to play an important role in memory consolidation , but SWRs related global spatiotemporal pattern has not been fully explored, especially in different arousal states. Spindle and SWRs events were identified (Supplementary Fig. ) based on previous studies , . Using the neural-event-triggered (NET) fMRI approach , we observed SWRs evoked BOLD activations in HPF, RSP and ventral sensorimotor cortex (vSMC) during NREM state (Fig. , Supplementary Fig. – and Supplementary Data ). Interestingly, we also found significant deactivations in subcortical regions such as thalamus, hypothalamus and midbrain regions (Fig. and Supplementary Fig. ). In AW state, we obtained similar but much weaker spatiotemporal profiles (Fig. and Supplementary Fig. ), suggesting state dependency of SWRs evoked global patterns. Thus, we quantified above differences of SWRs evoked BOLD responses in medial prefrontal cortex (mPFC) and HPF (Fig. and Supplementary Fig. ) and found significantly higher BOLD responses in NREM state than those in AW state. Such difference of BOLD responses in mPFC was also significantly correlated with the difference of electrophysiological power in 3-45 Hz (Fig. ). In agreement with previous studies , SWRs exhibited a strong co-occurrence with spindle events (10-16 Hz) in our data (Fig. and Supplementary Fig. ). Thus, we explored whether the state dependent SWRs evoked BOLD response could attribute to the co-occurrence of spindle events. Using the same NET fMRI approach, we revealed that the spatiotemporal pattern of spindle-uncoupled SWRs in NREM state (Fig. , upper panel, Supplementary Fig. and Supplementary Data ) resembled the pattern of SWRs in AW state. And spindle-coupled SWRs exhibited similar spatiotemporal pattern but higher cortical activations, compared to the spindle-uncoupled SWRs (Fig. , lower panel, and Supplementary Fig. ). Difference of spindle-coupled and -uncoupled SWRs evoked BOLD responses (Fig. ) was significant and highly correlated with the difference of electrophysiological power in 4-42 Hz (Fig. ). Moreover, there was no significant difference of BOLD responses between SWRs in AW state and spindle-uncoupled SWRs in NREM state (Fig. and Supplementary Fig. ). The above results suggested that spindle co-occurrence enhanced the global activations of SWRs evoked BOLD responses, which was consistent with global positive BOLD responses of SWR-uncoupled spindle (Supplementary Fig. ). Therefore, the above results suggested that the co-occurrence of spindle events contributed to the enhanced SWRs evoked BOLD responses in NREM state. To investigate whether there were any synergistic effects of the SWRs and spindles triggered BOLD responses, we firstly summed the BOLD responses of solitary spindles (Fig. and Supplementary Fig. – ) and SWRs in each session (“summed responses”), as well as those of spindle coupled SWRs (“coupled responses”). Then, we conducted paired t-test across sessions between the above coupled and summed responses at regional (Fig. ) and whole brain levels (Fig. ). We found a “coupled > summed” response across cortical regions and “coupled <summed” response in thalamus (Fig. ), suggesting synergistic effects of the SWRs and spindles triggered BOLD responses. Slow waves are thought to participate in the regulation of NREM sleep process. Using the same NET-fMRI approach, we also calculated the slow wave triggered spatiotemporal map and found the BOLD activations in RSP and thalamus (Supplementary Fig. ). To investigate how arousal fluctuation contributes to the global dynamics, we established the mouse sleep fMRI setup using the simultaneous electrophysiology and fMRI. MR-compatible electrocorticography (ECoG) and depth electrodes were custom designed and fabricated for high MR-compatibility at 9.4 T, while maintaining good electrophysiological recording quality (Fig. ). Double-sided flexible printed circuit (FPC) with polyimide film as base substrate was used to minimize the thickness and width, and to maximize the flexibility of the array. Copper layer was used for its similar magnetic susceptibility to the brain (thus minimal MRI artefact), and gold was plated over copper layer in each recording site to avoid biological toxicity of copper. The 16-Channels ECoG electrodes covered a large portion of the cerebral cortex, including retrosplenial area (RSP), motor area, somatosensory area and posterior parietal association area (Fig. and Supplementary Data ). To record subcortical activity, we further fabricated a graphene fiber based depth electrodes. Graphene fiber was shown to be highly MR-compatible in our previous deep brain stimulation (DBS) study at 9.4 T, and was repurposed for LFP recording in the hippocampus (Fig. ). The pipeline of electrode implantation surgery, habituation training and data acquisition was shown in Fig. and described in Method. Both electrodes produced minimal MRI artifacts in T2 weighted anatomical images and T2* weighted functional images (Fig. ) of the relatively small sized mouse brain at 9.4 T. With highly MR-compatible electrodes providing the feasibility of monitoring animal’s arousal states, it remained challenging to record mouse sleep in the MRI environment. Based on our extensive previous experiences on awake mouse fMRI , and systematical optimization of stress level reduction , the current un-anesthetized mouse fMRI setup allowed simultaneous electrophysiological recording while minimizing animal’s stress and thus facilitating sleep (Fig. ). In particular, the mouse’s head was tilted for 30 degree and its forelimbs were allowed to move freely, both specifically designed for minimizing stress level . It is well known that the simultaneously acquired electrophysiological signal suffers from severe MRI artifacts. Thus, we established and evaluated an off-line de-noising preprocessing pipeline (Supplementary Fig. ), which greatly suppressed MRI artifacts to the extent that they no longer affected further analysis (Fig. ). Meanwhile, we also developed an fMRI preprocessing pipeline that was optimized for 4-hour sleep fMRI data. Raw fMRI data exhibited good (temporal) signal-noise-ratio (Fig. ) with small head motion and the “6 rp + 6 Δrp + 40 PCs” regression approach was modified from our previous studies , to minimize the effects of scanner drift, motion and other non-neural physiological noises (Fig. and Supplementary Fig. ). The 40 PCs were derived from fMRI signals outside the mouse brain (Supplementary Fig. ), largely capturing the non-neuronal signals , such as head motion (Supplementary Fig. ), physiological effects (Supplementary Fig. ) and infra-slow drift (Supplementary Fig. ). Thus, using the “6 rp + 6 Δrp + 40 PCs” regression approach, arousal induced non-neuronal nuisance effects on fMRI signals were minimized. With this simultaneous electrophysiology and fMRI recording setup, a representative session of mouse sleep fMRI with awake (AW), NREM and REM sleep states was shown in Fig. with denoised ECoG, electromyography (EMG) and fMRI signals. Sleep states were conventionally classified using electrophysiological signals (see Methods for details, Supplementary Fig. ). Similar to previous optical imaging studies of mouse sleep , we utilized a 4-hour long fMRI acquisition scheme. Mice exhibited notable and increasing portion of NREM sleep as the scan progressed, and REM sleep was also observed during the 4-hour scan (Fig. ). In total, 46 4-hour sessions (184 h) simultaneous ECoG/LFP-fMRI recordings from 24 mice (27 ECoG-fMRI sessions from 14 mice and 19 LFP-fMRI sessions from 10 mice) were acquired in our dataset, including 3588 min of NREM sleep and 342 min of REM sleep (Fig. ). The above dataset with both raw and preprocessed data is openly accessible at 10.12412/BSDC.1668502646.20001. Human sleep is characterized by long state durations, e.g., each NREM-REM cycle is approximately 90 min , which prevents conventional general linear model (GLM) analysis due to its low frequency beyond the detection limit of fMRI. However, mouse sleep is fragmented with short state durations (Supplementary Fig. , median duration of AW: 46.0 s, NREM: 41.0 s and REM: 121.5 s in our data), thus is convenient for analyzing state dependent activations. Using the conventional GLM analysis, we mapped the brain-wide BOLD activations of NREM and REM sleep, relative to the AW state (Fig. ). For the NREM state, a large part of cerebral cortex and hippocampus were activated, while subcortical regions, such as thalamus and part of midbrain, were deactivated compared to the AW state (Fig. ). Importantly, using the simultaneously acquired ECoG or LFP signals, we explored the potential relationship between sleep state dependent BOLD and electrophysiological signals. We calculated hemodynamic response function (HRF) -convolved ECoG/LFP band-limited power and its correlation with BOLD changes (Supplementary Fig. ). The relative changes of ECoG or hippocampal LFP were in generally consistent with current understanding of NREM and REM sleep , (Fig. , left panels). Importantly, we found a significant broadband (1–100 Hz) correlation between the relative ECoG power and relative cortical BOLD signal changes (Fig. , right panel) during NREM state, which was also evident in hippocampal CA1 recordings (Fig. , right panel). For REM state, significant correlation between theta band-limited power and BOLD changes was observed (Fig. , right panels), suggesting the potential relationship of theta band and BOLD changes. The above NREM and REM activation patterns were state-dependent features. However, previous studies indicated that sleep architectures are dynamic . Taking advantages of whole-brain mouse sleep fMRI data, we investigated the low dimensional dynamics within and between brain states. First, we conducted group principal component analysis (PCA) to BOLD signals for dimensional reduction. The −100th PC accounted for at least 0.3% explained variances and accumulatively the first 100 PCs resolved greater than 78% explained variances of BOLD signals. (Fig. , Supplementary Fig. and Supplementary Data ). The first four PCs were shown in Fig. . Each PC exhibited non-stationary temporal weights from the start to the end of each state (Fig. ), suggesting dynamic involvements of different functional networks within brain states. Further dynamics were expected across brain states, thus we characterized the state transition probabilities (Fig. ). Electrophysiological and whole brain BOLD features were shown in Fig. (upper panel) and Supplementary Fig. – for four state transitions, respectively. For each state transition, we observed sequential BOLD signal fluctuations traversing the mouse brain (Supplementary Fig. and Supplementary Data ). Further global spatiotemporal patterns of all four state transitions were shown in Supplementary Fig. – . Importantly, the temporal weights of PCs (tPCs) also exhibited diverse characteristics across state transitions (Fig. , lower panel, and Supplementary Fig. ), suggesting dynamic involvements of different functional networks across brain states. Moreover, in the low dimensional spaces spanned by BOLD PCs (Fig. and Supplementary Fig. ) and electrophysiological band-limited power ratios (Fig. ), activity flows of the manifolds exhibited dynamic properties within and across brain states. Interestingly, we found the separated trajectories between “AW to NREM” and “NREM to AW” transitions (Fig. ). And we further quantified such phenomenon and the significant asymmetry were observed around state transitions in both BOLD and electrophysiological space (Fig. ), suggesting “AW to NREM” and “NREM to AW” transitions were asymmetric processes. The dynamic characteristics of PCs across state transitions promoted us to investigate whether such transitions could be predicted by BOLD signals prior to electrophysiology defined transition time, and if so, which functional networks contributed to such predictions. Thus, we established a LSTM RNN model for state transition prediction, with BOLD signals (tPCs) preceding state transitions as model input and the brain state after such transition as model output (Fig. ). We systematically evaluated the model parameters, and the resulting optimized parameter set was 1 hidden LSTM layer with 50 hidden units (Fig. ). Prediction accuracy on the validation dataset was primarily related to the gap times preceding state transitions (Fig. ). Using the optimized parameter set, we found high prediction accuracy (Fig. ) of more than 85% across all seven prediction categories. The high prediction accuracy gradually decreased as the gap time increased (Fig. , upper panel), and the mean discriminate times were 10.55 s for “AW to NREM”, 3.28 s for “NREM to AW”, 17.84 s for “NREM to REM”, and 3.45 s for “REM to AW” state transitions (Fig. , middle panel). The null dataset was constructed by shuffling prediction categories. Correspondingly, sensitivity results of the LSTM RNN model revealed several regions significantly contributed to the prediction of state transitions (Fig. , lower panel, and Supplementary Fig. ), including (1) ventral thalamus (vThal), medial mammillary nucleus (MM), RSP, ventral tegmental area (VTA), median raphe nucleus (MnR), globus pallidus (external segment, GPe) and interposed cerebellar nucleus (Int) for “AW to NREM” state; (2) anterior cingulate area (ACA), agranular insular area (AI), visual areas (VIS), pontine reticular nucleus (ventral part, PnV), hippocampal formation (HPF) and primary somatosensory area (SSp) for “NREM to AW” state; (3) paragigantocellular reticular nucleus (PGRN), periaqueductal gray (PAG), HPF, VIS, medial preoptic area (MPO), supplemental somatosensory areas (SSs) and amygdala areas (AMY) for “NREM to REM” state; and (4) MM and pontine central gray (PCG) for “REM to AW” state. Interestingly, we found HPF exhibited significant contributions on the prediction of state transitions from NREM to AW and REM states. Utilizing the GF electrode placed in hippocampus CA1 region of our LFP-fMRI dataset, we compared the LFP power spectrum between “NREM to AW” and “NREM only” states, and found significantly higher power in three frequency bands, including (1) 7-37 Hz from −14 to −2 s, (2) 44-95 Hz from −12 to −1 s, and (3) 119-205 Hz from −11 to −3 s before state transitions (Fig. ). Also, we conducted same analysis between “NREM to REM” and “NREM only” states, and found significantly higher power in two frequency bands, including (1) 1-34 Hz from (beyond) −30 to −3 s, and (2) 128-141 Hz from (beyond) −30 to −13 s before state transitions (Fig. ). These “NREM only” epochs were randomly selected from the sleep fMRI dataset, and epochs of “NREM to AW”, “NREM to REM” and “NREM only” with short durations (<60 s) were excluded. Then, two sample t-test was conducted with threshold of p < 0.05. Furthermore, we isolated SWR and spindle events, among other events, based on LFP dataset (Fig. ). Altogether, SWR and spindle events showed a pronounced decrease of their occurrence probability before “NREM to AW (or REM)” state transition (Fig. ). To be note, the preceding time of SWR event changes (Fig. ) was similar to the mean discriminate time of “NREM to AW (or REM)” state transition (Fig. , middle panel) upon LSTM RNN predictions, further suggesting the validity of the above LSTM RNN prediction. Different arousal states are characterized by various neurophysiological events, such as abovementioned spindle and SWR in NREM sleep, sawtooth wave and pontine-geniculo-occipital wave in REM sleep. SWRs during AW or NREM state have been shown to play an important role in memory consolidation , but SWRs related global spatiotemporal pattern has not been fully explored, especially in different arousal states. Spindle and SWRs events were identified (Supplementary Fig. ) based on previous studies , . Using the neural-event-triggered (NET) fMRI approach , we observed SWRs evoked BOLD activations in HPF, RSP and ventral sensorimotor cortex (vSMC) during NREM state (Fig. , Supplementary Fig. – and Supplementary Data ). Interestingly, we also found significant deactivations in subcortical regions such as thalamus, hypothalamus and midbrain regions (Fig. and Supplementary Fig. ). In AW state, we obtained similar but much weaker spatiotemporal profiles (Fig. and Supplementary Fig. ), suggesting state dependency of SWRs evoked global patterns. Thus, we quantified above differences of SWRs evoked BOLD responses in medial prefrontal cortex (mPFC) and HPF (Fig. and Supplementary Fig. ) and found significantly higher BOLD responses in NREM state than those in AW state. Such difference of BOLD responses in mPFC was also significantly correlated with the difference of electrophysiological power in 3-45 Hz (Fig. ). In agreement with previous studies , SWRs exhibited a strong co-occurrence with spindle events (10-16 Hz) in our data (Fig. and Supplementary Fig. ). Thus, we explored whether the state dependent SWRs evoked BOLD response could attribute to the co-occurrence of spindle events. Using the same NET fMRI approach, we revealed that the spatiotemporal pattern of spindle-uncoupled SWRs in NREM state (Fig. , upper panel, Supplementary Fig. and Supplementary Data ) resembled the pattern of SWRs in AW state. And spindle-coupled SWRs exhibited similar spatiotemporal pattern but higher cortical activations, compared to the spindle-uncoupled SWRs (Fig. , lower panel, and Supplementary Fig. ). Difference of spindle-coupled and -uncoupled SWRs evoked BOLD responses (Fig. ) was significant and highly correlated with the difference of electrophysiological power in 4-42 Hz (Fig. ). Moreover, there was no significant difference of BOLD responses between SWRs in AW state and spindle-uncoupled SWRs in NREM state (Fig. and Supplementary Fig. ). The above results suggested that spindle co-occurrence enhanced the global activations of SWRs evoked BOLD responses, which was consistent with global positive BOLD responses of SWR-uncoupled spindle (Supplementary Fig. ). Therefore, the above results suggested that the co-occurrence of spindle events contributed to the enhanced SWRs evoked BOLD responses in NREM state. To investigate whether there were any synergistic effects of the SWRs and spindles triggered BOLD responses, we firstly summed the BOLD responses of solitary spindles (Fig. and Supplementary Fig. – ) and SWRs in each session (“summed responses”), as well as those of spindle coupled SWRs (“coupled responses”). Then, we conducted paired t-test across sessions between the above coupled and summed responses at regional (Fig. ) and whole brain levels (Fig. ). We found a “coupled > summed” response across cortical regions and “coupled <summed” response in thalamus (Fig. ), suggesting synergistic effects of the SWRs and spindles triggered BOLD responses. Slow waves are thought to participate in the regulation of NREM sleep process. Using the same NET-fMRI approach, we also calculated the slow wave triggered spatiotemporal map and found the BOLD activations in RSP and thalamus (Supplementary Fig. ). In the current study we developed the mouse sleep fMRI method based on simultaneous electrophysiological and fMRI recording in un-anesthetized mouse at 9.4 T. Our method provided the global view of sleep state-dependent patterns and state-transition dynamics. Furthermore, we revealed the potentiated SWR-evoked BOLD response in NREM state compared to that in AW state, largely attributed to the co-occurrence of spindle events. Therefore, our method demonstrated the unique capability of revealing the global sleep dependent features, and provides an accessible platform for sleep research and investigation of neural basis of the arousal related fMRI signal. It is well known that simultaneous recording of electrophysiological and fMRI signals is technically challenging, particularly in un-anesthetized animals. Such difficulty mainly arises from the mutual electromagnetic interference and head motion related complications in awake animals. To achieve high MR-compatibility, our 16-Channel ECoG electrodes and GF depth electrodes were specially designed to achieve minimal MRI distortions and high-quality electrophysiological recordings. Such good compatibility is important for simultaneous electrophysiological fMRI recording in mice, as the mouse brain is much smaller than brains of human and monkey, thus it is more prone to imaging artifacts of electrodes. With these developed electrodes, after de-noising procedures (Fig. , Supplementary Fig. and Supplementary Data ), the characteristics of electrophysiological signals were in line with those outside the MRI environment, such as amplitude and power spectrum , . Interestingly, we did not observe typical pulse artifact from the ECoG and LFP power spectrums (Fig. and Fig. ), which is often problematic in human EEG-fMRI studies . After our electrophysiological denoising procedure, no notable ballistocardiogram artifact was observed in 5–10 Hz frequency band, corresponding to the unanesthetized mouse heart rate between 300 and 600 beats per minute. This phenomenon might attribute to our intracranial placement of electrodes and the stable electrode fixation. In human EEG-fMRI studies, pulse artifact occurs when an EEG electrode is placed over a pulsating vessel . The pulsation can cause slow electrode movements that contaminate EEG activities. Our intracranial electrodes were tightly fixed on the mouse skull by dental cement. Thus, scalp pulse and cardiac-related motion were less likely to impact the electrode movement and further influence our electrophysiological signals. However, previous studies combined electrophysiology and fMRI were mainly conducted in anesthetized animals, e.g., rat and macaque, which prevented the utilization in sleep research. Based on our extensive experience on awake mouse fMRI , , in the current study we further demonstrated that mouse could sleep in the noisy MRI environment. This critical improvement paved way for multimodal fMRI research in mice, and would also be highly meaningful for other sleep related MRI investigation such as diffusion MRI . Mouse fMRI is uniquely suited to reveal the state dependent global patterns and state transition dynamics, as its sleep is fragmented with short state duration and thus frequent state transitions . In contrast, the long sleep state durations of human and monkey prevent the conventional GLM analysis to map the state dependent activations, due to its low frequency beyond the detection limit of fMRI. For example, each NREM-REM cycle is approximately 90 min in humans . Such long cycles also prevent fMRI analysis of state transitions as there would be so few transitions in each scan. For comparison, our 184 h mouse sleep fMRI data included 3851 “AW to NREM”, 3683 “NREM to AW”, 168 “NREM to REM” and 168 “REM to AW” transitions (Fig. ), which enabled further detailed analysis of state transitions. The dynamics of broadly defined state transition and fluctuations in resting-state fMRI is now widely investigated, and has been implicated in cognitive processes , and brain disorders, e.g. Alzheimer’s disease (AD) and obsessive–compulsive disorder (OCD) . The functional network of those patients with brain disorders exhibited abnormal dynamic rhythms , , indicating potential clinical relevance. And also, consciousness has been shown to modulate the diversity of the state dynamics across different sedation levels . Utilizing the rich transgenic mouse disease models, e.g., various AD mouse models, future research can be conducted based on our mouse sleep fMRI setup to further investigate the mechanisms of state transition dynamics and its role in brain disorders. The abundant sleep transitions in mouse sleep fMRI greatly facilitated the investigation of the macroscopic cerebral dynamics, for which we utilized group PCA. A major advantage of PCA over other analytic approaches is that it imposes orthogonality onto the components, which is crucial for providing a low-dimensional subspace to embed the state space manifold. Other dimension reduction methods (such as independent component analysis) find a different set of optimal solutions (such as maximal statistical independence), but these are not, in general, linearly independent. Notably, state space attractors (Fig. ) are invariant to linear transformations of their embedding phase space as long as the dimensions remain orthogonal. Hence, PCA enables analysis of the state space trajectory (or flow), which reflects the temporal evolution of the global brain state . Combined with rich resources and tools in mice, more features of dynamic microstructure within mouse sleep states can be explored in future studies. Particularly, the asymmetrical trajectories in both electrophysiological and BOLD spaces indicated the asymmetry between “AW to NREM” and “NREM to AW” transitions (Fig. ), further suggesting different neural circuits for awake-promoting and NREM sleep-promoting processes. Our results provided a framework for sleep research through the lens of complex dynamical systems, linking the flow of electrophysiological signatures to the dynamic reconfiguration of functional networks in the low-dimensional state space. The within- and between-state dynamics was also clearly demonstrated by prediction of state transitions prior to electrophysiology defined transition time point using LSTM RNN models on BOLD signals. The long preceding discriminate times (AW to NREM: 10.55 s, and NREM to REM: 17.84 s) were significantly longer than previous reported neurophysiological results and considering the hemodynamic delay of 1.5–2 s in mice , it may be slightly longer if direct neural signals were used. For example, the firing rate of sleep-promoting neurons in the POA and wake-promoting neurons in the Locus Coeruleus (LC), tuberomammillary nucleus (TMN), and BF altered before the transition for less than 1 s , and the firing rate of GABAergic neuron in ventral medulla (vM) increased precede the NREM-REM transition less than 15 s . Key regions contributing to the prediction accuracy of state transitions (Fig. ) in our LSTM RNN model may also be interesting for further research. Part of these regions have already been implicated in sleep modulation, such as ventrolateral periaqueductal gray (vlPAG) , PGRN and AMY in promoting REM sleep, and GPe and thalamus nucleus in promoting NREM sleep. Other sensitive regions might also play vital roles in modulating the awake-sleep cycle, e.g., VTA, MnR. Previous studies have shown that a subset cells of VTA drive NREM and REM sleep through the lateral hypothalamus and fatal insomnia diseases are associated with raphe nuclei degeneration . The remaining sensitive regions provide more candidates for further sleep research. Such long discriminate times and region-specific contribution demonstrated the global and sequential nature of sleep transitions, which further emphasized the need for a systematic whole-brain view to understand its fundamental mechanism. Sleep-wake cycle is believed to be tightly regulated by a distributed network of sleep and wake promoting neurons , , primarily located in subcortical regions. Recently, cortical regulation of sleep transition and maintenance has also been reported. Silencing neocortical layer 5 pyramidal neurons using SNAP25 knockout mouse decreased the cortico-subcortical communication and further markedly increased wakefulness . In our study, we found widespread changes of cortical activities during “AW to NREM” and “NREM to AW” state transitions (Supplementary Fig. ), and some changes occurred earlier than those in subcortical regions, highlighting the important role of the cortex. Another notable example of cortical involvement in sleep regulation is retrosplenial cortex (RSP). Two recent studies in mice both showed that RSP was critically involved in REM sleep initiation and progression , . The current study also found RSP was highly activated during both NREM and REM state, and the activity of posterior DMN-like network (PC4 in Fig. ), largely overlapping with RSP, apparently preceded the “NREM to REM” state transition. Therefore, our study agrees well with recent progress on the cortical involvement in sleep-wake transitions. Our MR-compatible LFP-fMRI setup enabled us to simultaneously record CA1 LFP signals and obtain good BOLD signals around the electrodes, including CA1 and RSP. As a characteristic event in hippocampus, SWRs have been shown to play an important role in memory consolidation and coordinated hippocampal-thalamic-cortical communication . Using NET fMRI approach, we provided the global view of SWR-evoked BOLD spatiotemporal pattern in NREM and AW states. Activation in RSP and hippocampus and deactivations in thalamus were consistent with previous results in macaque . And optical imaging during natural sleep and anesthesia in mice also showed significant activation of RSP during SWRs . Converging evidence, including the current results, clearly suggests the important role of RSP during SWRs, which may be related to the critical role of subiculum-retrosplenial pathway for SWR propagation from hippocampus to neocortex . Interestingly, larger SWR-evoked BOLD responses in NREM state might attribute to the co-occurrence of spindle events (Fig. ), as the spindle events showed whole-brain positive BOLD responses (Fig. and Supplementary Fig. ). In addition, we found the co-occurrence of SWRs and spindles elicited enhanced BOLD responses, compared to the sum of the responses of two solitary events (Fig. ). Such effect was most notable in hippocampus, RSP and thalamus. Previous studies have shown that the interaction between thalamocortical spindles and hippocampal ripples promoted memory consolidation . Interruption of the synchronization between ripple and spindle events appeared to interfere the efficiency of memory consolidation . Thus, we speculate that the synergistic effects of SWRs and spindles might be related to the memory consolidation process, and further research is needed to examine the functional relevance of such phenomenon. Moreover, slow waves are thought to be critical for initiating information transfer between hippocampus and neocortex . A previous human EEG-fMRI study showed significant slow wave evoked activations in right parahippocampal gyrus, precuneus and posterior cingulate cortex, which were in good accordance with our results (Supplementary Fig. ). It is known that the relationship among slow waves, spindles and SWRs is complex and may contribute to many cognition processes , especially memory consolidation , . Thus, the relationship among these events can be further explored, potentially using the current dataset. Our mouse sleep fMRI setup provides a tool for investigating global spatiotemporal patterns of state dependent neural events and the mutual relationship of those events. It is known that arousal may contribute to neuronal dynamics and non-neuronal variations , e.g., vascular effects, head motion, and physiology, and they both contributed to fMRI dynamics. Various studies investigated the methods to remove arousal related non-neuronal variations, including global signal regression , data-driven approaches, e.g., ICA-FIX and model based approaches, e.g., RETROICOR . In our study, we applied a regression-based de-noising method modified from our previous studies , , in which the “6 rp + 6 Δrp + 40 PCs” nuisance signals were used as regressors. The 40 PCs were derived from fMRI signals of non-brain tissues, largely capturing the non-neuronal signals , such as head motion, scanner drift, and physiological effects (Supplementary Fig. ). Using the above regression approach, we believed that arousal induced non-neuronal nuisance effects on fMRI signals were largely suppressed. Naturally, arousal fluctuations also modulate large scale brain activities , which has been shown to further contribute to resting-state fMRI dynamics . Across different brain states in human, the network structure of spontaneous BOLD fluctuations was associated with that of slow electrophysiological activities . Furthermore, arousal fluctuation synchronized the brain’s functional systems through global wave propagations based on human fMRI and macaque ECoG . Similar global wave propagation of infra-slow activity was shown in mouse based on the calcium and hemodynamic imaging in anesthetized and awake states . Meanwhile, another human fMRI study showed different infra-slow propagation patterns between slow wave sleep and wakefulness . Therefore, infra-slow brain activity across arousal states may orchestrate a wide range of interrelated neurophysiological and autonomic processes, and thus serve as a neural basis of low frequency spontaneous BOLD activity . As arousal fluctuations are intricately linked to both neural and non-neural components in BOLD signals, the current mouse sleep fMRI setup may be advantageous to further investigate the neural basis of arousal related BOLD activity, with the simultaneous recording of electrophysiological and BOLD signals. Our current method of mouse sleep fMRI, and the open source data acquired using this method, features substantial arousal fluctuations from awake to REM sleep with simultaneously acquired cortical (ECoG) and hippocampus (LFP) signals. Therefore, our method and dataset provide an avenue to investigate the neural basis of arousal related fMRI dynamics. In our study, there are several important limitations that need further improvement in the future. First, imaging coverage of deep brain regions was limited by the single loop receive coil used in our imaging setup. The signal-to-noise ratio can be further improved in deep brain regions, such as midbrain and pons, which are important for awake/sleep modulation . Then, the sleep scoring criteria employed in the current study (Supplementary Fig. and Supplementary Data ) was largely adopted from previous mouse sleep studies in the free-moving state. As no prior knowledge about ECoG/LFP and EMG signal characteristics during mouse sleep fMRI was known, the semi-automated sleep scoring approach developed here may be further refined. While we designed the 30° head holder to mimic the natural posture, sleep in the natural free-moving state may still be different from that in the head fixed state, which may result in different ECoG/LFP and EMG signal characteristics. Although other optical imaging studies of mouse sleep conducted in the head fixed condition also employed similar sleep scoring criteria , , further detailed examination is much needed for improving sleep scoring in the head fixed state. Thus, optimization of receive coils is expected in the future for wider and better coverage of brain. In addition, the BOLD signal is the indirect measurement of neural activities based on neurovascular coupling . Arousal fluctuation introduces physiological changes, e.g. respiration and heart rate, which could affect BOLD signal. In our fMRI preprocessing, the physiological nuisance could be largely captured by PCs of the signals from the outside brain (Supplementary Fig. ), in which the first 40 PCs showed significant correlation with respiration rate changes and were subsequently regressed out from fMRI data. Moreover, the significant correlation between electrophysiological and BOLD signals (Fig. ) indicated the potential neurophysiological relevance of NREM and REM state dependent BOLD activations. Nevertheless, as discussed above, such complex relationship between arousal related BOLD signal and neural activity is an important research direction itself, for better preprocessing strategies and more precise dissection of neural and non-neural contributions. Furthermore, recent advances in non-hemodynamic based fMRI such as direct imaging of neuronal activity (DIANA) methods , might eventually overcome such limitation. In conclusion, we developed the mouse sleep fMRI method in un-anesthetized mice based on highly MR-compatible simultaneous electrophysiology at 9.4 T. For sleep research, this method provides an important avenue to investigate sleep dynamics at a global scale, and has great potential to integrate local and global view of sleep when further combined with other circuitry level tools in mice. For fMRI research, our method provides a convenient platform to examine the neural basis of the arousal related fMRI signal. Furthermore, our open source dataset of mouse sleep fMRI would provide a valuable source for both experimental and theoretical neuroscientists, and may help to establish a general multiscale framework of sleep from molecular, neuronal, circuitry and whole-brain levels. Animals Male wide type C57BL/6 J mice were obtained from Shanghai Laboratory Animal Center (Shanghai, China) at 8–10 weeks of age, weighted 20–30 g. Animals were group housed (5–6/cage) in the standard laboratory condition (temperature: 23 ± 1 °C; humidity: 50–70%) under a 12 h light/dark cycle (light on from 7 a.m. to 7 p.m.) with food and water ad libitum . All animal experiments were approved by the Animal Care and Use Committee of the Institute of Neuroscience, Chinese Academy of Sciences, Shanghai, China. MR-compatible electrodes Two types of MR-compatible electrodes were developed. First, an 18-contacts MR-compatible surface electrode was custom designed and manufactured for recording electrocorticogram (ECoG) and electromyography (EMG) signals (Fig. ). The array was made up of a flexible printed circuit (FPC) and two insulated gold wires, all soldered to an Omnetics 18-pin MR-compatible connector (A70242-001, Omnetics, USA, Minneapolis). The FPC was consisted of a 30 μm thin polyimide film, a 12 μm copper film, and a 3 µm gold-plated layer, with 14 ECoG recording site and two recording reference/ground (100 µm diameter; impedance, ~150 kOhm at 1KHz). Two insulated gold wires were used for EMG recording (50 µm diameter; impedance ~500 kOhm at 1 KHz). Secondly, a MR-compatible depth electrode (Fig. ) was developed with 3 insulated gold wires, 2 bare silver wires and 1 graphene fiber (GF) electrode soldered to a flexible flat cable. One of three insulated gold wires (50 µm diameter; impedance, ~500 kOhm at 1 KHz) was used for recording ECoG, and the rest two for recording EMG. Two bare silver wires (100 µm diameter; impedance, ~60 kOhm at 1 kHz) were used as reference and ground. To record the hippocampal LFP signal, GF electrodes (impedance, ~60 kOhm at 1 kHz) were custom designed. GFs were fabricated through a one-step dimensionally confined hydrothermal process using suspensions of graphene oxide (GO) (monolayer, thickness: 0.8–1.2 nm; sheet diameter: 0.5–5 µm; #XF002-2, Nanjing/Jiangsu XFNANO Materials Technology, China). In a typical preparation, an 8 mg mL −1 aqueous GO suspension was injected into a glass pipeline with a 0.9 mm inner diameter using a syringe. After being baked at 230 °C for 2 h with the two ends of the pipeline sealed, a GF matching the pipe geometry was produced. This preformed GF was then released from the pipeline by flow of N 2 and dried in air. The dried GF had a reduction in diameter to ~75 µm due to water loss and drying-induced alignment of the GO sheets. A GF with diameter of ~75 µm and length of 3 mm was connected to a bare copper wire with diameter of 100 µm using elargol. Parylene-C film of ~5 µm thickness was deposited onto the fibers in a custom made low-pressure coating system to finish the GF electrodes fabrication. Thus, the GF electrodes enabled unbiased fMRI mapping and excellent electrochemical performance, including low impedance and high electrical conductivity, which were not achievable by other electrodes. Surgical procedures All surgical procedures were conducted under the standard aseptic condition. Mice were pretreated with 5 mg kg −1 dexamethasone intraperitoneally 1 h before surgery to prevent brain edema. After anesthetized with isoflurane, mice were secured in a stereotaxic apparatus with a heating mat (mouseSTAT, Kent scientific cooperation). A midline sagittal incision was made along the scalp to expose the skull. The periosteum from the skull was removed by saline with cotton-tip applicator. After skull was dried out, a coat of self-etch adhesive (3 M ESPE Adper Easy One) was applied followed by light curing. For MR-compatible ECoG electrode implantation, a sterile dental drill was used to drill off the surface of bone in the shape shown in Fig. (length ~4 mm; width, ~6 mm; centered on bregma), with the bone around raphe retained. The window area was kept wet with saline, and the skull was slowly detached from the dura and raised at about 30° using a nasal stripper and tip tweezers. Gelatin sponge was applied to stop bleeding and keep the dura moist. Afterwards, the FPC part of the ECoG electrode was implanted epidurally with reference and ground sites attached on the bone of raphe. 2% sterilized agarose saline solution was filled between the skull and FPC. After the agarose was solidified, the window area was covered with light curing flowable dental resin to fix the FPC. Later, the two gold wires were implanted into posterior neck muscles for recording EMG. A head holder for awake imaging was then attached on the skull above the cerebellum. Finally, dental cement was applied to smooth the surface of exposed skull. Mice were injected with meloxicam 0.5 mg kg −1 (Baoding sunlight herb medicament, CN) subcutaneously for seven consecutive days post-surgery. For MR-compatible depth electrode implantation, four 0.1 mm diameter holes were drilled in the skull. GF electrode was implanted into the dorsal CA1 region of the left hippocampus at stereotaxic coordinates AP = − 1.94 mm, ML = − 1.7 mm and DV = − 1.15 mm. One gold wire was inserted epidurally at AP = + 1 mm, ML = − 1 mm to record ECoG, and the other two gold wire were implanted into posterior neck muscles to record EMG. Two bare silver wires were used as reference and ground, and inserted epidurally at ML = 0, AP = − 5.5 mm and −6.5 mm, respectively. Light curing flowable dental resin was used for fixation of electrodes and head holder. Other procedures were identical as described above. Habituation After seven-day recovery, mice were then habituated for fMRI for another seven days. Mice were head fixed on the animal bed with the recorded acoustic MRI scanning noise based on previous work . Optical imaging studies usually record 3–5 h to obtain mice sleep more efficiently , , as head fixed mice frequently fall asleep after 1.5 h and REM sleep frequently occurs after 2.5 h . Therefore, 4 h head restraining was also utilized in our mouse sleep fMRI. The 30° head holder was designed to fit the natural sleep gesture of mice . The animal bed was modified to allow more space for forelimb movement to reduce stress level . The habituations were all carried out during 9a.m.−15p.m. with a fixed duration of 4 h and gradually increased noise levels. The detailed schedule was listed in Table . No reward was given during or after the habituation training. Simultaneous neurophysiological recording and MRI acquisition Neurophysiological signals were fed through a ZIF-CLIP anaglog headstage transmitter, PZ5 amplifier, RZ2 BioAmp processer, and finally recorded by the WS-8 workstation with Synapse software (all from Tucker-Davis Technologies, USA, Alachua). Neurophysiological signals were recorded at a sampling rate of 24414 Hz, high-pass filtered at 0.1 Hz and notch filtered at 50, 100 and 150 Hz, except for LFP signals, which were recorded without notch filter. PZ5 was connected to the external ground (waveguide tube of MRI) for stabilizing the signals. Respiration signal and MRI trigger signal were recorded at a sampling rate of 1024 Hz. All MRI data were acquired with a Bruker BioSpec 9.4 T scanner (Software: ParaVision 6.0.1). An 86 mm volume coil was used for transmission and a single loop mouse head coil (Bruker, 1 cm diameter) was used for receiving. Mouse was head-fixed as described in previous section without using any anesthesia. A T2 weighted RARE anatomical image (TR: 3200 ms; TE: 34 ms; matrix size: 256 × 128; FOV: 18 × 9 mm 2 ; slice thickness: 400 μm; resolution: 70 × 70 μm 2 ) was acquired for coregistration purpose. After local shimming using Mapshim, functional images were acquired using single-shot echo planar imaging (EPI) with the following parameters: TR 2000 ms, TE 14 ms, FA 70°, matrix size 90 × 45, nominal in-plane resolution 200 × 200 μm 2 , slice thickness 400 μm, slices number 22, 7200 EPI volumes (4 h). All 7200 volumes were acquired in a single EPI scan. Triggers were sent from the MRI console for each slice acquisition and recorded along with the electrophysiological signal. Electrophysiological signal processing The off-line correction of MRI gradient artifacts in the electrophysiological signals was conducted using the fMRI Artifact Slice Template Removal algorithm (FASTR) (Supplementary Fig. ). Briefly, the imaging artifact waveforms were segmented, averaged and iteratively subtracted from the raw electrophysiological signals, according to the concurrently acquired trigger signal from the MRI scanner. This procedure was performed through FMRIB in EEGLAB ( https://fsl.fmrib.ox.ac.uk/eeglab/fmribplugin/ ). After careful visual inspection, we found a few noisy ECoG channels in a small number of scans. Thus, we interpolated the noisy ECoG signal by weighted averaging signals from other good ones, similar to the previous study using the neighbor interpolation method (Supplementary Fig. and Supplementary Data ). Considering there were k ( k < 4) bad channels in 14 channels, the weighted averaging method could be simply formulated as 1 [12pt]{minimal} $${S}^{{k}_{i}}=_{j=0}^{14-k}{({{Dis}}_{{k}_{i}-{good}})}^{ }{ S}_{{good}}^{\,j}$$ S k i = 1 14 − k ∑ j = 0 14 − k D i s k i − g o o d λ ⋅ S g o o d j where [12pt]{minimal} $${S}^{{k}_{i}}$$ S k i , [12pt]{minimal} $${S}_{{good}}$$ S g o o d and [12pt]{minimal} $${{Dis}}_{{k}_{i}-{good}}$$ D i s k i − g o o d were the ECoG signals of the [12pt]{minimal} $${k}_{i}$$ k i th interpolated channel, all good channels and the corresponding Euclidean distance, and [12pt]{minimal} $$$$ λ represented the exponential constraints ( [12pt]{minimal} $$$$ λ <0) using the weighted averaging method. Parameter [12pt]{minimal} $$$$ λ was estimated using other scans without noisy ECoG channels based on the least square fitting. Negative [12pt]{minimal} $$$$ λ indicated farther good channels contributed lower weights on the interpolation of [12pt]{minimal} $${k}_{i}$$ k i th noisy channel. The denoised electrophysiological signals were further down-sampled to 1024 Hz for subsequent analysis. For brain state classification , one channel within each session was selected for further analysis, and the selected channels were listed in the Supplementary Data . Then, we calculated the power spectrum for the ECoG/iEEG and EMG (160–250 Hz) data with 3 s sliding windows and 1 s step size, using the multi-taper method implemented in Chronux ( http://chronux.org/ ). Next, we computed the theta (6–12 Hz) and delta (1–4 Hz) power and theta/delta power ratio, which were further smoothed using “medfilt1” (20 points) in MATLAB. For each session, we used the following criteria for tentatively defining brain states: (1) a time point was classified as NREM sleep if the smoothed delta power was higher than its mean; (2) a time point was assigned as REM sleep if the smoothed theta/delta power ratio was two standard deviations higher than its mean and the EMG power was one standard deviation lower than its mean; and (3) all remaining time points were classified as AW state. Then, we further manually adjusted the classification of brain states as following: (1) For NREM state, according to the ECoG power spectrum, we adjusted the start or end point to the point with the greatest ascent or descent speed of smoothed delta power. For REM state, the start point was adjusted to the end point of the previous NREM, and the end point was adjusted to the point with the greatest descent speed of smoothed theta power. If the greatest ascent or descent point of EMG power was different from that of ECoG/iEEG signals, the midpoint between the two was defined as the transition point; (2) For sessions with noisy EMG recordings, we used the head motion (framewise displacement) estimated from fMRI data as substitute for EMG power; and (3) epochs with short durations (<5 s) were manually merged to the nearest sleep stage. An example of the above classification procedure (session 1: channel 10, 11500–13000 s) was shown in Supplementary Fig. . Spindle and SWRs events were identified in the LFP dataset based on previous studies , . To identify the spindle events, raw iEEG signals in mPFC were bandpass filtered (10–16 Hz) with Butterworth filter. A spindle event was identified if the envelope of the filtered iEEG signal was larger than its mean + 1.5 s.d. in NREM state (Supplementary Fig. ). To identify the SWRs events, raw LFP signals in CA1 of hippocampus were bandpass filtered (120–250 Hz) with Butterworth filter. A SWRs event was detected if (1) the envelope of the filtered LFP signal larger than its mean + 3 s.d. and (2) the power of the filtered LFP signal larger than its temporal mean. The center of spindle or SWRs event was defined as the time of maximum peak of the threshold-passed envelope. The beginning and the end (i.e., event duration) was measured before and after this maximum peak when the amplitude dropped below the mean value of the corresponding envelope (Supplementary Fig. ). Only spindles with 0.4–3 s durations were included. Slow waves were identified based on the procedures described previously . The raw signal was first down-sampled to 1024 kHz. Then, for slow waves detection, mPFC iEEG signals were filtered between 0.3 and 4.5 Hz with a two-order Butterworth bandpass filter. A slow wave was detected in NREM state if the following three criteria were all fulfilled: (1) the interval (T) of negative wave between 0.4 and 2.0 s; (2) top 35% negative amplitude (N) and (3) top 45% negative-to-positive peak-to-peak amplitude (M). Slow wave onset and offset was defined by the time of the first and third zero crossing, respectively (Supplementary Fig. ). fMRI data preprocessing All subsequent procedures were performed using custom scripts in MATLAB 2020a (MathWorks, Natick, MA) and SPM12 ( http://www.fil.ion.ucl.ac.uk/spm/ ). The mouse brain was extracted manually using ITK-SNAP ( http://www.itksnap.org/ ). First, each fMRI scan was slice-timing corrected and registered to the scan-specific structural image using rigid body transformation and the scan-specific structure image was then nonlinearly transformed to a study-specific mouse template ( https://atlas.brain-map.org/ ) for group analysis. Then, a light spatial smoothing (0.4 mm isotropic Gaussian kernel) was performed but no band-pass filter was applied to the BOLD time series. Furthermore, to minimize the effects of scanner drift, motion and other non-neural physiological noises, BOLD signals were regressed by “6 rp + 6 Δrp + 40 PCs” nuisance signals , (Supplementary Fig. ). “6 rp + 6 Δrp” nuisance signals represented 6 head motion parameters and their 1st order first derivatives, and “40 PCs” were the first 40 principal components from the BOLD signals of non-brain tissue, e.g., the muscles. The regression-based denoising strategy was adopted from a previous study , in which the PCs estimated from tissues outside the brain were used to model non-neural signal variations and produced a moderate improvement in specificity. The Pearson’s correlation coefficients between the frame-wise displacement (FD) and DVARS (D referring to temporal derivative of time courses, VARS referring to RMS variance over voxels) were calculated to quantitatively reflect the extent to which the motion related signal was reduced by given regressors (Fig. ). General linear model of brain state activation and electrophysiological validation GLM based statistical analysis was conducted using the mouse-specific HRF from our previous study , in which NREM and REM states were set as the predictors and thus the AW state was used implicitly as the baseline. Standard first level analysis was done for individual EPI scans. For second level analysis, flexible factorial model (brain states and mouse individuals) was conducted to generate the activation maps with FDR corrected p < 0.05. Furthermore, to investigate the electrophysiological relevance of BOLD signal variations in NREM and REM states, we estimated the Pearson’s correlation coefficients between relative ECoG (or LFP) powers (vs. AW state) and corresponding relative BOLD amplitudes. Group principal component analysis (PCA) of BOLD fMRI signal Each individual EPI scan comprising t time points and v voxels could be represented as a 2-dimensional time-space matrix ( [12pt]{minimal} $${S}_{t v}$$ S t × v ). In the initial scan-level PCA step, fMRI data of each scan was reduced to p components of dimension ( [12pt]{minimal} $${W}_{p v}$$ W p × v , p « t ). In the second PCA step, group PCA was performed on the concatenated data ( [12pt]{minimal} $${W}_{{pN} v}$$ W p N × v ) from all scans, in which data from N scans were stacked along the reduce dimension. Then, the group spatial PCs ( [12pt]{minimal} $${W}_{{p}_{0} v}$$ W p 0 × v ) were back-projected to raw data ( [12pt]{minimal} $${S}_{t v}$$ S t × v ) to reconstruct the time courses of each component (tPCs: [12pt]{minimal} $${S}_{t {p}_{0}}$$ S t × p 0 ) for each individual EPI scan. Brain state prediction using LSTM RNNs The long short-term memory (LSTM) recurrent neural networks (RNNs) model was built to predict the brain state based on its functional profile, i.e., tPCs, and their temporal dependency of BOLD signals on its preceding time points. Given the L most recent timesteps [12pt]{minimal} $$\{{X}_{t-L+1},{X}_{t-L+2},,\, {X}_{t}\}$$ X t − L + 1 , X t − L + 2 , … , X t , the goal at timestep t was to predict the state of M timesteps into the future, [12pt]{minimal} $${}_{t+M}$$ X ^ t + M . The architecture of the LSTM RNNs used in this study was illustrated in Fig. , including one (or two or three) hidden LSTM layer(s) and one fully connected layer. Multiple hidden LSTM layers could be used to encode the functional information with temporal dependency for each time point, and the fully connected layer was used to learn a mapping between the learned feature representation and brain states. The functional representation encoded in each LSTM layer was calculated as 2 [12pt]{minimal} $${f}_{t}^{l}= ({W}_{f}^{l} [{h}_{t-1}^{l},\, {x}_{t}^{l}]+{b}_{f}^{l})$$ f t l = σ W f l ⋅ h t − 1 l , x t l + b f l 3 [12pt]{minimal} $${i}_{t}^{l}= ({W}_{i}^{l} [{h}_{t-1}^{l},\, {x}_{t}^{l}]+{b}_{i}^{l})$$ i t l = σ W i l ⋅ h t − 1 l , x t l + b i l 4 [12pt]{minimal} $${}_{t}^{l}= ({W}_{C}^{l} [{h}_{t-1}^{l},\, \,{x}_{t}^{l}]+{b}_{C}^{l})$$ C ~ t l = tanh W C l ⋅ h t − 1 l , x t l + b C l 5 [12pt]{minimal} $${C}_{t}^{l}={f}_{t}^{l}*{C}_{t-1}^{l}+{i}_{t}^{l}*{}_{t}^{l}$$ C t l = f t l * C t − 1 l + i t l * C ~ t l 6 [12pt]{minimal} $${o}_{t}^{l}= ({W}_{o}^{l} [{h}_{t-1}^{l},\, \,{x}_{t}^{l}]+{b}_{o}^{l})$$ o t l = σ W o l ⋅ h t − 1 l , x t l + b o l 7 [12pt]{minimal} $${h}_{t}^{l}={o}_{t}^{l}* ({C}_{t}^{l})$$ h t l = o t l * tanh C t l where [12pt]{minimal} $${f}_{t}^{l}$$ f t l , [12pt]{minimal} $${i}_{t}^{l}$$ i t l , [12pt]{minimal} $${C}_{t}^{l}$$ C t l , [12pt]{minimal} $${h}_{t}^{l}$$ h t l and [12pt]{minimal} $${x}_{t}^{l}$$ x t l denoted the output of forget gate, input gate, cell state, hidden state and the input feature vector of the l -th LSTM layer ( l = 1, 2 or 3) at the t -th time point, respectively. In addition, [12pt]{minimal} $$$$ σ represented the sigmoid function and [12pt]{minimal} $${W}^{l}$$ W l and [12pt]{minimal} $${b}^{l}$$ b l denoted the gates weights and biases of the l -th LSTM layer. A fully connected layer with s output nodes was utilized for predicting the brain state as 8 [12pt]{minimal} $${s}_{t}={softmax}({W}_{s} {h}_{t}^{2}+{b}_{s})$$ s t = s o f t m a x W s ⋅ h t 2 + b s where s was the number of brain state to be predicted, and [12pt]{minimal} $${h}_{t}^{2}$$ h t 2 was the hidden state output of the last LSTM layer which encoded the input functional signature at the t -th time point and the temporal dependency information encoded in the cell state from its preceding time points. Softmax cross-entropy between empirical and predicted brain states was used as the objective function to optimize the LSTM RNNs model. Based on the abovementioned group PCA results, the time series of each principal component (tPCs) were normalized to z scores, and then used as the input of the LSTM RNNs model to predict their corresponding brain states. We chose the first 100 tPCs for characterizing the functional profiles of BOLD signal dynamics along with brain state changes. Firstly, we divided the mouse brain states into 7 prediction categories, including (1) AW only, (2) NREM only, (3) REM only, (4) AW to NREM, (5) NREM to AW, (6) NREM to REM and (7) REM to AW state. Then, we utilized the bootstrap method to avoid the prediction bias from highly unbalanced samples of brain states, e.g., a low percentage of REM state and REM related state transitions. Thus, we randomly selected the BOLD clips (100 tPCs × 120 s) with same sample sizes for 7 categories, which was the minimum number of BOLD clips among 7 prediction categories. The resulting clips were used as the input features of our LSTM RNNs model. Finally, to minimize the prediction bias underlying the sampling procedure, we conducted the bootstrap method for 500 times and repeated the predictions using the correspondingly 500 groups of BOLD clips. For each group of BOLD clips, a 10-fold cross-validation was carried out to improve the robustness of the prediction performance. Particularly, we adopted the “adaptive moment estimation (ADAM)” optimizer with a learning rate of 0.01, which was updated every 1000 training steps with a decay rate of 0.5, and the total number or training steps was set to 8000. Batch size was set to 128 during the training procedure. Parameters including number of hidden layers (1, 2 and 3) and number of nodes in hidden layers (25, 50, …, 200) were selected based on their prediction performance on the validation dataset. The parameter selection was performed on empirical dataset, and the selected parameters were used for the null dataset without further optimization. The null dataset was constructed by randomly shuffling the predicted categories for each group of BOLD clips. Thus, the discriminate time of brain state transition was defined as the farthest time point when prediction accuracy from the empirical dataset was out of the 95% confidence interval (CI) of that from the null dataset. To further reveal which brain region(s) contributed to the prediction most, we carried out a sensitivity analysis based on our 500 times bootstrap sampled dataset. The sensitivity analysis was conducted by evaluating how changes of the 100 tPCs affected the prediction accuracy. Briefly, with the trained LSTM RNNs model remaining unchanged, time courses of 100 PCs were set to 0 one by one from the model input. Changes ( [12pt]{minimal} $${ {ACC}}_{{p}_{0}}$$ △ A C C p 0 ) in the prediction accuracy were then reconstructed by multiplying the spatial weights matrix ( [12pt]{minimal} $${W}_{{p}_{0} v}$$ W p 0 × v ) of group PCA. The resulting spatial weighted accuracy ( [12pt]{minimal} $${ {ACC}}_{{p}_{0}} {W}_{{p}_{0} v}$$ △ A C C p 0 ⋅ W p 0 × v ) was defined as the sensitivities of LSTM RNNs model on brain state prediction. Reporting summary Further information on research design is available in the linked to this article. Male wide type C57BL/6 J mice were obtained from Shanghai Laboratory Animal Center (Shanghai, China) at 8–10 weeks of age, weighted 20–30 g. Animals were group housed (5–6/cage) in the standard laboratory condition (temperature: 23 ± 1 °C; humidity: 50–70%) under a 12 h light/dark cycle (light on from 7 a.m. to 7 p.m.) with food and water ad libitum . All animal experiments were approved by the Animal Care and Use Committee of the Institute of Neuroscience, Chinese Academy of Sciences, Shanghai, China. Two types of MR-compatible electrodes were developed. First, an 18-contacts MR-compatible surface electrode was custom designed and manufactured for recording electrocorticogram (ECoG) and electromyography (EMG) signals (Fig. ). The array was made up of a flexible printed circuit (FPC) and two insulated gold wires, all soldered to an Omnetics 18-pin MR-compatible connector (A70242-001, Omnetics, USA, Minneapolis). The FPC was consisted of a 30 μm thin polyimide film, a 12 μm copper film, and a 3 µm gold-plated layer, with 14 ECoG recording site and two recording reference/ground (100 µm diameter; impedance, ~150 kOhm at 1KHz). Two insulated gold wires were used for EMG recording (50 µm diameter; impedance ~500 kOhm at 1 KHz). Secondly, a MR-compatible depth electrode (Fig. ) was developed with 3 insulated gold wires, 2 bare silver wires and 1 graphene fiber (GF) electrode soldered to a flexible flat cable. One of three insulated gold wires (50 µm diameter; impedance, ~500 kOhm at 1 KHz) was used for recording ECoG, and the rest two for recording EMG. Two bare silver wires (100 µm diameter; impedance, ~60 kOhm at 1 kHz) were used as reference and ground. To record the hippocampal LFP signal, GF electrodes (impedance, ~60 kOhm at 1 kHz) were custom designed. GFs were fabricated through a one-step dimensionally confined hydrothermal process using suspensions of graphene oxide (GO) (monolayer, thickness: 0.8–1.2 nm; sheet diameter: 0.5–5 µm; #XF002-2, Nanjing/Jiangsu XFNANO Materials Technology, China). In a typical preparation, an 8 mg mL −1 aqueous GO suspension was injected into a glass pipeline with a 0.9 mm inner diameter using a syringe. After being baked at 230 °C for 2 h with the two ends of the pipeline sealed, a GF matching the pipe geometry was produced. This preformed GF was then released from the pipeline by flow of N 2 and dried in air. The dried GF had a reduction in diameter to ~75 µm due to water loss and drying-induced alignment of the GO sheets. A GF with diameter of ~75 µm and length of 3 mm was connected to a bare copper wire with diameter of 100 µm using elargol. Parylene-C film of ~5 µm thickness was deposited onto the fibers in a custom made low-pressure coating system to finish the GF electrodes fabrication. Thus, the GF electrodes enabled unbiased fMRI mapping and excellent electrochemical performance, including low impedance and high electrical conductivity, which were not achievable by other electrodes. All surgical procedures were conducted under the standard aseptic condition. Mice were pretreated with 5 mg kg −1 dexamethasone intraperitoneally 1 h before surgery to prevent brain edema. After anesthetized with isoflurane, mice were secured in a stereotaxic apparatus with a heating mat (mouseSTAT, Kent scientific cooperation). A midline sagittal incision was made along the scalp to expose the skull. The periosteum from the skull was removed by saline with cotton-tip applicator. After skull was dried out, a coat of self-etch adhesive (3 M ESPE Adper Easy One) was applied followed by light curing. For MR-compatible ECoG electrode implantation, a sterile dental drill was used to drill off the surface of bone in the shape shown in Fig. (length ~4 mm; width, ~6 mm; centered on bregma), with the bone around raphe retained. The window area was kept wet with saline, and the skull was slowly detached from the dura and raised at about 30° using a nasal stripper and tip tweezers. Gelatin sponge was applied to stop bleeding and keep the dura moist. Afterwards, the FPC part of the ECoG electrode was implanted epidurally with reference and ground sites attached on the bone of raphe. 2% sterilized agarose saline solution was filled between the skull and FPC. After the agarose was solidified, the window area was covered with light curing flowable dental resin to fix the FPC. Later, the two gold wires were implanted into posterior neck muscles for recording EMG. A head holder for awake imaging was then attached on the skull above the cerebellum. Finally, dental cement was applied to smooth the surface of exposed skull. Mice were injected with meloxicam 0.5 mg kg −1 (Baoding sunlight herb medicament, CN) subcutaneously for seven consecutive days post-surgery. For MR-compatible depth electrode implantation, four 0.1 mm diameter holes were drilled in the skull. GF electrode was implanted into the dorsal CA1 region of the left hippocampus at stereotaxic coordinates AP = − 1.94 mm, ML = − 1.7 mm and DV = − 1.15 mm. One gold wire was inserted epidurally at AP = + 1 mm, ML = − 1 mm to record ECoG, and the other two gold wire were implanted into posterior neck muscles to record EMG. Two bare silver wires were used as reference and ground, and inserted epidurally at ML = 0, AP = − 5.5 mm and −6.5 mm, respectively. Light curing flowable dental resin was used for fixation of electrodes and head holder. Other procedures were identical as described above. After seven-day recovery, mice were then habituated for fMRI for another seven days. Mice were head fixed on the animal bed with the recorded acoustic MRI scanning noise based on previous work . Optical imaging studies usually record 3–5 h to obtain mice sleep more efficiently , , as head fixed mice frequently fall asleep after 1.5 h and REM sleep frequently occurs after 2.5 h . Therefore, 4 h head restraining was also utilized in our mouse sleep fMRI. The 30° head holder was designed to fit the natural sleep gesture of mice . The animal bed was modified to allow more space for forelimb movement to reduce stress level . The habituations were all carried out during 9a.m.−15p.m. with a fixed duration of 4 h and gradually increased noise levels. The detailed schedule was listed in Table . No reward was given during or after the habituation training. Neurophysiological signals were fed through a ZIF-CLIP anaglog headstage transmitter, PZ5 amplifier, RZ2 BioAmp processer, and finally recorded by the WS-8 workstation with Synapse software (all from Tucker-Davis Technologies, USA, Alachua). Neurophysiological signals were recorded at a sampling rate of 24414 Hz, high-pass filtered at 0.1 Hz and notch filtered at 50, 100 and 150 Hz, except for LFP signals, which were recorded without notch filter. PZ5 was connected to the external ground (waveguide tube of MRI) for stabilizing the signals. Respiration signal and MRI trigger signal were recorded at a sampling rate of 1024 Hz. All MRI data were acquired with a Bruker BioSpec 9.4 T scanner (Software: ParaVision 6.0.1). An 86 mm volume coil was used for transmission and a single loop mouse head coil (Bruker, 1 cm diameter) was used for receiving. Mouse was head-fixed as described in previous section without using any anesthesia. A T2 weighted RARE anatomical image (TR: 3200 ms; TE: 34 ms; matrix size: 256 × 128; FOV: 18 × 9 mm 2 ; slice thickness: 400 μm; resolution: 70 × 70 μm 2 ) was acquired for coregistration purpose. After local shimming using Mapshim, functional images were acquired using single-shot echo planar imaging (EPI) with the following parameters: TR 2000 ms, TE 14 ms, FA 70°, matrix size 90 × 45, nominal in-plane resolution 200 × 200 μm 2 , slice thickness 400 μm, slices number 22, 7200 EPI volumes (4 h). All 7200 volumes were acquired in a single EPI scan. Triggers were sent from the MRI console for each slice acquisition and recorded along with the electrophysiological signal. The off-line correction of MRI gradient artifacts in the electrophysiological signals was conducted using the fMRI Artifact Slice Template Removal algorithm (FASTR) (Supplementary Fig. ). Briefly, the imaging artifact waveforms were segmented, averaged and iteratively subtracted from the raw electrophysiological signals, according to the concurrently acquired trigger signal from the MRI scanner. This procedure was performed through FMRIB in EEGLAB ( https://fsl.fmrib.ox.ac.uk/eeglab/fmribplugin/ ). After careful visual inspection, we found a few noisy ECoG channels in a small number of scans. Thus, we interpolated the noisy ECoG signal by weighted averaging signals from other good ones, similar to the previous study using the neighbor interpolation method (Supplementary Fig. and Supplementary Data ). Considering there were k ( k < 4) bad channels in 14 channels, the weighted averaging method could be simply formulated as 1 [12pt]{minimal} $${S}^{{k}_{i}}=_{j=0}^{14-k}{({{Dis}}_{{k}_{i}-{good}})}^{ }{ S}_{{good}}^{\,j}$$ S k i = 1 14 − k ∑ j = 0 14 − k D i s k i − g o o d λ ⋅ S g o o d j where [12pt]{minimal} $${S}^{{k}_{i}}$$ S k i , [12pt]{minimal} $${S}_{{good}}$$ S g o o d and [12pt]{minimal} $${{Dis}}_{{k}_{i}-{good}}$$ D i s k i − g o o d were the ECoG signals of the [12pt]{minimal} $${k}_{i}$$ k i th interpolated channel, all good channels and the corresponding Euclidean distance, and [12pt]{minimal} $$$$ λ represented the exponential constraints ( [12pt]{minimal} $$$$ λ <0) using the weighted averaging method. Parameter [12pt]{minimal} $$$$ λ was estimated using other scans without noisy ECoG channels based on the least square fitting. Negative [12pt]{minimal} $$$$ λ indicated farther good channels contributed lower weights on the interpolation of [12pt]{minimal} $${k}_{i}$$ k i th noisy channel. The denoised electrophysiological signals were further down-sampled to 1024 Hz for subsequent analysis. For brain state classification , one channel within each session was selected for further analysis, and the selected channels were listed in the Supplementary Data . Then, we calculated the power spectrum for the ECoG/iEEG and EMG (160–250 Hz) data with 3 s sliding windows and 1 s step size, using the multi-taper method implemented in Chronux ( http://chronux.org/ ). Next, we computed the theta (6–12 Hz) and delta (1–4 Hz) power and theta/delta power ratio, which were further smoothed using “medfilt1” (20 points) in MATLAB. For each session, we used the following criteria for tentatively defining brain states: (1) a time point was classified as NREM sleep if the smoothed delta power was higher than its mean; (2) a time point was assigned as REM sleep if the smoothed theta/delta power ratio was two standard deviations higher than its mean and the EMG power was one standard deviation lower than its mean; and (3) all remaining time points were classified as AW state. Then, we further manually adjusted the classification of brain states as following: (1) For NREM state, according to the ECoG power spectrum, we adjusted the start or end point to the point with the greatest ascent or descent speed of smoothed delta power. For REM state, the start point was adjusted to the end point of the previous NREM, and the end point was adjusted to the point with the greatest descent speed of smoothed theta power. If the greatest ascent or descent point of EMG power was different from that of ECoG/iEEG signals, the midpoint between the two was defined as the transition point; (2) For sessions with noisy EMG recordings, we used the head motion (framewise displacement) estimated from fMRI data as substitute for EMG power; and (3) epochs with short durations (<5 s) were manually merged to the nearest sleep stage. An example of the above classification procedure (session 1: channel 10, 11500–13000 s) was shown in Supplementary Fig. . Spindle and SWRs events were identified in the LFP dataset based on previous studies , . To identify the spindle events, raw iEEG signals in mPFC were bandpass filtered (10–16 Hz) with Butterworth filter. A spindle event was identified if the envelope of the filtered iEEG signal was larger than its mean + 1.5 s.d. in NREM state (Supplementary Fig. ). To identify the SWRs events, raw LFP signals in CA1 of hippocampus were bandpass filtered (120–250 Hz) with Butterworth filter. A SWRs event was detected if (1) the envelope of the filtered LFP signal larger than its mean + 3 s.d. and (2) the power of the filtered LFP signal larger than its temporal mean. The center of spindle or SWRs event was defined as the time of maximum peak of the threshold-passed envelope. The beginning and the end (i.e., event duration) was measured before and after this maximum peak when the amplitude dropped below the mean value of the corresponding envelope (Supplementary Fig. ). Only spindles with 0.4–3 s durations were included. Slow waves were identified based on the procedures described previously . The raw signal was first down-sampled to 1024 kHz. Then, for slow waves detection, mPFC iEEG signals were filtered between 0.3 and 4.5 Hz with a two-order Butterworth bandpass filter. A slow wave was detected in NREM state if the following three criteria were all fulfilled: (1) the interval (T) of negative wave between 0.4 and 2.0 s; (2) top 35% negative amplitude (N) and (3) top 45% negative-to-positive peak-to-peak amplitude (M). Slow wave onset and offset was defined by the time of the first and third zero crossing, respectively (Supplementary Fig. ). All subsequent procedures were performed using custom scripts in MATLAB 2020a (MathWorks, Natick, MA) and SPM12 ( http://www.fil.ion.ucl.ac.uk/spm/ ). The mouse brain was extracted manually using ITK-SNAP ( http://www.itksnap.org/ ). First, each fMRI scan was slice-timing corrected and registered to the scan-specific structural image using rigid body transformation and the scan-specific structure image was then nonlinearly transformed to a study-specific mouse template ( https://atlas.brain-map.org/ ) for group analysis. Then, a light spatial smoothing (0.4 mm isotropic Gaussian kernel) was performed but no band-pass filter was applied to the BOLD time series. Furthermore, to minimize the effects of scanner drift, motion and other non-neural physiological noises, BOLD signals were regressed by “6 rp + 6 Δrp + 40 PCs” nuisance signals , (Supplementary Fig. ). “6 rp + 6 Δrp” nuisance signals represented 6 head motion parameters and their 1st order first derivatives, and “40 PCs” were the first 40 principal components from the BOLD signals of non-brain tissue, e.g., the muscles. The regression-based denoising strategy was adopted from a previous study , in which the PCs estimated from tissues outside the brain were used to model non-neural signal variations and produced a moderate improvement in specificity. The Pearson’s correlation coefficients between the frame-wise displacement (FD) and DVARS (D referring to temporal derivative of time courses, VARS referring to RMS variance over voxels) were calculated to quantitatively reflect the extent to which the motion related signal was reduced by given regressors (Fig. ). GLM based statistical analysis was conducted using the mouse-specific HRF from our previous study , in which NREM and REM states were set as the predictors and thus the AW state was used implicitly as the baseline. Standard first level analysis was done for individual EPI scans. For second level analysis, flexible factorial model (brain states and mouse individuals) was conducted to generate the activation maps with FDR corrected p < 0.05. Furthermore, to investigate the electrophysiological relevance of BOLD signal variations in NREM and REM states, we estimated the Pearson’s correlation coefficients between relative ECoG (or LFP) powers (vs. AW state) and corresponding relative BOLD amplitudes. Each individual EPI scan comprising t time points and v voxels could be represented as a 2-dimensional time-space matrix ( [12pt]{minimal} $${S}_{t v}$$ S t × v ). In the initial scan-level PCA step, fMRI data of each scan was reduced to p components of dimension ( [12pt]{minimal} $${W}_{p v}$$ W p × v , p « t ). In the second PCA step, group PCA was performed on the concatenated data ( [12pt]{minimal} $${W}_{{pN} v}$$ W p N × v ) from all scans, in which data from N scans were stacked along the reduce dimension. Then, the group spatial PCs ( [12pt]{minimal} $${W}_{{p}_{0} v}$$ W p 0 × v ) were back-projected to raw data ( [12pt]{minimal} $${S}_{t v}$$ S t × v ) to reconstruct the time courses of each component (tPCs: [12pt]{minimal} $${S}_{t {p}_{0}}$$ S t × p 0 ) for each individual EPI scan. The long short-term memory (LSTM) recurrent neural networks (RNNs) model was built to predict the brain state based on its functional profile, i.e., tPCs, and their temporal dependency of BOLD signals on its preceding time points. Given the L most recent timesteps [12pt]{minimal} $$\{{X}_{t-L+1},{X}_{t-L+2},,\, {X}_{t}\}$$ X t − L + 1 , X t − L + 2 , … , X t , the goal at timestep t was to predict the state of M timesteps into the future, [12pt]{minimal} $${}_{t+M}$$ X ^ t + M . The architecture of the LSTM RNNs used in this study was illustrated in Fig. , including one (or two or three) hidden LSTM layer(s) and one fully connected layer. Multiple hidden LSTM layers could be used to encode the functional information with temporal dependency for each time point, and the fully connected layer was used to learn a mapping between the learned feature representation and brain states. The functional representation encoded in each LSTM layer was calculated as 2 [12pt]{minimal} $${f}_{t}^{l}= ({W}_{f}^{l} [{h}_{t-1}^{l},\, {x}_{t}^{l}]+{b}_{f}^{l})$$ f t l = σ W f l ⋅ h t − 1 l , x t l + b f l 3 [12pt]{minimal} $${i}_{t}^{l}= ({W}_{i}^{l} [{h}_{t-1}^{l},\, {x}_{t}^{l}]+{b}_{i}^{l})$$ i t l = σ W i l ⋅ h t − 1 l , x t l + b i l 4 [12pt]{minimal} $${}_{t}^{l}= ({W}_{C}^{l} [{h}_{t-1}^{l},\, \,{x}_{t}^{l}]+{b}_{C}^{l})$$ C ~ t l = tanh W C l ⋅ h t − 1 l , x t l + b C l 5 [12pt]{minimal} $${C}_{t}^{l}={f}_{t}^{l}*{C}_{t-1}^{l}+{i}_{t}^{l}*{}_{t}^{l}$$ C t l = f t l * C t − 1 l + i t l * C ~ t l 6 [12pt]{minimal} $${o}_{t}^{l}= ({W}_{o}^{l} [{h}_{t-1}^{l},\, \,{x}_{t}^{l}]+{b}_{o}^{l})$$ o t l = σ W o l ⋅ h t − 1 l , x t l + b o l 7 [12pt]{minimal} $${h}_{t}^{l}={o}_{t}^{l}* ({C}_{t}^{l})$$ h t l = o t l * tanh C t l where [12pt]{minimal} $${f}_{t}^{l}$$ f t l , [12pt]{minimal} $${i}_{t}^{l}$$ i t l , [12pt]{minimal} $${C}_{t}^{l}$$ C t l , [12pt]{minimal} $${h}_{t}^{l}$$ h t l and [12pt]{minimal} $${x}_{t}^{l}$$ x t l denoted the output of forget gate, input gate, cell state, hidden state and the input feature vector of the l -th LSTM layer ( l = 1, 2 or 3) at the t -th time point, respectively. In addition, [12pt]{minimal} $$$$ σ represented the sigmoid function and [12pt]{minimal} $${W}^{l}$$ W l and [12pt]{minimal} $${b}^{l}$$ b l denoted the gates weights and biases of the l -th LSTM layer. A fully connected layer with s output nodes was utilized for predicting the brain state as 8 [12pt]{minimal} $${s}_{t}={softmax}({W}_{s} {h}_{t}^{2}+{b}_{s})$$ s t = s o f t m a x W s ⋅ h t 2 + b s where s was the number of brain state to be predicted, and [12pt]{minimal} $${h}_{t}^{2}$$ h t 2 was the hidden state output of the last LSTM layer which encoded the input functional signature at the t -th time point and the temporal dependency information encoded in the cell state from its preceding time points. Softmax cross-entropy between empirical and predicted brain states was used as the objective function to optimize the LSTM RNNs model. Based on the abovementioned group PCA results, the time series of each principal component (tPCs) were normalized to z scores, and then used as the input of the LSTM RNNs model to predict their corresponding brain states. We chose the first 100 tPCs for characterizing the functional profiles of BOLD signal dynamics along with brain state changes. Firstly, we divided the mouse brain states into 7 prediction categories, including (1) AW only, (2) NREM only, (3) REM only, (4) AW to NREM, (5) NREM to AW, (6) NREM to REM and (7) REM to AW state. Then, we utilized the bootstrap method to avoid the prediction bias from highly unbalanced samples of brain states, e.g., a low percentage of REM state and REM related state transitions. Thus, we randomly selected the BOLD clips (100 tPCs × 120 s) with same sample sizes for 7 categories, which was the minimum number of BOLD clips among 7 prediction categories. The resulting clips were used as the input features of our LSTM RNNs model. Finally, to minimize the prediction bias underlying the sampling procedure, we conducted the bootstrap method for 500 times and repeated the predictions using the correspondingly 500 groups of BOLD clips. For each group of BOLD clips, a 10-fold cross-validation was carried out to improve the robustness of the prediction performance. Particularly, we adopted the “adaptive moment estimation (ADAM)” optimizer with a learning rate of 0.01, which was updated every 1000 training steps with a decay rate of 0.5, and the total number or training steps was set to 8000. Batch size was set to 128 during the training procedure. Parameters including number of hidden layers (1, 2 and 3) and number of nodes in hidden layers (25, 50, …, 200) were selected based on their prediction performance on the validation dataset. The parameter selection was performed on empirical dataset, and the selected parameters were used for the null dataset without further optimization. The null dataset was constructed by randomly shuffling the predicted categories for each group of BOLD clips. Thus, the discriminate time of brain state transition was defined as the farthest time point when prediction accuracy from the empirical dataset was out of the 95% confidence interval (CI) of that from the null dataset. To further reveal which brain region(s) contributed to the prediction most, we carried out a sensitivity analysis based on our 500 times bootstrap sampled dataset. The sensitivity analysis was conducted by evaluating how changes of the 100 tPCs affected the prediction accuracy. Briefly, with the trained LSTM RNNs model remaining unchanged, time courses of 100 PCs were set to 0 one by one from the model input. Changes ( [12pt]{minimal} $${ {ACC}}_{{p}_{0}}$$ △ A C C p 0 ) in the prediction accuracy were then reconstructed by multiplying the spatial weights matrix ( [12pt]{minimal} $${W}_{{p}_{0} v}$$ W p 0 × v ) of group PCA. The resulting spatial weighted accuracy ( [12pt]{minimal} $${ {ACC}}_{{p}_{0}} {W}_{{p}_{0} v}$$ △ A C C p 0 ⋅ W p 0 × v ) was defined as the sensitivities of LSTM RNNs model on brain state prediction. Further information on research design is available in the linked to this article. Supplementary Information Peer Review File Description of Additional Supplementary Files Supplementary Data 1 Supplementary Data 2 Supplementary Data 3 Supplementary Data 4 Supplementary Data 5 Supplementary Data 6 Reporting Summary |
Gender and Racial Representation Trends Among Internal Medicine Department Chairs from 2010–2020 | fe8dd44c-c051-4b85-8f12-951c7b7e1655 | 10039186 | Internal Medicine[mh] | Quality medical education, reduction in health disparities, and healthcare research that includes all members of society are enhanced by diversity in departments of internal medicine (IM). , , , As the largest subspecialty in medicine, IM trains the largest number of medical students and postgraduate trainees and produces a majority of physicians. , Internal medicine (IM), therefore, has a key role in healthcare. However, racial and gender disparities in academic rank and promotion in IM departments persist. While women are nearly equal to men in representation at instructor and assistant professor levels, their representation drops dramatically at higher faculty ranks. , A deeper analysis of IM faculty data reveals more nuanced trends. When parsed out by gender, the percentage of under-represented in medicine (URM) female faculty remains consistently above that of URM male faculty. Such variations in the data lead to Ibrahim’s call for a more granular analysis of the data to highlight areas of disparity that need attention in IM. One area that needs further study is diversity within IM leadership especially at the level of department chairs. Diverse representation in leadership positions can help ensure that minority voices are included in policy decisions, thereby broadening the dialogue around diversity and inclusivity. This then has the potential to disrupt systemic inequities. Most research on increasing diversity within the academic medicine student body or faculty notes the important role of leadership. , , – Yet, there is a lack of in-depth research into diversity in leadership and the need for such diversity. This lack of data negatively impacts structural change which requires a clear picture of the baseline and the desired state. To address the gap in research and identify the baseline data on diversity in leadership at IM departments, we conducted an exploratory study on the racial/ethnic and gender parity representation at IM department chair positions. We analyzed the racial and gender parity between IM faculty and IM department chairs. This study sought to answer the question: What is the level of parity representation, by gender and race, at department chair positions in academic IM departments?
In this study, we examined the data of all racial/ethnic and gender groups. We have used two distinct terms in this manuscript when referring to racial/ethnic and gender groups. Underrepresented in Medicine (URM) : In 2004, the AAMC adopted the term “underrepresented in medicine” to refer to groups whose representation in medical schools falls below their representation in the general population of the USA. For example, Blacks and African Americans constitute approximately 13.4% of the general population of the USA. However, they only make up about 6.1% of medical school matriculants and they are, therefore, categorized as URM. Marginalized : Women and Asian faculty in medical education are no longer considered URM as their percentage representation in medical school faculty is comparable to (women) or more than (Asians) their percentage representation in the general US population. Therefore, we have chosen to use the term “marginalized groups” to refer to women and Asian faculty. The over-representation of Asians in aggregate faculty numbers has moved them into the non-underrepresented category with Whites. The result of this categorization is that there is sparse research into Asians in academic medicine. While their numbers might preclude them from being classified as URM in academic medicine, Asians are viewed as “different” and experience discrimination and bias as other minority groups. Asian physicians experience ethnic and racially offensive remarks from patients and co-workers and have also experienced physical harm. Furthermore, representation of Asian and women faculty drops at higher academic levels and in positions of leadership. Asians and women encounter barriers in their career trajectory and are marginalized as they are excluded from positions of power. In this study, we examined the data for URM and marginalized groups in medical education to provide a comprehensive overview of the current leadership landscape at IM departments in medical schools.
A cross-sectional analysis of race/ethnicity and gender of IM faculty and IM department chairs in IM medical school departments from 2010 to 2020 was conducted. We selected this time frame to provide a sufficient breadth of data to study current trends. Institutional Review Board clearance was not required as only publicly available de-identified data were used in this study. Findings are reported in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies. Data Sets The September 30, 2021, snapshot of the US Medical School Faculty (USMSF) data from the American Association of Medical College’s (AAMC) Faculty Roster was used as the data set. This study used 11 years of data to identify trends in minority representation at IM department chairs levels. Data were extracted from two data sets: (1) the distribution of department chairs by department, sex, and race/ethnicity, and (2) US medical school faculty by gender, race/ethnicity, rank, and department. These data sets represent self-reported data compiled annually by AAMC from all medical schools in the USA. Data relevant to IM were extracted for analysis from these primary data sets. The AAMC data used the term “gender” which was classified as a binary value of women or men . We use this classification in this study. The race/ethnicity categories represented in the data sets were as follows: (1) American Indian or Alaska Native; (2) Asian; (3) Black or African American; (4) Hispanic, Latino, or of Spanish Origin; (5) Native Hawaiian or Other Pacific Islander; (6) White; (7) Other; (8) Multiple Race–Hispanic; (9) Multiple Race–Non-Hispanic; and (10) Unknown Race/Ethnicity. The data for 2010–2020 comprised of 448,986 IM faculty (professors, associate professors, assistant professors, instructors, and others reported as faculty) and 1830 IM department chairs. Methodological Framework Parity indices have been used to calculate the global gender gap, inform legislative initiatives on health equity, corporate leadership gaps, and rank equity in academic medicine. – Parity studies in academic medicine have found that minority faculty are predominantly represented at lower academic ranks (instructor, assistant professor). , Minority representation at the levels of associate and full professor or in leadership (department chairs and deans) are not in parity with their representation at the lower levels. Studies into parity move beyond aggregate numbers to provide more nuanced analysis of data. This study used the Leadership Parity Index (LPI) adapted from the Executive Parity Index (EPI) as the unit of analysis. The EPI was developed in 2015 to assess parity in corporate workforce leadership representation. The EPI has also been adapted to calculate Rank Equity Indices (REI), examining the academic pipeline for faculty in medical schools. , In this study, the Leadership Parity Index (LPI) is calculated as: [12pt]{minimal}
$$ Leadership\ Parity\ Index=^{ }s\ percentage\ representation\ as\ leaders}{\ {The\ group}^{ }\ s\ percentage\ representation\ as\ faculty} $$ Leadership Parity Index = A grou p ′ s percentage representation as leaders The group ′ s percentage representation as faculty Parity in the percentage of leaders and faculty is represented by an LPI of 1.00. Values below 1.00 indicate under-representation, and values over 1.00 indicate over-representation. Studies using parity calculations (gender parity, EPI, REI) comment on over-representation (parity index above 1) and under-representation (parity index below 1) with “1” being seen as the ideal. , , Data Analysis Descriptive statistics for the IM department chairs and faculty were calculated. The proportion of IM department chairs to IM faculty by race/ethnicity for each year (2010–2020) was used to calculate the LPI by race/ethnicity for each of the four race/ethnicity categories in this study. In addition, LPI by gender and gender and race/ethnicity were also calculated for each year.
The September 30, 2021, snapshot of the US Medical School Faculty (USMSF) data from the American Association of Medical College’s (AAMC) Faculty Roster was used as the data set. This study used 11 years of data to identify trends in minority representation at IM department chairs levels. Data were extracted from two data sets: (1) the distribution of department chairs by department, sex, and race/ethnicity, and (2) US medical school faculty by gender, race/ethnicity, rank, and department. These data sets represent self-reported data compiled annually by AAMC from all medical schools in the USA. Data relevant to IM were extracted for analysis from these primary data sets. The AAMC data used the term “gender” which was classified as a binary value of women or men . We use this classification in this study. The race/ethnicity categories represented in the data sets were as follows: (1) American Indian or Alaska Native; (2) Asian; (3) Black or African American; (4) Hispanic, Latino, or of Spanish Origin; (5) Native Hawaiian or Other Pacific Islander; (6) White; (7) Other; (8) Multiple Race–Hispanic; (9) Multiple Race–Non-Hispanic; and (10) Unknown Race/Ethnicity. The data for 2010–2020 comprised of 448,986 IM faculty (professors, associate professors, assistant professors, instructors, and others reported as faculty) and 1830 IM department chairs.
Parity indices have been used to calculate the global gender gap, inform legislative initiatives on health equity, corporate leadership gaps, and rank equity in academic medicine. – Parity studies in academic medicine have found that minority faculty are predominantly represented at lower academic ranks (instructor, assistant professor). , Minority representation at the levels of associate and full professor or in leadership (department chairs and deans) are not in parity with their representation at the lower levels. Studies into parity move beyond aggregate numbers to provide more nuanced analysis of data. This study used the Leadership Parity Index (LPI) adapted from the Executive Parity Index (EPI) as the unit of analysis. The EPI was developed in 2015 to assess parity in corporate workforce leadership representation. The EPI has also been adapted to calculate Rank Equity Indices (REI), examining the academic pipeline for faculty in medical schools. , In this study, the Leadership Parity Index (LPI) is calculated as: [12pt]{minimal}
$$ Leadership\ Parity\ Index=^{ }s\ percentage\ representation\ as\ leaders}{\ {The\ group}^{ }\ s\ percentage\ representation\ as\ faculty} $$ Leadership Parity Index = A grou p ′ s percentage representation as leaders The group ′ s percentage representation as faculty Parity in the percentage of leaders and faculty is represented by an LPI of 1.00. Values below 1.00 indicate under-representation, and values over 1.00 indicate over-representation. Studies using parity calculations (gender parity, EPI, REI) comment on over-representation (parity index above 1) and under-representation (parity index below 1) with “1” being seen as the ideal. , ,
Descriptive statistics for the IM department chairs and faculty were calculated. The proportion of IM department chairs to IM faculty by race/ethnicity for each year (2010–2020) was used to calculate the LPI by race/ethnicity for each of the four race/ethnicity categories in this study. In addition, LPI by gender and gender and race/ethnicity were also calculated for each year.
Demographic Distribution From 2010 to 2020 there were 448,986 IM faculty members and 1830 IM department chairs. White and male faculty were in the majority for both levels (faculty and department chairs). Black or African American faculty and Hispanic, Latino, or of Spanish Origin faculty remain under-represented in academic IM, each making up approximately 4% of the total IM faculty. American Indian or Alaskan Native faculty or Native Hawaiian or Other Pacific Islander faculty are also under-represented constituting only 0.1% of IM faculty. (See Table for details.) Race/Ethnicity LPI From 2010 to 2020, White and Black or African American faculty have achieved leadership parity index of 1 as IM department chairs. While the leadership parity index for Hispanic, Latino, or of Spanish Origin faculty has been moving downwards since 2010, it still remains above 1. Asian faculty, however, have remained under-represented with LPI ranging from 0.17 in 2010 to 0.54 in 2020 (a 46% gap in achieving parity representation). (See Fig. .) From 2010 to 2020, there have been no American Indian or Alaska Native or Native Hawaiian or Other Pacific Islander department chairs. Therefore, data in these two categories have been removed from Figs. , , , and . Data about Other, Multiple Race–Hispanic, Multiple Race–Non-Hispanic, and Unknown Race/Ethnicity have also been excluded to maintain clarity in the representation of ethnicities. Figures with all the groups are available in Supplemental materials . Gender LPI In IM department chairs, from 2010 to 2020, men are consistently over-represented and women, consistently under-represented. LPI for IM women has stayed at or below an LPI of 0.4 while the LPI for men has moved between 1.32 in 2010 and 1.43 in 2020 (Fig. ). Gender and Race LPI Across the four ethnic groups of White, Asian, Black or African American and Hispanic, Latino, or of Spanish Origin, Asian men are the only group under-represented as IM department chairs with LPI ranging from 0.16 (2010) to 0.90 (2020). (See Fig. .) While women faculty as a group are under-represented as IM department chairs (see Fig. ), when studying the data for only women faculty, White and Asian women faculty have been under-represented across the 11 years. Hispanic, Latino, or of Spanish Origin women faculty representation as department chairs has seen more movement beginning with over-representation in 2010 and ending at under-representation in 2020. (See Fig. .)
From 2010 to 2020 there were 448,986 IM faculty members and 1830 IM department chairs. White and male faculty were in the majority for both levels (faculty and department chairs). Black or African American faculty and Hispanic, Latino, or of Spanish Origin faculty remain under-represented in academic IM, each making up approximately 4% of the total IM faculty. American Indian or Alaskan Native faculty or Native Hawaiian or Other Pacific Islander faculty are also under-represented constituting only 0.1% of IM faculty. (See Table for details.)
From 2010 to 2020, White and Black or African American faculty have achieved leadership parity index of 1 as IM department chairs. While the leadership parity index for Hispanic, Latino, or of Spanish Origin faculty has been moving downwards since 2010, it still remains above 1. Asian faculty, however, have remained under-represented with LPI ranging from 0.17 in 2010 to 0.54 in 2020 (a 46% gap in achieving parity representation). (See Fig. .) From 2010 to 2020, there have been no American Indian or Alaska Native or Native Hawaiian or Other Pacific Islander department chairs. Therefore, data in these two categories have been removed from Figs. , , , and . Data about Other, Multiple Race–Hispanic, Multiple Race–Non-Hispanic, and Unknown Race/Ethnicity have also been excluded to maintain clarity in the representation of ethnicities. Figures with all the groups are available in Supplemental materials .
In IM department chairs, from 2010 to 2020, men are consistently over-represented and women, consistently under-represented. LPI for IM women has stayed at or below an LPI of 0.4 while the LPI for men has moved between 1.32 in 2010 and 1.43 in 2020 (Fig. ).
Across the four ethnic groups of White, Asian, Black or African American and Hispanic, Latino, or of Spanish Origin, Asian men are the only group under-represented as IM department chairs with LPI ranging from 0.16 (2010) to 0.90 (2020). (See Fig. .) While women faculty as a group are under-represented as IM department chairs (see Fig. ), when studying the data for only women faculty, White and Asian women faculty have been under-represented across the 11 years. Hispanic, Latino, or of Spanish Origin women faculty representation as department chairs has seen more movement beginning with over-representation in 2010 and ending at under-representation in 2020. (See Fig. .)
This study reveals patterns and potential gaps that exist in parity representation of racial/ethnic and gender groups in positions of IM department chairs at medical schools. When compared to the general population, the under-represented in medicine (URM) faculty remain under-represented both as faculty and department chairs in IM departments. The parity calculations in this study, however, compared percentage representation within IM department faculty and department chairs. These LPI calculations present a different picture. In comparison to their representation within IM faculty, URM faculty in this study are at parity representation as IM department chairs. IM departments seem to be making conscious efforts to diversify representation at leadership positions. Hence, over the last 11 years, Hispanic, Latino, or of Spanish Origin and Black or African American faculty representation in leadership has been at or above parity with their numbers within the faculty. However, it should be noted that almost 16% of Black or African American faculty are at the three historically Black institutions. Therefore, the parity distribution is likely not equally distributed across all academic medicine institutions. The representation of women and Asian faculty in IM leadership is different. While women are almost equal to men in aggregate numbers in IM faculty, they fall below parity representation at department chair positions. Over the past 11 years, the LPI for women faculty has not moved beyond 0.4, indicating a 60% gap in achieving parity in department chair representation. In 2020, Asian faculty made up about 25% of the IM faculty but only 8% of IM department chairs. Over the 11 years studied, Asian faculty have consistently fallen below parity representation in leadership. When parsed out by gender, both Asian men and Asian women faculty are below parity compared to their representation within the IM faculty. But the representation of Asian men faculty as IM department chairs has been moving towards parity from 0.16 in 2010 to 0.9 in 2020. However, Asian women faculty have never exceeded a parity of index of 0.17 which was in 2010. These findings show that discrimination can occur in different ways and representation in aggregate numbers does not tell the whole story. , As Wesson et al. note, “Discrimination is often subtle but pervasive. It often appears as unrecognized assumptions and attitudes that work systematically against minorities and women.” The under-representation of Asian and women faculty in IM leadership is concerning given the ramifications. Limiting access to leadership can be a manifestation of discrimination as certain voices are excluded. The practical implications of this are that not all perspectives are equally considered in decision making processes and the process itself begins to lack credibility. This also engenders non-inclusive learning environments. Diversity in faculty and leadership is a visible demonstration of an institution’s commitment to diversity and a testament to what is possible. The lack of parity in representation for women and Asian faculty in IM department chair positions conveys a concerning message that while diversity is encouraged, there may be a ceiling on the achievement of certain groups. There is an implicit institutional message that it is not possible for members of these groups to move into IM department chair positions and the gap between espoused values and actual practice is revealed. The terminology of “underrepresented in medicine” has enabled institutions to develop focused programs to address issues that are specific to this population. Despite these initiatives, there have been no American Indian or Alaska Native and Native Hawaiian or Other Pacific Islander department chairs from 2010 to 2020. The diversity initiatives to draw more URM into academic medicine need to be more robust. The unintended consequence of the term URM has been that Asian faculty are now placed in the same category as the majority whites. Asians are seen as the “model minority” who are white adjacent and “WASPs with brown skin.” . Yet, their experiences of discrimination show that they do not share the privilege of the majority. Asian faculty find themselves de-minoritized and occupying a liminal space neither URM nor white. The findings from this study highlight the disparities that can creep in when the primary focus is placed on representation in terms of aggregate numbers. Representation in aggregate numbers can lead to a sense of complacency wherein more subtle forms of discrimination are overlooked. This study also calls into question the use of broad classifications such as URM or Asian. Asian is an umbrella term representing diverse communities from the affluent Indians and Chinese to the poorest populations in the USA (Hmong, Vietnamese, and Cambodian). Asian groups such as the Hmong are far from being over-represented in medicine. Yet, their situation is completely overlooked. Broad classifications such as URM or non-URM can hide discrimination that is experienced and manifested in ways that are specific to each sub-group. Calculations such as the leadership parity index (LPI) used in this study can help provide a nuanced picture of the baseline of the leadership landscape in IM departments. The LPI could help identify gaps in the IM leadership pipeline by revealing groups that are being marginalized within departments and guide interventions to increase the diversity of department chairs.
This study is based on datasets obtained from the AAMC faculty roster and shares the limitations of the dataset. The data set only provided IM leadership in the role of department chair. Therefore, other leadership roles such as associate deans, program/course directors, etc. could not be explored. But the findings of this study begin a conversation about the lack of parity representation at the level of department chairs. Furthermore, the datasets only identified “men” and “women” and did not allow for a deeper analysis of other gender identities. This study highlighted the consequences of using broad categorizations and the gender categorization in this dataset could be obscuring other patterns that need to be studied. We recognize that there might be variation in gender and racial representation of department chairs based on regional location. However, the dataset we used provided only aggregate information and did not parse out the data based on geographical regions. Since this is an initial exploratory study into the trends in leadership representation, we decided to use this data. Future studies could replicate this study across different regions to identify other trends. The lack of leadership representation could be due to various reasons including the fact that not all faculty might want or accept leadership positions. Further research is needed to examine why Asians and women are not in leadership positions. This study used the parity index calculation which is fairly new in the medical education literature. Therefore, questions such as acceptable levels of variation from a parity of 1 have yet to be explored.
Studying representation in IM leadership through a parity lens provides another perspective on diversity in medical education. Women and Asians encounter various challenges in their progression to leadership positions which have been termed the “glass ceiling” and “bamboo ceiling” respectively. This study shows that a ceiling effect may also be at play in IM departments. The Hispanic, Latino, or of Spanish Origin and Black or African American URM faculty groups have attained parity representation in department chairs at IM departments. While IM departments are showing movement in a positive direction, this study shows that we cannot become complacent in our efforts at diversity. Issues of discrimination play out in different ways, and it is important for us to remain vigilant and work at uncovering hidden biases.
ESM 1 (PNG 135 kb) ESM 2 (PNG 128 kb) ESM 3 (PNG 133 kb)
|
Study on the effectiveness and safety of ciprofol in anesthesia in gynecological day surgery: a randomized double-blind controlled study | b31438ad-057a-42ea-a7a4-0bfab92d2e3a | 10039513 | Gynaecology[mh] | Day surgery has become popular in different countries because of its advantages in reducing hospitalization time, improving bed utilization and reducing hospitalization costs. With the development of medical technology, the proportion of day surgery is growing rapidly. Guidelines from the Association of Anaesthetists and the British Association of day surgery argue that day surgery anesthesia should be selected with the goal of minimizing patient stress and optimizing comfort. Therefore, general anesthesia has become the most commonly used anesthesia method for ambulatory surgery. Day surgery requires anesthetic drugs with rapid onset, rapid elimination, short duration of action, good sedative and analgesic effects, minimal effects on cardiopulmonary function, and no serious adverse effects or discomfort. Propofol is a widely used intravenous anesthetic with the advantages of rapid onset, rapid recovery and no accumulation.Due to its disadvantages such as dose-dependent blood pressure reduction and injection pain, its application in the elderly, circulatory dysfunction, etc. is limited.Injection pain is one of the most common adverse events of propofol, the incidence of Injection pain in adults is 28-90%, Ciprofol is a new type of intravenous anesthetic developed independently in China, which is a short-acting γ-aminobutyric acid A receptor (γ-aminobutyric acid subtype A receptor (GABAA) agonist,and has now completed a Phase 3 clinical trial. In the previous experiment , it was proved that ciprofol has the characteristics of fast onset of action, rapid recovery, no accumulation, less pain and small respiratory depression after injection, which has potential clinical application value. However, there is still little experience in the application of ciprofol in clinical practice,,More trials are needed to analyze the safety and efficacy of ciprofol. This study intend to use a randomized double-blind control to explore the safety and efficacy of ciprofol by comparing the adverse events and anesthesia effects between ciprofol and propofol in gynecological day surgery, in order to provide reference for clinical application. Patients and study protocol The study was conducted at Weifang People’s Hospital, has been approved by the hospital ethics committee(2,021,037). This study was registered with the Chinese Clinical Trial Registry (ChiCTR2100053444) in 21/11/2021 and informed consent was signed by patients or their legal guardian. This study was a randomized double-blind controlled study,and the primary endpoint was the overall incidence of adverse events.Patients (18 ~ 64years),with American Society of Anesthesiologists physical classification status I or II, BMI between 18 and 28 kg/m 2 , who were about to undergo gynaecological ambulatory surgery from January 2022 to June 2022 were eligible.Patients were excluded if they sufered from egg/soy/propofol allergies,significant cardiovascular, respiratory or hepatic and renal diseases in this study.In addition, women who were pregnant, or planning to become pregnant were excluded.An independent investigator used random number table to assign patients to the ciprofol group and the propofol group, with 64 patients in each group.The distribution list is placed in an envelope that is opened by the nurse anesthetist on the day of surgery to prepare study medications in the anesthesia preparation room. The nurse anesthetist is not directly involved in patient care. Anesthesiologist performed anesthesia without knowing the grouping. Participants and outcome assessors were blinded to group allocation. The patient had no pre-anesthetic medication.Following arrival in the operating room, patients were monitored via electrocardiography, pulse oximetry, bispectral index(BIS,Mindray), and continuous noninvasive blood pressure and established intravenous access of the upper limb.Intravenous flurbiprofen axetil (50 mg), dexamethasone (5 mg) were used to start general anesthesia induction, followed by pump injection with Medical Syringe Pump(Silugao) for 60 s of ciprofol (0.5 mg/kg; Liaoning Haisco Pharmaceutical Co.Ltd,National Drug Administration (NDA)H20200013)or propofol (2 mg/kg; Fresenius Kabi AB,National Drug Administration (NDA)HJ20170305). The time of loss of consciousness from the beginning of study drug administration were assessed every 5 s by calling for eye opening or by mild prodding or shaking.Once the patient reached the modified observer’s assessment of alertness/ sedation (MOAA/S) ≤ 1 (no response after mild prodding or shaking), 0.2 mg/kg mivacurium chloride and 20ug/kg alfentanil were administered immediately. After the spontaneous breathing disappears, oxygen was administered under face mask pressure, and after the skeletal muscles were relaxed, the glottis was exposed with a visual laryngoscope and 2% lidocaine(3 ml) was sprayed into the subglottis using Single-use ENT anesthetic nebulizer.Then a tracheal catheter was inserted and properly fixed. Mechanical ventilation was performed with parameters set at 6 ~ 8 ml / kg for V T ,60% for FiO2, 12 ~ 16 times / min for RR, 1: 2 for I:E. If the patient failed to achieve MOAA/S ≤ 1 within 1 min after the full induction dose was administered, one-half of the initial dose was given.If the patient failed to achieve MOAA/S ≤ 1 was not reached within 2 min, it would be regarded as a failure of general anesthesia induction of the drug in this study.Anesthesia was maintained with ciprofol (1 mg/kg/h) or propofol(5 mg/kg/h) and alfentanil (40 ug/kg/h). During the operation, when blood pressure or heart rate rose to 20% of the basal value,study druy( 0.05 ml/kg) was given; when the blood pressure dropped by 30% of the basal value, ephedrine (6 mg) was given; when the heart rate was less than 50 beats/min, atropine(0.3 mg) is given. Stop the infusion of all medications when the surgical operation is stopped.After the endotracheal tube was removed, patient was transferred to the postanesthesia care unit (PACU). Primary outcomes : The overall incidence of adverse events. Overall adverse event is defined as an event that occurs in the perioperative period that affects the safety of anesthesia.Overall adverse events included: (1) bradycardia (HR < 50 beats/min, > 30s); (2) Tachycardia (HR > 100 beats/min, > 30s) (3) Hypotension (30% reduction in SBP compared to baseline value); (4) Hypertension (SBP is 20% higher than baseline value); (5) injection pain(We asked patients if they feel pain in the arm when the drug was injected); (6) Intraoperative body movements(The patient had no conscious movement of the limbs). Secondary outcomes included (1) success rate of induction of anesthesia,(2) the time of loss of consciousness (time of initiation of study drug infusion to MOAA/S ≤ 1), (3) time of awakening (time of drug discontinuation to extubation), (4) study drug top-up doses, (5)rescue drug use. The study was conducted at Weifang People’s Hospital, has been approved by the hospital ethics committee(2,021,037). This study was registered with the Chinese Clinical Trial Registry (ChiCTR2100053444) in 21/11/2021 and informed consent was signed by patients or their legal guardian. This study was a randomized double-blind controlled study,and the primary endpoint was the overall incidence of adverse events.Patients (18 ~ 64years),with American Society of Anesthesiologists physical classification status I or II, BMI between 18 and 28 kg/m 2 , who were about to undergo gynaecological ambulatory surgery from January 2022 to June 2022 were eligible.Patients were excluded if they sufered from egg/soy/propofol allergies,significant cardiovascular, respiratory or hepatic and renal diseases in this study.In addition, women who were pregnant, or planning to become pregnant were excluded.An independent investigator used random number table to assign patients to the ciprofol group and the propofol group, with 64 patients in each group.The distribution list is placed in an envelope that is opened by the nurse anesthetist on the day of surgery to prepare study medications in the anesthesia preparation room. The nurse anesthetist is not directly involved in patient care. Anesthesiologist performed anesthesia without knowing the grouping. Participants and outcome assessors were blinded to group allocation. The patient had no pre-anesthetic medication.Following arrival in the operating room, patients were monitored via electrocardiography, pulse oximetry, bispectral index(BIS,Mindray), and continuous noninvasive blood pressure and established intravenous access of the upper limb.Intravenous flurbiprofen axetil (50 mg), dexamethasone (5 mg) were used to start general anesthesia induction, followed by pump injection with Medical Syringe Pump(Silugao) for 60 s of ciprofol (0.5 mg/kg; Liaoning Haisco Pharmaceutical Co.Ltd,National Drug Administration (NDA)H20200013)or propofol (2 mg/kg; Fresenius Kabi AB,National Drug Administration (NDA)HJ20170305). The time of loss of consciousness from the beginning of study drug administration were assessed every 5 s by calling for eye opening or by mild prodding or shaking.Once the patient reached the modified observer’s assessment of alertness/ sedation (MOAA/S) ≤ 1 (no response after mild prodding or shaking), 0.2 mg/kg mivacurium chloride and 20ug/kg alfentanil were administered immediately. After the spontaneous breathing disappears, oxygen was administered under face mask pressure, and after the skeletal muscles were relaxed, the glottis was exposed with a visual laryngoscope and 2% lidocaine(3 ml) was sprayed into the subglottis using Single-use ENT anesthetic nebulizer.Then a tracheal catheter was inserted and properly fixed. Mechanical ventilation was performed with parameters set at 6 ~ 8 ml / kg for V T ,60% for FiO2, 12 ~ 16 times / min for RR, 1: 2 for I:E. If the patient failed to achieve MOAA/S ≤ 1 within 1 min after the full induction dose was administered, one-half of the initial dose was given.If the patient failed to achieve MOAA/S ≤ 1 was not reached within 2 min, it would be regarded as a failure of general anesthesia induction of the drug in this study.Anesthesia was maintained with ciprofol (1 mg/kg/h) or propofol(5 mg/kg/h) and alfentanil (40 ug/kg/h). During the operation, when blood pressure or heart rate rose to 20% of the basal value,study druy( 0.05 ml/kg) was given; when the blood pressure dropped by 30% of the basal value, ephedrine (6 mg) was given; when the heart rate was less than 50 beats/min, atropine(0.3 mg) is given. Stop the infusion of all medications when the surgical operation is stopped.After the endotracheal tube was removed, patient was transferred to the postanesthesia care unit (PACU). Primary outcomes : The overall incidence of adverse events. Overall adverse event is defined as an event that occurs in the perioperative period that affects the safety of anesthesia.Overall adverse events included: (1) bradycardia (HR < 50 beats/min, > 30s); (2) Tachycardia (HR > 100 beats/min, > 30s) (3) Hypotension (30% reduction in SBP compared to baseline value); (4) Hypertension (SBP is 20% higher than baseline value); (5) injection pain(We asked patients if they feel pain in the arm when the drug was injected); (6) Intraoperative body movements(The patient had no conscious movement of the limbs). Secondary outcomes included (1) success rate of induction of anesthesia,(2) the time of loss of consciousness (time of initiation of study drug infusion to MOAA/S ≤ 1), (3) time of awakening (time of drug discontinuation to extubation), (4) study drug top-up doses, (5)rescue drug use. (1) success rate of induction of anesthesia,(2) the time of loss of consciousness (time of initiation of study drug infusion to MOAA/S ≤ 1), (3) time of awakening (time of drug discontinuation to extubation), (4) study drug top-up doses, (5)rescue drug use. In type I error 0.05(bilateral), Power of test is 80%.The overall incidence of adverse events was approximately 36.4% in the ciprofol group, and 60.6% in the propofol group.Finally a total of 128 patients were includeda in this study. Using SPSS 25. 0 Statistical software for data analysis. normal distribution measurement data is expressed as mean ± standard deviation (x ± s), with two independent samples t-test used for inter-group comparisons; The enumeration data is represented by the example method, the inter-group comparison is represented by the χ 2 test or the Fisher exact method. P<0. 05 indicates that the difference is statistically significant. 128 patients were included in this study. Data from 128 patients were obtained for statistical analysis (Fig. ). There were no significant differences between the ciprofol and propofol groups with respect to patient age, weight, height, BMI, ASA physical status, type of surgery and anesthesia time (Table ). A total of 137 adverse events occurred in 128 patients, of which 44 adverse events occurred in 64 patients in the ciprofol group and 93 adverse events occurred in 64 patients in the propofol group. Among those who had 2 or more adverse events in one person, 11% were in the ciprofol group and 48.5% were in the propofol group. The overall incidence of adverse events was significantly lower in the ciprofol group compared with the propofol group (56.2% vs. 92.2%,P < 0.05). The incidence of hypotension was the highest, accounting for 44.5% of the total adverse events. Those who required intraoperative ephedrine to boost blood pressure were 5 cases in the ciprofol group and 9 cases in the propofol group, of which 2 patients in the propofol group used it twice and the rest used used it once, with no statistically significant difference between the two groups (P > 0.05), as shown in Table . The success rate of induction was 100% in both groups, and the comparison between the two groups was not statistically significant (P > 0.05).The time of loss of consciousness in the ciprofol group was longer than that in the propofol group (1.6 ± 0.4 min vs. 1.4 ± 0.2 min), which was statistically significant (P < 0.05). The time of awakening was not statistically significant in the two groups(P > 0.05).Top-up dose during the operation was not statistically significant in the two groups (P > 0.05), (Table ). The intraoperative blood pressure, heart rate, and BIS trends were similar, with a transient decrease in blood pressure after injection, followed by an increase in blood pressure and tending to stabilize. Compared with T1, SBP、HR were lower at T2 ~ T9 and the difference was statistically significant.The difference between SBP at T 1 and at T 3 ~ T 6 in the ciprofol group was lower than that in the propofol group(P < 0.05).The difference between BIS at T 1 and at T4、T 6 ~ T 8 in the ciprofol group was larger than that in the propofol group(P < 0.05)(Table 、Fig. ). Figure Changes in blood pressure, heart rate, and BIS at different time points during the patient’s perioperative period. Ciprofol is a new type of intravenous anesthetic independently developed in China, its chemical name is 2-[(1R)-1-cyclopropylethyl]-6-isopropylphenol, which is an isomer of propofol, and the ciprofol group is introduced into the chemical structure of propofol, forming a chiral structure and increasing the stereo effect. By enhancing GABAA receptor-mediated ion channels, it influxes chloride ions, causing hyperpolarization of nerve cell membranes to achieve central nervous system suppression . Because of its low injection pain and light respiratory depression, it has attracted much attention since its listing. In previous clinical trials, ciprofol has been shown to have similar safety and tolerability to propofol during induction and maintenance of anaesthesia, and ciprofol has potential clinical application value. Teng Y et al. in the IIa and IIb study of ciprofol concluded that 0.4–0.5 mg/kg ciprofol for colonoscopy had equivalent anesthesia to 2.0 mg/kg propofol and had a similar safety profile with no serious adverse events. This trial investigated the safety of higher doses of ciprofol in gynecologic ambulatory surgery by comparing 0.5 mg/kg ciprofol with 2 mg/kg propofol. In this study, the success rate of induction was 100% in both groups, a result that suggests that ciprofol has good anesthetic efficacy when applied to gynecologic ambulatory surgery. In a phase III clinical trial of gastroscopy , the induction time was 1.1 ± 0.5 min in the ciprofol group and 1.1 ± 0.4 min in the propofol group (P = 0.405), and the mean time to complete awakening was 3.3 ± 3.1 min in the ciprofol group compared with 2.0 ± 2.1 min in the propofol group (P < 0.05). In a phase 3 multicenter study of elective surgery, the time to successful induction was 0.91 ± 0.03 min in the ciprofol group and 0.80 ± 0.03 min in the propofol group (P < 0.05),and the time to disappearance of the eyelash reflex was 0.80 ± 0.03 min and 0.71 ± 0.03 min(P < 0.05 ). The time of loss of consciousness in this experiment was 1.6 ± 0.4 min in the ciprofol group and 1.4 ± 0.2 min in the propofol group (P < 0.05), which was longer than in the previous study and may be related to the speed of drug injection. In this study, a syringe pump was used to limit the drug infusion to 1 min to reduce the effect of infusion speed on drug onset time.The longer induction time of ciprofol than propofol in this study may be related to the relatively lower lipophilicity of ciprofol due to the introduction of the cyclopropyl structure, which affects the type of formulation (e.g., lower oil content) and reduces the free ciprofol concentration and the rate of ciprofol crossing the blood-brain barrier, etc. In a study of ciprofol used in gynecological surgery, the incidence of adverse events was significantly reduced in the ciprofol group, (20% vs. 48.33%, P = 0.0019), not including injection pain. However, this study included injection pain in the observation of adverse events, so the overall incidence of adverse events was high(56.2% vs. 92.2%,P < 0.05), The incidence of body movement was higher in the ciprofol group than in the propofol group, and the difference was statistically significant. Whether intraoperative body movements are related to our adoption of a shallow depth of anesthesia with a sudden increase in surgical stimulation remains to be observed in further studies. Injection pain is one of the most common adverse effects in propofol anesthesia.There are many factors influencing injection pain, including injection site and injection speed, venous size, etc. In order to improve patient comfort and reduce patient pain, clinically, anesthesiologists seek different ways to alleviate propofol injection pain, including drug interventions (lidocaine, opioids, dexmedetomidine, propofol medium and long chain fat emulsion injection, etc.), physical interventions (selection of coarser blood vessels, dilution of propofol, low-dose desensitization, etc.),but the effect was not good. In this study, the incidence of ciprofol injection pain was significantly lower than that of the propofol group (1.6% vs. 76.6%), Ciprofol is an isomer of propofol, and the cyclopropyl group is introduced into the chemical structure of propofol, which improves the pharmacological and physicochemical properties, eliciting less pain on injection. Hypotension is also a common adverse effect of propofol. Systolic blood pressure < 90 mmHg is the threshold for associated myocardial and renal injury, and a brief (> 5 min) reduction in systolic blood pressure by 41 ~ 50 mmHg from baseline increases the incidence of myocardial infarction by a factor of 3. Moreover, MAP < 80 mmHg for more than 10 min increases mortality in patients, and the longer the time and the lower the MAP, the greater the risk , so the inhibition of propofol for circulation is one of the reasons why its use in anesthesia is limited. In this study, the blood pressure in both groups decreased with a similar trend, which mostly decreased within 3 min after drug administration, and then gradually stabilized. Compared with T1, SBP were lower at T2 ~ T9 and the difference between SBP at T 1 and at T 3 ~ T 6 in the ciprofol group was lower than that in the propofol group,which indicates that ciprofol is more beneficial for hemodynamic stability of patients. Adverse events such as hypotension that occurred in both groups recovered on after administration of small amounts of cardiovascular active drugs or within a short period of time without serious adverse consequences. The present trial had several limitations. First, this study used a cuff for noninvasive blood pressure testing. There is a delay in the observation of blood pressure changes and it is unknown if more severe blood pressure changes have occurred. But performing invasive blood pressure monitoring for short procedures is unnecessarily traumatic for the patient. Second, this trial was conducted in patients with ASA I or II, and further studies are needed in Elderly, frail and seriously ill patients. Third, due to the short duration of the procedure, the use of inotropic drugs may have an impact on the time to awakening. Overall, the results of this study suggest that ciprofol is as effective as propofol in anesthesia in gynecological ambulatory surgery, while having a lower incidence of adverse events. |
The epidemiological and medico-legal characteristics of violent deaths and spousal homicides through a population of women autopsied within the Forensic Medicine Department of the University Hospital of Annaba | 5814cf1c-2934-4d79-bb9f-293f17129931 | 10039587 | Forensic Medicine[mh] | The mortality of women, and more particularly that which occurs as a result of domestic violence, represents a real public health problem with lethal consequences. The international literature on spousal homicide is poor and the references are limited. Nevertheless, some studies have attempted to draw up the profile of the victims or that of the perpetrators, especially from a criminological and/or psychopathological point of view. According to the latest estimates from the World Health Organization (WHO), the intimate partner is the perpetrator of the homicide in approximately 38% of homicides committed against women . According to official statistics from the Government of Canada, during the year of 2003, seventy-eight (78) people were killed by a spouse, including 64 women and 14 men. This homicide rate fell slightly in 2003 compared to 2002, when 83 victims were counted . In Italy, the incidence of femicides is increasing more and more in recent years and it accounted for 30.9% of all homicides in 2011 . An Italian study assessed the characteristics of female homicides and femicides based on autopsy reports of female homicides that occurred in the judicial district of Bologna, over a period of 70 years, on a sample of 172 female homicides, including 103 femicides. This study concluded that there was a statistically significant association between homicide and the victim-offender relationship and that over time the main age of victims increased . Another study from Italy examined all cases of female homicides using autopsy reports at the Institute of Forensic Medicine in Milan over twenty years (1999–2019). Two hundred (200) murders of women were noted, among more than 15,000 autopsies performed, and 535 homicides were recorded at the Institute of Forensic Medicine in Milan over the period considered, with an average of 9.5 femicides per year . A Turkish study, in Eskişehir province, investigated the cases of women killed by their partners (husband, boyfriend, ex-husband, or ex-boyfriend) over 25 years (1992 and 2016), through the data collected from autopsy and toxicology reports. According to the results of this study, 148 women were victims of femicide over 25 years, and the number of cases continues to increase every year . Another Turkish study was conducted in Istanbul, on a sample of 20,486 forensic autopsies carried out for five years (2006, 2010). Of these deaths, 537 were related to female violence and 12.9% of female deaths were due to intimate partner violence, which accounted for 2.6% of all autopsies performed . A third Turkish study accounted 162 cases of femicides that occurred in 12 cities in Turkey over ten years (January 1st, 2000, to December 31st, 2010). According to this study, 80 women were killed by their intimate partner, and 81 women were killed by one of their relatives (friends or sexual partners) . In the United Kingdom, 37% of all women were murdered by their current or former intimate partner compared to 6% of men . This very question of spousal homicide arose the interest of the team of the medico-legal institute of Paris, which tried to highlight the seriousness of this phenomenon. A study was conducted to evaluate the frequency of spousal homicides on a representative sample of 652 deaths over a period of seven years. This study showed that 31% of female homicides were committed by intimate partners, 20% by sexual partners, and 15% by individuals unknown to the victims . A French team from the Forensic Institute of Tours was also interested in this phenomenon of spousal homicide and carried out a study for four years (2000 -2003). This study analyzed two years of autopsies of spousal homicides and showed that frequently there was a link between the victim and the perpetrator of the homicide. Indeed, in 50% of homicides, the victims knew their assailants; in 25% of cases, the aggressor was a family member, and in 17% of the cases, it was spousal homicide . In France, a frequency of 10% of male deaths and 7% of female deaths were attributed exclusively to violent acts. Violence occupies the third etiology of mortality after cardio-circulatory diseases and tumors. Domestic violence is considered to be one of the main causes of female mortality . The World Health Organization (WHO) recently released its report on violence-related mortality data. No less than 38% of women who were murdered were killed by their intimate partners, and 42% of these women received physical injuries or sexual violence from a partner . It is important to note that we do not have enough data on spousal homicides in Algeria and its neighboring countries. The objective of this study is to determine the epidemiological and medico-legal characteristics of the violent deaths of women, and more particularly, of spousal homicide.
This study was based on a double survey, initially, retrospective, for two years (2017, 2018) and secondarily perspective, also for two years (2019, 2020), relating respectively to the violent mortality of women and intimate partner homicide. The initial, retrospective, survey focused on women who died in a violent context without considering the circumstances of the occurrence and regardless of the medico-legal forms, during the period between January 1st, 2017, and December 31st, 2018. The second, prospective, survey concerned exclusively women who died as a result of violence caused by an intimate partner and who underwent a forensic autopsy during the period between January 1st, 2019, and December 31st, 2020. During this double survey, all women aged 18 and over (the most prominent age group in Algeria) who died in a context of violence and on whom a medico-judicial autopsy was performed, were included. Our investigation took place within the thanatology unit of the Forensic Medicine department at IBN ROCHD Hospital and University Center, taking charge of autopsy activities and post-mortem explorations while seeking justice. Our protocol included a detailed external examination to determine a complete lesion assessment by identifying the type of lesions, and their locations, and evoking the vulnerable instruments used to cause these injuries, as well as the medico-legal reasoning in terms of imputations and causes of death. The internal examination made it possible to highlight the complete lesion assessment of the organs and viscera affected, as well as the fatal injuries (number, depth, and cause of death). Our protocol ends with toxicological and histological examinations according to the medico-legal circumstances and the legal requirements. For the data collection of our study (sociodemographic parameters, forensic characteristics, and data related to the circumstances of death), we based investigation on the questioning of the women's relatives and those around them, reports from the police and national gendarmerie services, as well as judicial files carried out by the competent court. The data obtained were processed using the Epi info6 software. All measures and provisions related to the principles applied to the ethics of research on human beings, namely the confidentiality of the information obtained and the informed consent of the families and beneficiaries of the participants in the double survey, were rigorously respected and in full compliance with the Helsinki Biomedical Research Ethics Guide). The Ethics and Research Committee in Health Sciences of the IBN ROCHD University Hospital gave a favorable opinion to this double survey on the violent mortality of women and spousal homicide by approving the research protocol and emphasizing the confidentiality of the collected data. Sheets in the triple language (Arabic, French, and English), explaining the objectives and framework of the study, were prepared beforehand and presented to families, relatives, and beneficiaries before each collection of information.
The retrospective survey During the two years of the retrospective study (2017 and 2018), women who underwent a forensic autopsy for reasons of violent death represented a frequency of (5.71%): 35 women victims of violent death out of a total number of 570 autopsies performed. All age groups were concerned, but to a greater extent young women, particularly those aged between 31 and 40 (31.42%). Figure shows the respective frequencies of the age groups. Married women were the most affected. Indeed, nearly three-quarters of the victims are legally married (71.42%). The remaining proportion was shared between single women (14.28%), widowed women (11.42%), and divorced women (2.89%). The overwhelming majority of female victims of violent death (93%) were inactive at the time of the murderous act. Eighty-eight point fifty-seven percent of them had no profession, and 5.71% were retired. A frequency of 82.85% died in the marital home and at the hospital following complications from the initial trauma. Figure lists the sites where violent deaths occur among women. Homicide is the most represented forensic form, with a frequency of 51.42%, followed by suicide with a frequency of around 40%. Figure shows the respective frequencies of the three forensic forms of violent female death. In terms of injury tools of homicide victims, blows with blunt objects represent the most frequent operative mechanism (44.44%), followed by injuries caused by sharp instruments (39%) while the use of firearms was relatively lower (17%) (Fig. ). During suicides, hanging is the mostly used method by the victims (42.85%). However, the use of other processes, namely jumping off a high place (28.57%) was added to the ingestion of caustic products (21.42%) and drugs (7.14%). In homicides, the spouse is the perpetrator in 61.11% of cases. The perpetrator is unknown to the victim in 33.33% of the cases. Figure shows the link between the perpetrator and their victims. Forensic examination showed the following frequencies: Non-specific blunt wounds (47%), simple wounds (26%), and blunt gunshot wounds (6%) (Fig. ). Thoracic and abdominal lesions were the most frequently observed injuries in victims (34%). Figure shows the different lesions of internal violence highlighted by the forensic autopsy. The cephalic extremity was the region of predilection for this type of fatal trauma (51.42%), followed by the category “ multiple anatomical regions of the body ” where several body segments were affected (28.57%). The other traumatic locations were indicated in detail in Fig. . The use of complementary investigations represented an essential tool in the identification of the violent causes of death and in particular those secondary to intoxication. However, the results remain highly dependent on several parameters (date of the autopsy, date of sampling, nature of the sample and its storage, cause of death, etc.). Toxicological research was almost systematic among victims who died following a homicide: (N: 14/18: 77.77%) of all homicides and (N14/35: 40%) of all victims of violent deaths. The recourse to histological examination was rare. Thus, according to our results, we note the following frequencies: (N: 6/18), a frequency of 33.33% of all homicides, and (N: 6 /35), representing a frequency of 17.14% of all violent deaths. Table illustrates the different examinations carried out during autopsies of victims of violent deaths. The prospective survey During the prospective survey (2019—2020), 12 spousal homicides were reported out of a total number of 670 forensic autopsies carried out during the aforementioned period, a frequency of nearly two percent (1.79%) of all medical-legal deaths. The sample studied had an average age of 33 ± 12.91 years, with extremes of 19 to 56 years. The age group between 21 and 30 is the most represented one, with a frequency of (50, 33%), followed by those aged over 50 years old, with a frequency of (25%). The perpetrators of spousal homicides had an average age slightly higher than that of the victims: about 42 ± 10.73 years with extremes ranging from 30 to 60 years old. Three-quarters (72%) of the perpetrators were between 31 and 40 years old and 25% were over 50 years old. Two-thirds of the victims in our sample are married (66.66%), while one-third maintain an unofficial relationship (sexual partner) (Fig. ). The description of the Socio-demographic characteristics of victims and perpetrators of spousal homicide is detailed in Table . Only one victim worked full-time. More than half of the victims (58.33%) were unemployed unlike the rest of the victims (33.33%) who were students. Contrary to the professional status of the victims, that of the perpetrators was composed in a proportion of 84% of employees and civil servants, whereas for the perpetrators, only two of them were unemployed at the time of the family tragedy. In terms of the circumstances of death, our study evoked above all the notion of misunderstanding and argument which ended in tragedy (66.66%), financial reasons (16.66%), and beatings without the intention of causing death (8 0.33%); however, in rare cases, no reason has been reported. A third of the victims (33.33%) were known to the healthcare environment and consulted for injuries and the need for medical care. Eight victims were unknown to the health services. Two women consulted at the forensic medicine unit of the legal medicine department, one at the gynecology and obstetrics department, and the other at the surgical emergency department for physical violence caused by an intimate partner. Regarding the period of the potential risk of violent acts, except in cases where the information relating to this parameter was missing (33.33%), a spousal homicide occurred much more during the period of marriage than divorce. Respective frequencies of 33.33% and 25% were noted, while temporary separation was mentioned only rarely (8.33%). The site of death was the marital home (41.66%); the second was the hospital (33.33%). For two victims, death took place in a public space. In terms of modus operandi or mechanisms of death, these were multiple and shared equally between three mechanisms: wounds by bladed weapons, wounds by projectiles from firearms, and blows by a blunt object. The medico-legal examination revealed a varied and characteristic external assessment, consisting mainly of non-specific contused wounds, characteristically sharp and stinging weapon wounds, and finally particular contused wounds caused by firearm projectiles. The internal assessment was mainly characterized by traumatic craniocerebral lesions (58.33%). The other fatal lesions were represented in Fig. . Toxicological investigations showed the presence of alcohol in a single victim and cannabis in two victims. However, they came back negative in the rest of the victims ( n = 09). The cause of death in spousal homicide is mainly represented by hemorrhagic shock (50%), followed by craniocerebral trauma (33.33%).
During the two years of the retrospective study (2017 and 2018), women who underwent a forensic autopsy for reasons of violent death represented a frequency of (5.71%): 35 women victims of violent death out of a total number of 570 autopsies performed. All age groups were concerned, but to a greater extent young women, particularly those aged between 31 and 40 (31.42%). Figure shows the respective frequencies of the age groups. Married women were the most affected. Indeed, nearly three-quarters of the victims are legally married (71.42%). The remaining proportion was shared between single women (14.28%), widowed women (11.42%), and divorced women (2.89%). The overwhelming majority of female victims of violent death (93%) were inactive at the time of the murderous act. Eighty-eight point fifty-seven percent of them had no profession, and 5.71% were retired. A frequency of 82.85% died in the marital home and at the hospital following complications from the initial trauma. Figure lists the sites where violent deaths occur among women. Homicide is the most represented forensic form, with a frequency of 51.42%, followed by suicide with a frequency of around 40%. Figure shows the respective frequencies of the three forensic forms of violent female death. In terms of injury tools of homicide victims, blows with blunt objects represent the most frequent operative mechanism (44.44%), followed by injuries caused by sharp instruments (39%) while the use of firearms was relatively lower (17%) (Fig. ). During suicides, hanging is the mostly used method by the victims (42.85%). However, the use of other processes, namely jumping off a high place (28.57%) was added to the ingestion of caustic products (21.42%) and drugs (7.14%). In homicides, the spouse is the perpetrator in 61.11% of cases. The perpetrator is unknown to the victim in 33.33% of the cases. Figure shows the link between the perpetrator and their victims. Forensic examination showed the following frequencies: Non-specific blunt wounds (47%), simple wounds (26%), and blunt gunshot wounds (6%) (Fig. ). Thoracic and abdominal lesions were the most frequently observed injuries in victims (34%). Figure shows the different lesions of internal violence highlighted by the forensic autopsy. The cephalic extremity was the region of predilection for this type of fatal trauma (51.42%), followed by the category “ multiple anatomical regions of the body ” where several body segments were affected (28.57%). The other traumatic locations were indicated in detail in Fig. . The use of complementary investigations represented an essential tool in the identification of the violent causes of death and in particular those secondary to intoxication. However, the results remain highly dependent on several parameters (date of the autopsy, date of sampling, nature of the sample and its storage, cause of death, etc.). Toxicological research was almost systematic among victims who died following a homicide: (N: 14/18: 77.77%) of all homicides and (N14/35: 40%) of all victims of violent deaths. The recourse to histological examination was rare. Thus, according to our results, we note the following frequencies: (N: 6/18), a frequency of 33.33% of all homicides, and (N: 6 /35), representing a frequency of 17.14% of all violent deaths. Table illustrates the different examinations carried out during autopsies of victims of violent deaths.
During the prospective survey (2019—2020), 12 spousal homicides were reported out of a total number of 670 forensic autopsies carried out during the aforementioned period, a frequency of nearly two percent (1.79%) of all medical-legal deaths. The sample studied had an average age of 33 ± 12.91 years, with extremes of 19 to 56 years. The age group between 21 and 30 is the most represented one, with a frequency of (50, 33%), followed by those aged over 50 years old, with a frequency of (25%). The perpetrators of spousal homicides had an average age slightly higher than that of the victims: about 42 ± 10.73 years with extremes ranging from 30 to 60 years old. Three-quarters (72%) of the perpetrators were between 31 and 40 years old and 25% were over 50 years old. Two-thirds of the victims in our sample are married (66.66%), while one-third maintain an unofficial relationship (sexual partner) (Fig. ). The description of the Socio-demographic characteristics of victims and perpetrators of spousal homicide is detailed in Table . Only one victim worked full-time. More than half of the victims (58.33%) were unemployed unlike the rest of the victims (33.33%) who were students. Contrary to the professional status of the victims, that of the perpetrators was composed in a proportion of 84% of employees and civil servants, whereas for the perpetrators, only two of them were unemployed at the time of the family tragedy. In terms of the circumstances of death, our study evoked above all the notion of misunderstanding and argument which ended in tragedy (66.66%), financial reasons (16.66%), and beatings without the intention of causing death (8 0.33%); however, in rare cases, no reason has been reported. A third of the victims (33.33%) were known to the healthcare environment and consulted for injuries and the need for medical care. Eight victims were unknown to the health services. Two women consulted at the forensic medicine unit of the legal medicine department, one at the gynecology and obstetrics department, and the other at the surgical emergency department for physical violence caused by an intimate partner. Regarding the period of the potential risk of violent acts, except in cases where the information relating to this parameter was missing (33.33%), a spousal homicide occurred much more during the period of marriage than divorce. Respective frequencies of 33.33% and 25% were noted, while temporary separation was mentioned only rarely (8.33%). The site of death was the marital home (41.66%); the second was the hospital (33.33%). For two victims, death took place in a public space. In terms of modus operandi or mechanisms of death, these were multiple and shared equally between three mechanisms: wounds by bladed weapons, wounds by projectiles from firearms, and blows by a blunt object. The medico-legal examination revealed a varied and characteristic external assessment, consisting mainly of non-specific contused wounds, characteristically sharp and stinging weapon wounds, and finally particular contused wounds caused by firearm projectiles. The internal assessment was mainly characterized by traumatic craniocerebral lesions (58.33%). The other fatal lesions were represented in Fig. . Toxicological investigations showed the presence of alcohol in a single victim and cannabis in two victims. However, they came back negative in the rest of the victims ( n = 09). The cause of death in spousal homicide is mainly represented by hemorrhagic shock (50%), followed by craniocerebral trauma (33.33%).
Violence against women is frequent; it is a source of multiple consequences ranging from simple trauma to death. It is difficult at our level to know the exact number of homicides resulting from domestic violence. During the initial investigation, we identified 35 cases of violent death involving women, representing a frequency of 5.71% of all medico-legal deaths; whereas, during the prospective cross-sectional survey 12 spousal homicides were recorded out of some 670 forensic autopsies, representing a frequency of 1.79% of all forensic deaths. As an indication and according to the results of a systematic review conducted by a research team, under the direction of the World Health Organization during the year 2013, a frequency of approximately 13.5% of homicides was committed by the intimate female partner and this frequency is six times higher than that of homicides against male partners . Young victims were more affected by this form of violent death: respective frequencies of 31.42% and 50, 33% were noted for victims aged between 31 and 40 years old during the retrospective survey and 21 and 30 years old during the prospective survey. According to the Italian study described earlier in this article, 50% of the victims were between 18 and 49 years old, and the spousal homicides were committed in the domestic context (78.5%) by male perpetrators (85%), having a relationship with the victims, as intimate or ex-intimate partners (73.5%) . While for the Turkish study, female victims of spousal homicides had an age range between 19 to 29 (38.5%), and (15.5%) of them were 18 or younger . The average age of the victims was evaluated at 33 ± 12.91 years, with extremes of 19 to 56 years. Regarding this criterion, a retrospective analysis of forensic autopsy records of adult femicide victims in Taiwan over ten years showed that the average age of victims killed by intimate partners (40.0 years) was younger than those killed by non-intimate partners (48.6 years) . This result concerns us directly and requires objective answers because this observation is completely contrary to other series where the average age of the victims was relatively higher (average age 45.8 with extremes of 18 and 85 years . According to our study, elderly victims are remarkably less affected. For the Turkish series, the medium age of victims was 43 years old and a significant proportion of victims (49.7%) were between 21 and 40 years old . The average age of perpetrators of spousal homicides is 42 ± 10.73 years with extremes ranging from 30 to 60 years old. Three-quarters of the perpetrators were between 31 and 40 years old and 25% are over 50 years old. This rate joins that of the Canadian series and other series of the European continent where an average age of 43.2 years was noted and it is relatively high compared to the Tunisian series, where this rate was 35, 2 years . A clear predominance of married women with respective frequencies of (71.42%) and (66.66%). This frequency may be superimposed on that noted by the various series among living victims where domestic violence is widespread . A remarkably high frequency of unemployed victims was recorded, both by the retrospective survey (88.57%) and by the prospective survey (58.33%). This confirms the frequent implication of socio-economic conditions where the share of unemployment and inactivity were frequently reported. The presence of previous violent episodes could be considered an early warning sign of spousal homicide. Unlike the two Canadian and Parisian series, our survey showed a frequency of more than half (56.3%( of victims who are known to the care services and who have suffered previous domestic violence. This frequency is comparable to the Tunisian series . For our study, 12.5% of the authors had a history of staying in a psychiatric environment, this result is in disagreement with the Tunisian series, where 80% of the authors suffered from a mental disorder . This highlights the importance of a systematic screening of situations of domestic violence at the level of basic care units and the possibility of alarming the victims who present serious injuries which could end in a family tragedy. The marital home represents the primary site of predilection for violent deaths and marital homicides. The hospital comes in the second position where death occurs following the various complications of the initial domestic injuries. This observation is shared with a French study . In terms of the period of cohabitation during which the death took place. It appears that spousal homicide occurs much more during marriage than divorce, contrary to the results of the two series above ; nearly a third (31.3%) of the same proportion of perpetrators were in a separated couple, and more than a third (37.5%) were in the process of separation at the time of the murderous act. According to the results of the retrospective survey, in more than half of the cases (51.42%), the violent deaths are the result of a criminal violent death while in 40% of the cases the victims presented injuries similar to those observed in suicides. The results of an American study in North Carolina analyzing the precursors to female suicides suggested that intimate partner violence was a precursor in at least 4.5% of single suicides and intimate partner violence was present in 6.1% of suicides overall . The modus operandi in homicide was represented in almost half of the cases by blows with blunt objects and bladed weapons. Firearms are rarely used, contrary to the results of a French series where 70% of murders were committed with a firearm . In spousal homicides, perpetrators appear to use as many sharp and pointed weapons as firearms and blunt objects. For the Canadian series, 62.5% used a firearm and 43.8% of the perpetrators of spousal homicides resorted to excessive violence . While according to the Italian study, the homicides were mainly perpetrated, with sharp and prickly instruments (32%), blunt instruments (21.5%), gunshots (18.5%), and asphyxiation (16.5%) . As for the mechanism of death, in the Turkish study, spousal homicides were most often committed with a sharp object (49.3%) . For the Istanbul City Series, violent deaths were committed by the following mechanisms: Gunshot wounds (50.1%), strangulation (8.4%), and stabbing wounds (28.3%) . According to another Italian study conducted on a sample of 319 cases of spousal homicide, sharp and spinous weapons were used in (31.97%), firearms in (27.27%), and asphyxiation in (16.30%) of the cases. These homicides were committed inside the marital home in (65.51%) of the cases . In deaths by suicide, we have found that the most common means used is hanging due to its radical nature, in addition to other means namely precipitation from a high place, ingestion of caustic products, and drug intoxication. In homicides, the alleged perpetrators are represented in nearly two-thirds (61.11%) of the cases by the spouse. This immediately raises the urgency of a systematic screening of situations of intra-family violence and in particular violence provoked by the intimate partner. In terms of motivations for spousal homicide. This study evoked above all the notion of disagreement and dispute which leads to a dramatic tragedy in addition to financial motives and blows without the intention to cause death, contrary to the Canadian series where 43.8% of the perpetrators report respectively, and in decreasing order of frequency: marital separation (18.8%), mental disorders (6.3%), and pecuniary motivations. For the Parisian series , the reasons reported were respectively the perpetuation of violence (94%), separation (16%), and verbal threats of death or suicide by the perpetrators (12%). For the Turkish series, the perpetrators declared, during their trials, before the court that the most frequent reason for homicide was the request for divorce or separation (61.5%) . According to a study conducted on homicide-suicide in the province of KONYA in Turkey, out of ten cases of homicide-suicides, the precipitating reason for this act was an imminent divorce in four cases, a depressive syndrome in two cases, an antisocial personality disorder in one case, and a pedophile suffering from reactive depression . The results of a study conducted on a sample of 86 cases of femicide suggest that it was perpetrated disproportionately by intimate partners (current or past) rather than by strangers with high levels of litigation and conflict representing significant precursors to spousal homicide . The external examination revealed traumatic bodily injuries in almost all of the abused victims. Their nature depends closely on the instruments used: non-specific contused wounds, located mainly at the level of the cephalic extremity, and simple wounds caused by a stab wound in a third of the victims, with their distribution over several segments of the body. Signs of passive constriction of the neck, in the form of a cervical hanging furrow, are not insignificant while blunt force injuries characteristic of gunshot wounds have been noted in one-third of domestic homicide victims. The medico-legal autopsy and the complementary investigations are of fundamental interest in the knowledge of the causes of violent death and particularly in marital homicide; various and multiple traumatic lesions were highlighted by the autopsy. These are mainly traumatic cranio-encephalic lesions, thoracoabdominal visceral lesions, and, to a lesser extent, deep cervical lesions. For the Jordanian study, the most common cause of death was gunshot wounds, and the severity of the violence was judged by the highest number of serious injuries . Heart wounds were noted in one victim of violent death, in 10, and a quarter of victims of spousal homicide. Toxicological research is systematic in violent deaths by intentional homicide and suicide, given its involvement in medico-legal circumstances, whereas it is less often carried out in the event of accidental death. The histological examination is of less utility compared to the toxicological examinations in the establishment of violent causes of death. Finally, in spousal homicide, the cause of death is essentially due to hemorrhagic shock or complications of traumatic brain injury.
The strengths of this survey are the dual nature of the survey, (retrospective and prospective), the methodology and source of study data (autopsy reports), and its duration extending over four years. The limits of our study are the cross-sectional survey sample which is small despite the two-year duration. Some parameters could be incomplete because the lifting of bodies is not systematic and remains dependent on the judicial decision. Some data related to the context of the death may lack precision as it was collected from the entourage and police reports during the preliminary investigation.
This study highlights the reality of violent deaths against women in the shape of spousal homicides, which were hidden despite social, cultural, and legal charges. The real figures remain difficult to grasp due to the underestimation and the underreporting of family violence and violence caused by intimate partners for social, cultural, and religious reasons, etc. Particular attention on the part of professionals working in the field of forensic medicine is perfectly justified and essential in the reception and care services for abused women in the forensic investigation units. Young married women without professional activity represent the most vulnerable category of homicide victims. While in terms of predictive signs, our study identified two main factors, which are the place of occurrence of the violent death, represented by the marital home, and the period of marriage. For the mechanism of death in homicide, our study identified the three classically used modes namely blows by blunt objects, wounds by sharp and prickly weapons, and finally wounds by firearms. Three main factors have been identified as motives for committing the criminal act: the notion of argument and disagreement, financial motives and difficulties encountered by couples, and recurring altercations. We underline the interest in screening for domestic violence as well as the search for risk factors during the management of violence induced by the intimate partner; this could increase the probability for the actors of the management (police, social and judicial forces) to anticipate the murder before its occurrence and thus saving the women from irreparable harm. This study indicates that forensic medicine services could be a reliable source of statistical data and confirms that forensic research could help identify predictors of domestic homicides, which absolutely, must be taken into account in strategies to combat domestic violence and in the design of support systems and prevention strategies in the fight against domestic violence. The results of this work must be pursued by other large-scale studies and bigger samples to have more solid scientific knowledge allowing the establishment of strategies, aimed at evaluating the potential risk of conjugal homicide at an early stage and providing appropriate support to victims on time.
|
The experience of caregivers providing therapeutic patient education for people living with bipolar disorder: a qualitative study | 69c86e66-273a-4d04-b8be-bde093546410 | 10039597 | Patient Education as Topic[mh] | The caregivers interviewed who lead therapeutic education groups described a change in their caregiving posture, which was more horizontal and collaborative. Therapeutic education thus fostered empowerment and destigmatization, promoting patients appeasement and openness. The observations overlap with personal recovery levers. Facilitation of therapeutic education could help caregivers move towards a more recovery-oriented care posture.
The World Health Organization considers bipolar disorder (BD) to be one of the ten most disabling chronic pathologies , and it has a prevalence of 2.6% in the general population . Frequent relapses in bipolar disorder have serious impacts on the psychosocial functioning, cognition, quality of life, and survival of affected individuals . Management recommendations for BD [ – ] call for a combination of a pharmacological approach and psychosocial interventions. Among the latter, there is a good level of evidence for psychoeducation . Psychoeducation is an individual or group-based intervention that aims to provide knowledge and skills to individuals affected by a chronic psychiatric disorder so that they can better manage their lives. Quantitative studies have indicated that psychoeducation has a beneficial impact on BD by reducing the rate of relapse, increasing the time between episodes of decompensation, decreasing the number and length of hospitalizations [ – ], and improving patients' treatment adherence and social functioning [ – ]. This therapeutic approach is considered promising for other psychiatric disorders such as schizophrenia. Its favourable effects were found in the field of schizophrenia, as shown by two systematic reviews . Psychoeducation has been shown to be effective in reducing relapse, length of hospitalization and medication compliance . Although the level of evidence is still modest for brief programs, there is a short-term effect on the number of relapses and a medium-term effect on medication compliance . In France, the “Hospital, Patients, Health and Territories” Act of 2009 included therapeutic patient education (TPE) in the Public Health Code and thereby established a regulatory framework for TPE. Access to TPE programs is recommended for any chronic disease. TPE "aims to help patients acquire or maintain the skills they need to manage their lives with a chronic disease" and must be adapted to the specific needs of each participant via an "educational diagnosis" evaluation. This degree of personalization in TPE differs from that in psychoeducation programs. The regulatory framework recommends the co-construction and co-facilitation of TPE programs with people concerned about the specific disease. Only a few studies have explored the experience of caregivers as facilitators in TPE. In the field of psychiatry, there are no articles to our knowledge that have explored the experiences of the facilitators of these programs. A recent study was interested in the representations of caregivers concerning the practice of TPE, and highlighted the great reluctance to use this care tool, which would conflict with a traditional model more paternalistic than collaborative . The aim of our work was therefore to qualitatively explore the experience of caregivers and peer helpers working in psychiatry as facilitators of a TPE program for people living with BD. Our intention was to describe what happens during TPE sessions from the facilitators' point of view and what they believe could promote the well-being and recovery process of people living with BD. The therapeutic patient education program Program construction and implementation The studied TPE program was dedicated to people living with BD and was developed by the research team in 2016; the team developed the program according to the national recommendations in force (HAS, 2007). The program has since been facilitated by the professionals of the Psychosocial Rehabilitation Center (PSR): nurses, peer helpers, and psychiatric physicians. It had been running on a regular basis for 5 years when the study was conducted. Program progress The program was co-facilitated by two to three health professionals and peer helpers trained in TPE. Each TPE group included 10 patients. It was conducted following four stages: i) an individual assessment interview, leading to an “educational diagnosis”; ii) nine weekly group sessions lasting two hours (except for the eighth session, which lasted three hours); iii) a final individual interview to evaluate the skills developed by the participants and their satisfaction with the program; and iv) a group consolidation session three months after the ninth session. Description of the program sessions Each session began with an "ice-breaker", followed by a debriefing on the previous session using participatory animation tools. A theme was addressed during the first hour; after a short break, a second theme was addressed, and the session ended with a period for synthesis and a mood self-assessment. An individual time could be proposed to participants who felt the need. The team created a booklet that followed the themes of the program that was given to the participants so that they could write down information throughout the sessions. Table presents the themes addressed and the objectives of the sessions.
Program construction and implementation The studied TPE program was dedicated to people living with BD and was developed by the research team in 2016; the team developed the program according to the national recommendations in force (HAS, 2007). The program has since been facilitated by the professionals of the Psychosocial Rehabilitation Center (PSR): nurses, peer helpers, and psychiatric physicians. It had been running on a regular basis for 5 years when the study was conducted. Program progress The program was co-facilitated by two to three health professionals and peer helpers trained in TPE. Each TPE group included 10 patients. It was conducted following four stages: i) an individual assessment interview, leading to an “educational diagnosis”; ii) nine weekly group sessions lasting two hours (except for the eighth session, which lasted three hours); iii) a final individual interview to evaluate the skills developed by the participants and their satisfaction with the program; and iv) a group consolidation session three months after the ninth session. Description of the program sessions Each session began with an "ice-breaker", followed by a debriefing on the previous session using participatory animation tools. A theme was addressed during the first hour; after a short break, a second theme was addressed, and the session ended with a period for synthesis and a mood self-assessment. An individual time could be proposed to participants who felt the need. The team created a booklet that followed the themes of the program that was given to the participants so that they could write down information throughout the sessions. Table presents the themes addressed and the objectives of the sessions.
The studied TPE program was dedicated to people living with BD and was developed by the research team in 2016; the team developed the program according to the national recommendations in force (HAS, 2007). The program has since been facilitated by the professionals of the Psychosocial Rehabilitation Center (PSR): nurses, peer helpers, and psychiatric physicians. It had been running on a regular basis for 5 years when the study was conducted.
The program was co-facilitated by two to three health professionals and peer helpers trained in TPE. Each TPE group included 10 patients. It was conducted following four stages: i) an individual assessment interview, leading to an “educational diagnosis”; ii) nine weekly group sessions lasting two hours (except for the eighth session, which lasted three hours); iii) a final individual interview to evaluate the skills developed by the participants and their satisfaction with the program; and iv) a group consolidation session three months after the ninth session.
Each session began with an "ice-breaker", followed by a debriefing on the previous session using participatory animation tools. A theme was addressed during the first hour; after a short break, a second theme was addressed, and the session ended with a period for synthesis and a mood self-assessment. An individual time could be proposed to participants who felt the need. The team created a booklet that followed the themes of the program that was given to the participants so that they could write down information throughout the sessions. Table presents the themes addressed and the objectives of the sessions.
Type of research Our study was based on an inductive and monocentric qualitative exploratory methodology. For this purpose, we conducted one focus group (FG) of health professionals (nurses, psychiatrists and psychiatric residents) and peer helpers facilitating a TPE program for bipolar disorders in a psychosocial rehabilitation center located in a Psychiatry and Mental Health Department. This FG was conducted in January 2021. The framework of the FG was developed by a public health intern and a psychiatric physician who facilitated the FG and were not involved in the care. The themes that were addressed following a semidirective guide were i) the facilitators' motivation to facilitate a TPE program; ii) representations of the role of the TPE facilitators; iii) caregivers' experiences; iv) changes observed in the participants by the facilitators; and v) dimensions associated with the observed changes. Study population Our sample targeted the study population of health professionals and expert patients running the "Bipolar Disorders" TPE program. The inclusion criteria for this study were i) to have facilitated at least one session of the bipolar disorders TPE program studied in the past year; ii) to have received 40 h of academic training in the facilitation of TPE programs; and iii) to be available for the duration of the FG. Mode of collection After receiving the agreement of the head of the PSR center, we presented our study project to the health care team in charge of running the "Bipolar Disorders" TPE program. At the time of the study, two sessions of the program were taking place in parallel, with each program having its own team of facilitators. We offered all the TPE program facilitators the opportunity to participate in the FG, and all agreed. A single FG was convened with all facilitators from the ongoing programs. The FG, which lasted one and a half hours, was facilitated by two health professionals, a public health intern trained in FG facilitation and a psychiatric physician; the latter were not involved in facilitating the TPE program. Two observers were also present: a public health intern not involved in the program, and an advanced practice nurse who was familiar with the program from having previously co-facilitated it. The observers wrote down the participants' reactions and responses. With the participants' consent, an audio recording was made and stored on a password-protected audio file. Transcription was performed manually without software support. The recordings were deleted once they were transcribed. The sociodemographic data of the sample were also collected during the FG. Method of analysis The FG corpus was transcribed manually by the first author. To process the data, we relied on the method of thematic analysis which consisted of capturing all of the themes of the data collected during the FG to identify the central themes that answered the research question. This method was composed of the following steps: i) coding/thematization; ii) categorization; iii) linking of the data; iv) presentation of the data; and v) verification of the data. This method made it possible to understand and interpret the data while remaining faithful to what the participants said. The reading of the FG transcript and its analysis were carried out by an advanced practice nurse and a psychiatric physician in a blinded fashion. Neither of these two individuals had been involved in the facilitation of this program under review, but they had previous experience in facilitating it. A consultation between these two authors took place at the end of the coding process to compare the themes identified by each and to create links between the categories. All themes were identified by both researchers, and agreement between them on the organization of the categories was reached. Ethical considerations Each participant was given an oral presentation of the study design, and signed informed consent was obtained. The protocol was validated by the ethics committee (Groupe Nantais d'Ethique dans le Domaine de la Santé, GNEDS) on 23 July 2020. To avoid the identification of any participant, the data collected in this study were processed in an entirely anonymous manner. Anonymity was ensured by a numerical code accessible only by the researchers.
Our study was based on an inductive and monocentric qualitative exploratory methodology. For this purpose, we conducted one focus group (FG) of health professionals (nurses, psychiatrists and psychiatric residents) and peer helpers facilitating a TPE program for bipolar disorders in a psychosocial rehabilitation center located in a Psychiatry and Mental Health Department. This FG was conducted in January 2021. The framework of the FG was developed by a public health intern and a psychiatric physician who facilitated the FG and were not involved in the care. The themes that were addressed following a semidirective guide were i) the facilitators' motivation to facilitate a TPE program; ii) representations of the role of the TPE facilitators; iii) caregivers' experiences; iv) changes observed in the participants by the facilitators; and v) dimensions associated with the observed changes.
Our sample targeted the study population of health professionals and expert patients running the "Bipolar Disorders" TPE program. The inclusion criteria for this study were i) to have facilitated at least one session of the bipolar disorders TPE program studied in the past year; ii) to have received 40 h of academic training in the facilitation of TPE programs; and iii) to be available for the duration of the FG.
After receiving the agreement of the head of the PSR center, we presented our study project to the health care team in charge of running the "Bipolar Disorders" TPE program. At the time of the study, two sessions of the program were taking place in parallel, with each program having its own team of facilitators. We offered all the TPE program facilitators the opportunity to participate in the FG, and all agreed. A single FG was convened with all facilitators from the ongoing programs. The FG, which lasted one and a half hours, was facilitated by two health professionals, a public health intern trained in FG facilitation and a psychiatric physician; the latter were not involved in facilitating the TPE program. Two observers were also present: a public health intern not involved in the program, and an advanced practice nurse who was familiar with the program from having previously co-facilitated it. The observers wrote down the participants' reactions and responses. With the participants' consent, an audio recording was made and stored on a password-protected audio file. Transcription was performed manually without software support. The recordings were deleted once they were transcribed. The sociodemographic data of the sample were also collected during the FG.
The FG corpus was transcribed manually by the first author. To process the data, we relied on the method of thematic analysis which consisted of capturing all of the themes of the data collected during the FG to identify the central themes that answered the research question. This method was composed of the following steps: i) coding/thematization; ii) categorization; iii) linking of the data; iv) presentation of the data; and v) verification of the data. This method made it possible to understand and interpret the data while remaining faithful to what the participants said. The reading of the FG transcript and its analysis were carried out by an advanced practice nurse and a psychiatric physician in a blinded fashion. Neither of these two individuals had been involved in the facilitation of this program under review, but they had previous experience in facilitating it. A consultation between these two authors took place at the end of the coding process to compare the themes identified by each and to create links between the categories. All themes were identified by both researchers, and agreement between them on the organization of the categories was reached.
Each participant was given an oral presentation of the study design, and signed informed consent was obtained. The protocol was validated by the ethics committee (Groupe Nantais d'Ethique dans le Domaine de la Santé, GNEDS) on 23 July 2020. To avoid the identification of any participant, the data collected in this study were processed in an entirely anonymous manner. Anonymity was ensured by a numerical code accessible only by the researchers.
In this study, our sample consisted of eight facilitators from the TPE program studied. Simple description of sociodemographic data The average age of the participants was 40.5 years; the minimum age was 29 years, and the maximum age was 66 years. Half of the participants were between 33 and 49 years of age, 3 were between 18 and 33 years of age, and 1 was over 50 years of age. The standard deviation of ages was 11.32, and the median age was 37.5 years. The majority of the participants were women ( N = 6). Of the eight facilitators, two were psychiatric physicians (PPs), two were state-registered nurses (Ns), two were volunteer peer helpers (VPHs) (one retired, one a craftsman), one was a psychiatric intern (PI), and one was a hospital-employed peer health mediator (EPH). All had experience working in a psychosocial rehabilitation department. Experience in facilitating TPE sessions ranged from less than 5 sessions facilitated ( N = 1) to more than 30 sessions facilitated ( N = 1). More than half of the facilitators had facilitated at least eleven TPE sessions ( N = 6). Two-thirds of the co-facilitators ( N = 6) had participated in the facilitation of at least four different sessions of the studied TPE program; two of these facilitators had co-facilitated the program four times, two six times, one seven times and one eight times. The other two had co-facilitated the TPE program one and twice. Discourse analysis Four dimensions were identified during the qualitative analysis of the corpus: i) facilitators' experiences of the TPE sessions, ii) being a TPE facilitator, iii) the role of the TPE sessions, and iv) perceived changes in patients. For each dimension, themes and subthemes emerged. A summary of the results is presented in Table . The detailed presentation of the results of the discourse analysis is presented next; the participants are identified anonymously with the following information: role, gender, and age. Facilitators' experiences of TPE sessions This dimension was associated with four themes: a climate of safety, belonging to a group, sharing experiences and the pleasure of facilitation. The theme of safety was associated with three subthemes: feeling confident, allowing oneself to be authentic, and the notion of caring. One of the main elements in providing a safe climate was the desire to create a space of trust: "It is my desire to facilitate a climate of authenticity and listening in the group" (EPH, female, 33). One of the facilitators noted what a patient said about being listened to: "It reminds me of what a patient said at one point, as if it was very surprising. He said, 'It's incredible here they listen to us,' as if in a psychiatric hospital, you can't be listened to. He couldn't believe it. It was quite spectacular to hear this come from his own mouth" (VPH, male, 66 years old). The notion of authenticity was taken up and explained in these terms: "To have an identity that is my own, that can be different from others" (EPH, female, 45 years old). One facilitator mentioned the desire to "create this space where (…) he feeds both the feeling of freedom and the need for relationship" (PP, female, 32 years old). This intention for relationships and connection was taken up again by another participant: "A good atmosphere and to make sure that there is a link between people, that there is exchange, and that people can have moments of listening with their peers that they don't necessarily have outside the sessions, of support in fact from others, to develop a feeling of belonging, to live better with it afterward" (N, female, 44 years old). Creating this space of security was reinforced by the notion of benevolence. "which is quite present in TPE and which is good for the group and good for ourselves" (PI, male, 29 years old). This benevolence created a climate of security for the participants and the facilitators: "There is something benevolent in the group in general, which makes me feel good" (N, female, 42 years old). Half of the facilitators interviewed ( N = 4) mentioned the feeling of belonging to a group shared with the users: "It is first of all that we are in a group, that the patients I meet in another context than the crisis" (N, female, 42 years old); "to be part of this adventure, which is something I believe in, which is everything that is mutual aid groups" (VPH, female, 45 years old). The majority of the facilitators ( N = 7) verbalized the feeling of belonging with colleagues and associated it with a notion of pleasure: "One of my motivations for participating is already to find people from the service; being able to lead the groups is always a pleasure for me" (PI, male, 29 years old). For one of the facilitators, the fact that the program was run in a group setting encouraged the sharing of experiences: "The idea that they are in a group, that they share with their peers… The sharing of experiences seems to me to be very rich" (PP, female, 32 years old). This point was supported by other participants: "There is this sharing between peers, which is very important. It is very interesting because it leaves room for knowledge through experience" (PP, female, 32 years old). Others mentioned the effects of this sharing, stating that "these TPE groups allow for an exchange between participants and an awareness of certain symptoms or disorders" (PI, male, 29 years old) and adding that "there is also the fact that exchanging between peers allows some people to get out of the denial in which they are to varying degrees" (VPH, male, 66 years old). All the participants in the FG expressed their pleasure in leading these TPE sessions: "It feels good in general. After a TPE session, I feel good" (PI, male, 29 years old). Being a facilitator of TPE sessions This dimension highlighted the following themes: the place of peer helpers, having an expert role, taking an educational posture, taking a listening posture, having a horizontal relationship and being the guarantor of the framework. The place of peer helpers was associated with being. "a reference point. That's what I felt when I had a peer helper as a facilitator, that he was a pillar of recovery" (VPH, female, 45 years old). For this facilitator, embodying. "a witnessing role (…) of a patient with bipolar disorder in recovery (…) that exists, that you can see" meant being a bearer of hope toward a recovery process. The lived experience of the illness was also associated with an expert role. One of the peer-helper facilitators described his role as follows: "When I am in the TPEs, it is to validate certain questions (asked by the participants in the TPE program) with my experience, my experimentation. It gives legitimacy to a whole bunch of sometimes quite delirious things that we may have experienced but that exist" (VPH, female, 45 years old). This function of expert of experience was contrasted with the role of medical expertise. One of the participating doctors showed his discomfort with. "delivering medical information. I feel like I'm taking on a bit of a posture (…); like the doctor who is a guest on a TV show, ‘So doctor, tell us’ (laughter). At the same time, I feel that sometimes, there are questions that need to be answered and that we have some answers, and I want to pass on knowledge and information, and it can be nice to have this medical hat on" (PP, female, 32 years old). One of the psychiatrists involved in TPE mentioned that the expert posture could give way to an educational posture: "When I am a facilitator, I feel that I am leaving my medical and caregiving posture to be more an educator. I have that feeling. I am trying to create a climate conducive to learning by transmitting information" (PP, female, 32 years old). Another facilitator emphasized being in a listening posture: "During the TPE groups, I spend a lot of time in a listening position. Everyone around the table, I feel like I am listening to them" (PP, female, 32 years old). Most of the facilitators had perceived a tendency to be in a horizontal posture of collaborative partnership in interaction with their co-facilitators and with patients: "What we aim for in TPE is horizontality" (PP, female, 32 years old);"It's about being on the same level in fact" (VPH, female, 45 years old). Finally, one of the psychiatrists mentioned the importance of being a guarantor by ensuring. "the notion of a framework, the organization of the session, paying attention to time, sharing the word (…) The framework and the flexibility in the framework" (PP, female, 32 years old). The role of TPE sessions The following three themes were used to clarify this dimension: knowledge provision, empowerment, self-normativity and destigmatization. The TPE facilitators mentioned the contribution of knowledge both for themselves and for the patients through exchanges: "I am almost in the position of the one who is also learning, who is learning, who is growing at the same time as the group" (N, female, 42 years old). Another added "… it taught me a lot of things in fact. I think it improves my daily practice and the support I can offer my patients" (PP, female, 32 years old). Others underscored the pleasure of. "being able to accompany, encourage, transmit knowledge" (VPH, woman, 45 years old) and "to bring clarity to people concerned by psychological disorders, on the disorders themselves" (PP, woman, 32 years old). Associated with this theme was the promotion of empowerment: "To create this space where we tell them that there is another possible way than the one of dependence on care, to find their own resources, to better understand themselves" (PP, female, 32 years old). This notion of empowerment could also be promoted through experiential sharing: "We can explain things, but it won't have the same impact as someone who has experienced the same disorders or who has gone through certain things or who has taken certain treatments, who has realized that with certain treatments, it was effective or not" (PI, male, 29 years old). The notion of empowerment could also be promoted through the provision of tools to develop autonomy: "I like this image of the toolbox, that they can leave at the end with a toolbox of the TPE session, that they can choose the tools they prefer to work with" (N, female, 44 years old). For some participants, the role of TPE sessions was to promote self-normativity by helping them. "to really find their own standards within themselves by trying to get away from the injunctions of the outside world, of society, to really build their recovery on who they are, on their needs" (EPH, female, 33 years old). For others, the role of TPE sessions was to promote destigmatization by talking about. "the illness, in general, which is quite stigmatized. For personal reasons, I really want to be able to put words to it, to have things said, to be able to put words to it" (PP, female, 32 years old). Changes perceived by the TPE facilitators in patients who had used the TPE program This dimension included seven themes: relief, awareness of a shared experience, openness, identification, quality of life, a first stage of re-engagement, and grief. The notion of appeasement was the first change identified in patients; this change emerged through the destigmatization of words and experiences: "It's about appeasement. The fact of being able to verbalize in the group. The fact that it can be heard by others, that it is not judged, accepted, validated because we have gone through the same things" (N, female, 42 years old). Another change observed was awareness, which is an internal process of transformation that changes representations of the self and the world. One caregiver observed among patients: "the awareness that there is a common experience of the disorder" (N, female, 44 years old). This awareness may also have had a negative effect. As one participant described, the awareness. "of one's real limitations, that's something that can seem negative in TPE; patients are all at different stages of their disease, not necessarily all the same" (VPH, female, 45). This TPE facilitator verbalized that she observed a grieving process among the patients, understood here as a cognitive and affective process that accompanies the loss of someone or something: "the different phases of grief; in any process there is grief. What's funny is that when you're in this TPE, they're all addressed together. There are some who are in acceptance, recovery (…). A patient was quite in denial at the beginning and all that. It progresses; it progresses, and finally, all the phases of mourning are approached. Depending on where each person is at, they have the possibility of moving on to the next phase; there is a process that takes place" (VPH, woman, 45 years old). However, the phenomenon of identification with peers, and in particular with the peer helper, could make it possible to overcome this negative effect: "People identify with the person who has a pathology and who has recovered. They say to themselves "why not me?" False shame disappears; for example, when you start talking about your delusions, "he too", this necessarily breaks down barriers. The peer identifies with the participants who are also there. Double movement. It is because there is a double identification that complex exchanges take place and everyone identifies" (VPH, male, 66 years old). The FG participants also perceived a change in openness on the part of the patients. This openness was first described toward their peers: "As the sessions progress, participants will allow themselves to say things that they have experienced or gone through because they know that it will be heard and validated, because it will be experienced or in any case welcomed without being stigmatized" (N, female, 42); "at the end of the TPE, there is often a group created on WhatsApp® to keep in touch; they have become aware that it is a key to recovery to be supported by peers" (VPH, female, 45). For one participant, one of the effects of openness was the destigmatization: "(…) Taboos that disintegrate. There are people who say 'if he or she talks about it, I can talk about it too'. Under certain conditions of prudence, it can be an opportunity to talk about it with those close to them. It frees up speech" (VPH, male, 66 years old). This transition toward openness toward loved ones was also highlighted by another participant: "In the session with the loved ones, accepting that what you do as work is shared with the most intimate people and that it succeeds like the ones we have been able to live that pffff, (sic.)… it mobilizes a lot of things" (VPH, female, 45). Another participant identified an openness toward psychiatrists via. "a kind of destigmatization of psychiatrists (…) they are also there to help people" (PI, male, 29 years old). This interviewee also suggested an impact on the quality of life of the patients, defined by the World Health Organization (WHO) as. “individuals' perception of their position in life in the context of the culture and value systems in which they live and in relation to their goals, expectations, standards and concerns”. :"With (…) the toolbox, we can manage daily life a little bit, improve in the end to have a more fulfilled life with what we have seen in TPE" (PI, male, 29 years old). The notion of commitment was mentioned: "Participating in the TPE group is a first step to get back into action, to take care of oneself again" (PP, female, 32 years old). This empowerment was underlined through this example: "A participant was able to build her own information booklet based on her skills and transmit them to the whole group, and that was very important for her and useful for the other participants" (PP, female, 32 years old). This action reinforced the participant’s feeling of personal efficiency in managing her disease in her daily life.
The average age of the participants was 40.5 years; the minimum age was 29 years, and the maximum age was 66 years. Half of the participants were between 33 and 49 years of age, 3 were between 18 and 33 years of age, and 1 was over 50 years of age. The standard deviation of ages was 11.32, and the median age was 37.5 years. The majority of the participants were women ( N = 6). Of the eight facilitators, two were psychiatric physicians (PPs), two were state-registered nurses (Ns), two were volunteer peer helpers (VPHs) (one retired, one a craftsman), one was a psychiatric intern (PI), and one was a hospital-employed peer health mediator (EPH). All had experience working in a psychosocial rehabilitation department. Experience in facilitating TPE sessions ranged from less than 5 sessions facilitated ( N = 1) to more than 30 sessions facilitated ( N = 1). More than half of the facilitators had facilitated at least eleven TPE sessions ( N = 6). Two-thirds of the co-facilitators ( N = 6) had participated in the facilitation of at least four different sessions of the studied TPE program; two of these facilitators had co-facilitated the program four times, two six times, one seven times and one eight times. The other two had co-facilitated the TPE program one and twice.
Four dimensions were identified during the qualitative analysis of the corpus: i) facilitators' experiences of the TPE sessions, ii) being a TPE facilitator, iii) the role of the TPE sessions, and iv) perceived changes in patients. For each dimension, themes and subthemes emerged. A summary of the results is presented in Table . The detailed presentation of the results of the discourse analysis is presented next; the participants are identified anonymously with the following information: role, gender, and age.
This dimension was associated with four themes: a climate of safety, belonging to a group, sharing experiences and the pleasure of facilitation. The theme of safety was associated with three subthemes: feeling confident, allowing oneself to be authentic, and the notion of caring. One of the main elements in providing a safe climate was the desire to create a space of trust: "It is my desire to facilitate a climate of authenticity and listening in the group" (EPH, female, 33). One of the facilitators noted what a patient said about being listened to: "It reminds me of what a patient said at one point, as if it was very surprising. He said, 'It's incredible here they listen to us,' as if in a psychiatric hospital, you can't be listened to. He couldn't believe it. It was quite spectacular to hear this come from his own mouth" (VPH, male, 66 years old). The notion of authenticity was taken up and explained in these terms: "To have an identity that is my own, that can be different from others" (EPH, female, 45 years old). One facilitator mentioned the desire to "create this space where (…) he feeds both the feeling of freedom and the need for relationship" (PP, female, 32 years old). This intention for relationships and connection was taken up again by another participant: "A good atmosphere and to make sure that there is a link between people, that there is exchange, and that people can have moments of listening with their peers that they don't necessarily have outside the sessions, of support in fact from others, to develop a feeling of belonging, to live better with it afterward" (N, female, 44 years old). Creating this space of security was reinforced by the notion of benevolence. "which is quite present in TPE and which is good for the group and good for ourselves" (PI, male, 29 years old). This benevolence created a climate of security for the participants and the facilitators: "There is something benevolent in the group in general, which makes me feel good" (N, female, 42 years old). Half of the facilitators interviewed ( N = 4) mentioned the feeling of belonging to a group shared with the users: "It is first of all that we are in a group, that the patients I meet in another context than the crisis" (N, female, 42 years old); "to be part of this adventure, which is something I believe in, which is everything that is mutual aid groups" (VPH, female, 45 years old). The majority of the facilitators ( N = 7) verbalized the feeling of belonging with colleagues and associated it with a notion of pleasure: "One of my motivations for participating is already to find people from the service; being able to lead the groups is always a pleasure for me" (PI, male, 29 years old). For one of the facilitators, the fact that the program was run in a group setting encouraged the sharing of experiences: "The idea that they are in a group, that they share with their peers… The sharing of experiences seems to me to be very rich" (PP, female, 32 years old). This point was supported by other participants: "There is this sharing between peers, which is very important. It is very interesting because it leaves room for knowledge through experience" (PP, female, 32 years old). Others mentioned the effects of this sharing, stating that "these TPE groups allow for an exchange between participants and an awareness of certain symptoms or disorders" (PI, male, 29 years old) and adding that "there is also the fact that exchanging between peers allows some people to get out of the denial in which they are to varying degrees" (VPH, male, 66 years old). All the participants in the FG expressed their pleasure in leading these TPE sessions: "It feels good in general. After a TPE session, I feel good" (PI, male, 29 years old).
This dimension highlighted the following themes: the place of peer helpers, having an expert role, taking an educational posture, taking a listening posture, having a horizontal relationship and being the guarantor of the framework. The place of peer helpers was associated with being. "a reference point. That's what I felt when I had a peer helper as a facilitator, that he was a pillar of recovery" (VPH, female, 45 years old). For this facilitator, embodying. "a witnessing role (…) of a patient with bipolar disorder in recovery (…) that exists, that you can see" meant being a bearer of hope toward a recovery process. The lived experience of the illness was also associated with an expert role. One of the peer-helper facilitators described his role as follows: "When I am in the TPEs, it is to validate certain questions (asked by the participants in the TPE program) with my experience, my experimentation. It gives legitimacy to a whole bunch of sometimes quite delirious things that we may have experienced but that exist" (VPH, female, 45 years old). This function of expert of experience was contrasted with the role of medical expertise. One of the participating doctors showed his discomfort with. "delivering medical information. I feel like I'm taking on a bit of a posture (…); like the doctor who is a guest on a TV show, ‘So doctor, tell us’ (laughter). At the same time, I feel that sometimes, there are questions that need to be answered and that we have some answers, and I want to pass on knowledge and information, and it can be nice to have this medical hat on" (PP, female, 32 years old). One of the psychiatrists involved in TPE mentioned that the expert posture could give way to an educational posture: "When I am a facilitator, I feel that I am leaving my medical and caregiving posture to be more an educator. I have that feeling. I am trying to create a climate conducive to learning by transmitting information" (PP, female, 32 years old). Another facilitator emphasized being in a listening posture: "During the TPE groups, I spend a lot of time in a listening position. Everyone around the table, I feel like I am listening to them" (PP, female, 32 years old). Most of the facilitators had perceived a tendency to be in a horizontal posture of collaborative partnership in interaction with their co-facilitators and with patients: "What we aim for in TPE is horizontality" (PP, female, 32 years old);"It's about being on the same level in fact" (VPH, female, 45 years old). Finally, one of the psychiatrists mentioned the importance of being a guarantor by ensuring. "the notion of a framework, the organization of the session, paying attention to time, sharing the word (…) The framework and the flexibility in the framework" (PP, female, 32 years old).
The following three themes were used to clarify this dimension: knowledge provision, empowerment, self-normativity and destigmatization. The TPE facilitators mentioned the contribution of knowledge both for themselves and for the patients through exchanges: "I am almost in the position of the one who is also learning, who is learning, who is growing at the same time as the group" (N, female, 42 years old). Another added "… it taught me a lot of things in fact. I think it improves my daily practice and the support I can offer my patients" (PP, female, 32 years old). Others underscored the pleasure of. "being able to accompany, encourage, transmit knowledge" (VPH, woman, 45 years old) and "to bring clarity to people concerned by psychological disorders, on the disorders themselves" (PP, woman, 32 years old). Associated with this theme was the promotion of empowerment: "To create this space where we tell them that there is another possible way than the one of dependence on care, to find their own resources, to better understand themselves" (PP, female, 32 years old). This notion of empowerment could also be promoted through experiential sharing: "We can explain things, but it won't have the same impact as someone who has experienced the same disorders or who has gone through certain things or who has taken certain treatments, who has realized that with certain treatments, it was effective or not" (PI, male, 29 years old). The notion of empowerment could also be promoted through the provision of tools to develop autonomy: "I like this image of the toolbox, that they can leave at the end with a toolbox of the TPE session, that they can choose the tools they prefer to work with" (N, female, 44 years old). For some participants, the role of TPE sessions was to promote self-normativity by helping them. "to really find their own standards within themselves by trying to get away from the injunctions of the outside world, of society, to really build their recovery on who they are, on their needs" (EPH, female, 33 years old). For others, the role of TPE sessions was to promote destigmatization by talking about. "the illness, in general, which is quite stigmatized. For personal reasons, I really want to be able to put words to it, to have things said, to be able to put words to it" (PP, female, 32 years old).
This dimension included seven themes: relief, awareness of a shared experience, openness, identification, quality of life, a first stage of re-engagement, and grief. The notion of appeasement was the first change identified in patients; this change emerged through the destigmatization of words and experiences: "It's about appeasement. The fact of being able to verbalize in the group. The fact that it can be heard by others, that it is not judged, accepted, validated because we have gone through the same things" (N, female, 42 years old). Another change observed was awareness, which is an internal process of transformation that changes representations of the self and the world. One caregiver observed among patients: "the awareness that there is a common experience of the disorder" (N, female, 44 years old). This awareness may also have had a negative effect. As one participant described, the awareness. "of one's real limitations, that's something that can seem negative in TPE; patients are all at different stages of their disease, not necessarily all the same" (VPH, female, 45). This TPE facilitator verbalized that she observed a grieving process among the patients, understood here as a cognitive and affective process that accompanies the loss of someone or something: "the different phases of grief; in any process there is grief. What's funny is that when you're in this TPE, they're all addressed together. There are some who are in acceptance, recovery (…). A patient was quite in denial at the beginning and all that. It progresses; it progresses, and finally, all the phases of mourning are approached. Depending on where each person is at, they have the possibility of moving on to the next phase; there is a process that takes place" (VPH, woman, 45 years old). However, the phenomenon of identification with peers, and in particular with the peer helper, could make it possible to overcome this negative effect: "People identify with the person who has a pathology and who has recovered. They say to themselves "why not me?" False shame disappears; for example, when you start talking about your delusions, "he too", this necessarily breaks down barriers. The peer identifies with the participants who are also there. Double movement. It is because there is a double identification that complex exchanges take place and everyone identifies" (VPH, male, 66 years old). The FG participants also perceived a change in openness on the part of the patients. This openness was first described toward their peers: "As the sessions progress, participants will allow themselves to say things that they have experienced or gone through because they know that it will be heard and validated, because it will be experienced or in any case welcomed without being stigmatized" (N, female, 42); "at the end of the TPE, there is often a group created on WhatsApp® to keep in touch; they have become aware that it is a key to recovery to be supported by peers" (VPH, female, 45). For one participant, one of the effects of openness was the destigmatization: "(…) Taboos that disintegrate. There are people who say 'if he or she talks about it, I can talk about it too'. Under certain conditions of prudence, it can be an opportunity to talk about it with those close to them. It frees up speech" (VPH, male, 66 years old). This transition toward openness toward loved ones was also highlighted by another participant: "In the session with the loved ones, accepting that what you do as work is shared with the most intimate people and that it succeeds like the ones we have been able to live that pffff, (sic.)… it mobilizes a lot of things" (VPH, female, 45). Another participant identified an openness toward psychiatrists via. "a kind of destigmatization of psychiatrists (…) they are also there to help people" (PI, male, 29 years old). This interviewee also suggested an impact on the quality of life of the patients, defined by the World Health Organization (WHO) as. “individuals' perception of their position in life in the context of the culture and value systems in which they live and in relation to their goals, expectations, standards and concerns”. :"With (…) the toolbox, we can manage daily life a little bit, improve in the end to have a more fulfilled life with what we have seen in TPE" (PI, male, 29 years old). The notion of commitment was mentioned: "Participating in the TPE group is a first step to get back into action, to take care of oneself again" (PP, female, 32 years old). This empowerment was underlined through this example: "A participant was able to build her own information booklet based on her skills and transmit them to the whole group, and that was very important for her and useful for the other participants" (PP, female, 32 years old). This action reinforced the participant’s feeling of personal efficiency in managing her disease in her daily life.
General discussion Main results Our study examined the experiences of caregiver facilitators of TPE groups in the specific area of bipolar disorder in psychiatry. The facilitators emphasized that their particular posture in the TPE program involved horizontality, a feeling of belonging to a common group with the participating patients, the sharing of experiences and even self-disclosure. They also said that they observed the effects of these programs on the participants; in addition to mutual knowledge, they insisted on the notions of empowerment and destigmatization. Mutual identification with peer helpers seemed to favor identity reconstruction, or at least a form of appeasement. TPE and the personal recovery process Personal recovery from psychiatric disorders is a journey toward a satisfying and hopeful life despite the possible persistence of symptoms. It is described as. “a deeply personal, unique process of changing one’s attitudes, values, feelings, goals, skills and/or roles… a way of living a satisfying, hopeful and contributing life even with the limitations caused by illness” . The CHIME framework was developed based on a literature review conducted by Leamy and her colleagues , who identified five processes as important factors for recovery: connectedness, hope and optimism about the future, identity, meaning in life, and empowerment. In our study, some of the dimensions reported by facilitators appear to correlate with those of recovery in the CHIME model. i) The connectedness dimension seems to be found in the connection to others and in particular to peers – both fostered by the soothing and the source of soothing – based on the feeling of belonging to a group of peers and the support provided by them. ii) The hope dimension is represented by the presence of peers in the recovery process and fostered by the sharing of tools that enable people to manage their daily lives and thus have a more fulfilling outlook on life. iii) The identity dimension is present through the notions of destigmatization and the sharing of experiences among peers, regardless of their age, which allows younger people to project themselves in a more positive way, despite the limits imposed by the disease; this last point is a constitutive element of Anthony's definition of recovery . iv) The meaning in life dimension was more rarely mentioned by the FG participants. We formulate the hypothesis that the group situation is perhaps less efficient on this intimate dimension, and could be favoured by individual support for life projects centred on the person's objectives. v) The empowerment dimension was one of the main dimensions reported both in terms of the role of the TPE sessions and in terms of the changes observed. Thus, all our results are in favor of the action of TPE in the recovery process according to Leamy and her colleagues. Another qualitative study carried out in our department in the same time regarding patients participating in TPE programs confirmed the benefits of the TPE program on the personal recovery of people with bipolar disorder and indicated that it improved all dimensions of recovery in the CHIME model . Duval and her colleagues also stated that sharing experiences among peers promotes knowledge of the illness, a sense of being understood by others, and a sense of belonging to a group. These results are consistent, showing that patients' experience is similar to that of caregivers regarding the role and effects of the TPE program on elements of recovery in patients with bipolar disorder. Posture of the facilitators The data from our study highlighted the importance of the posture of the TPE facilitator, which seems to involve continuous oscillation between being an expert (medical, experiential) and engaging in collaborative partnership aiming at horizontality. This notion of horizontality is taken up in numerous works. According to Lang , facilitating a session leads to. "renouncing the omnipotence linked to one's knowledge in front of the patient, who was previously considered incompetent to manage his or her illness". To move toward a partnership, it is necessary to build a framework of security and benevolence, as confirmed by Le Rhun . Viard specified that the relationship. "must be based on exchanges, trust and the absence of hierarchy" to move toward a partnership . The recognition of the patient as a full person was discussed during our FG; according to Petré, this recognition fosters a sense of equality between the patient and the health professional as well as a more horizontal and less patriarchal relationship . Finally, in our study, the verbatim interview transcripts of the program facilitators emphasize the essential place of sharing experiences between peers, whether they are participants or facilitators, in the conduct of the sessions. This observation highlighted the value of the use of self-disclosure by caregivers in psychiatric care . Our study showed that the facilitators respected both the scientific and national recommendations for their position as facilitators. This posture as a TPE caregiver-facilitator appeared to be shared by many authors, beyond the strict field of psychosocial rehabilitation psychiatric care. This cannot be explained by the mere fact that these professionals carried out a partial or complete part of their activity in the rehabilitation psychosocial department with a so-called "recovery-oriented practice". Indeed, according to Lang , the thirteen principles for a psychosocial rehabilitation practice set out by Cnaan correlated with those of TPE in France: full human capacities, equipment of people with skills, self-determination, normalization, differential needs and care, commitment of staff, deprofessionalization of service, early intervention, environmental approach, changes in the environment, a lack of limits on participation, work-centered process, and social rather than medical supremacy. For Lagger, TPE promotes a change in posture and offers the opportunity to strengthen the bond, especially for the patient to be able to feel less alone with his or her chronic illness . We can formulate the hypothesis that this change in posture favored by the TPE as a caregiver-facilitator could be a tool favoring a new caregiver posture as a whole and could thereby contribute through its development to psychiatric care said to be "oriented toward the recovery of people". A 2017 study explored the skills of caregivers and supports for development of recovery-supportive care. Some caregiving interventions are reported as contributing: goal setting, conversing, early intervention and anxiety management strategies; our work suggests that TPE could help in these transformations of practices and caregiving posture . These results are similar to a recent review concerning the "recovery colleges" . These community psycho-educational practices, open to all, have shown very similar effects on the caregivers: increased motivation to work, destigmatization of users, change in power dynamics. To confirm this hypothesis, it would be important to carry out quantitative studies specifically focused on the effectiveness of TPE programs for people with bipolar disorder on their personal recovery.
Main results Our study examined the experiences of caregiver facilitators of TPE groups in the specific area of bipolar disorder in psychiatry. The facilitators emphasized that their particular posture in the TPE program involved horizontality, a feeling of belonging to a common group with the participating patients, the sharing of experiences and even self-disclosure. They also said that they observed the effects of these programs on the participants; in addition to mutual knowledge, they insisted on the notions of empowerment and destigmatization. Mutual identification with peer helpers seemed to favor identity reconstruction, or at least a form of appeasement.
Our study examined the experiences of caregiver facilitators of TPE groups in the specific area of bipolar disorder in psychiatry. The facilitators emphasized that their particular posture in the TPE program involved horizontality, a feeling of belonging to a common group with the participating patients, the sharing of experiences and even self-disclosure. They also said that they observed the effects of these programs on the participants; in addition to mutual knowledge, they insisted on the notions of empowerment and destigmatization. Mutual identification with peer helpers seemed to favor identity reconstruction, or at least a form of appeasement.
Personal recovery from psychiatric disorders is a journey toward a satisfying and hopeful life despite the possible persistence of symptoms. It is described as. “a deeply personal, unique process of changing one’s attitudes, values, feelings, goals, skills and/or roles… a way of living a satisfying, hopeful and contributing life even with the limitations caused by illness” . The CHIME framework was developed based on a literature review conducted by Leamy and her colleagues , who identified five processes as important factors for recovery: connectedness, hope and optimism about the future, identity, meaning in life, and empowerment. In our study, some of the dimensions reported by facilitators appear to correlate with those of recovery in the CHIME model. i) The connectedness dimension seems to be found in the connection to others and in particular to peers – both fostered by the soothing and the source of soothing – based on the feeling of belonging to a group of peers and the support provided by them. ii) The hope dimension is represented by the presence of peers in the recovery process and fostered by the sharing of tools that enable people to manage their daily lives and thus have a more fulfilling outlook on life. iii) The identity dimension is present through the notions of destigmatization and the sharing of experiences among peers, regardless of their age, which allows younger people to project themselves in a more positive way, despite the limits imposed by the disease; this last point is a constitutive element of Anthony's definition of recovery . iv) The meaning in life dimension was more rarely mentioned by the FG participants. We formulate the hypothesis that the group situation is perhaps less efficient on this intimate dimension, and could be favoured by individual support for life projects centred on the person's objectives. v) The empowerment dimension was one of the main dimensions reported both in terms of the role of the TPE sessions and in terms of the changes observed. Thus, all our results are in favor of the action of TPE in the recovery process according to Leamy and her colleagues. Another qualitative study carried out in our department in the same time regarding patients participating in TPE programs confirmed the benefits of the TPE program on the personal recovery of people with bipolar disorder and indicated that it improved all dimensions of recovery in the CHIME model . Duval and her colleagues also stated that sharing experiences among peers promotes knowledge of the illness, a sense of being understood by others, and a sense of belonging to a group. These results are consistent, showing that patients' experience is similar to that of caregivers regarding the role and effects of the TPE program on elements of recovery in patients with bipolar disorder.
The data from our study highlighted the importance of the posture of the TPE facilitator, which seems to involve continuous oscillation between being an expert (medical, experiential) and engaging in collaborative partnership aiming at horizontality. This notion of horizontality is taken up in numerous works. According to Lang , facilitating a session leads to. "renouncing the omnipotence linked to one's knowledge in front of the patient, who was previously considered incompetent to manage his or her illness". To move toward a partnership, it is necessary to build a framework of security and benevolence, as confirmed by Le Rhun . Viard specified that the relationship. "must be based on exchanges, trust and the absence of hierarchy" to move toward a partnership . The recognition of the patient as a full person was discussed during our FG; according to Petré, this recognition fosters a sense of equality between the patient and the health professional as well as a more horizontal and less patriarchal relationship . Finally, in our study, the verbatim interview transcripts of the program facilitators emphasize the essential place of sharing experiences between peers, whether they are participants or facilitators, in the conduct of the sessions. This observation highlighted the value of the use of self-disclosure by caregivers in psychiatric care . Our study showed that the facilitators respected both the scientific and national recommendations for their position as facilitators. This posture as a TPE caregiver-facilitator appeared to be shared by many authors, beyond the strict field of psychosocial rehabilitation psychiatric care. This cannot be explained by the mere fact that these professionals carried out a partial or complete part of their activity in the rehabilitation psychosocial department with a so-called "recovery-oriented practice". Indeed, according to Lang , the thirteen principles for a psychosocial rehabilitation practice set out by Cnaan correlated with those of TPE in France: full human capacities, equipment of people with skills, self-determination, normalization, differential needs and care, commitment of staff, deprofessionalization of service, early intervention, environmental approach, changes in the environment, a lack of limits on participation, work-centered process, and social rather than medical supremacy. For Lagger, TPE promotes a change in posture and offers the opportunity to strengthen the bond, especially for the patient to be able to feel less alone with his or her chronic illness . We can formulate the hypothesis that this change in posture favored by the TPE as a caregiver-facilitator could be a tool favoring a new caregiver posture as a whole and could thereby contribute through its development to psychiatric care said to be "oriented toward the recovery of people". A 2017 study explored the skills of caregivers and supports for development of recovery-supportive care. Some caregiving interventions are reported as contributing: goal setting, conversing, early intervention and anxiety management strategies; our work suggests that TPE could help in these transformations of practices and caregiving posture . These results are similar to a recent review concerning the "recovery colleges" . These community psycho-educational practices, open to all, have shown very similar effects on the caregivers: increased motivation to work, destigmatization of users, change in power dynamics. To confirm this hypothesis, it would be important to carry out quantitative studies specifically focused on the effectiveness of TPE programs for people with bipolar disorder on their personal recovery.
Our exploratory, single-center qualitative study was conducted using an FG. This method has limitations regarding the ability of each participant to express himself or herself in a group, to express his or her own opinions and to ignore the hierarchical functional relationships between the participants, which may have led to normative responses. In particular, we noted that during this focus group, we did not hear any expression of a negative experience of this group facilitation. The choice of sampling represents another limitation: in our study, a convenience sample was used. Additionally, the FG participants belonged to the same care unit and shared the same philosophy of care. Therefore, the results may not be representative of all TPE facilitators. It would be desirable to carry out further work exploring the experiences of facilitators in a larger number of different centers; this was not easily achievable at the beginning of our study, as very few centers had implemented similar programs. To strengthen these results, it would be interesting to use qualitative methods to reach data saturation and to combine the experience of facilitators and patients. The modalities of FG facilitation could have potentially guided the participants' responses. Our study does not specify the perceived effects on participants by characteristics such as severity or duration of the disorder, or by demographic characteristics. Studies exploring quantitative or qualitative variables on these subpopulations could contribute to our knowledge. The results of our study allow us to formulate the hypothesis that the facilitation of TPE programs could allow for a change in caregiver posture and peer-to-peer exchanges that would promote the personal recovery of people living with bipolar disorder. To confirm this hypothesis, it would also be important to carry out quantitative studies specifically focused on the effectiveness of TPE programs for the personal recovery of people with bipolar disorder. Such studies specific to TPE programs in bipolar disorder have not been published to date. There are several prerequisites for such studies, especially the choice of a relevant outcome and a suitable design. Indeed, it would be necessary to be able to precisely measure personal recovery in individuals living with bipolar disorder, as well as its determinants, to better understand how participation in the TPE program could sustain recovery. The implementation of a randomized controlled trial poses other problems in regard to psychotherapeutic actions, particularly concerning the question of the composition of the control group.
The objective of our study was to explore the TPE experience from the perspective of the caregiver-facilitators. We were able to show that the posture sought by the facilitators corresponded to the national recommendations and allowed for the establishment of a particular horizontalized relationship of collaborative partnership between the facilitators and the program participants. With regards to the changes observed in the program participants by the facilitators, it appeared that they corresponded to almost all the dimensions of the CHIME recovery model, since the study found the dimensions of Connectedness, Hope, Identity and Empowerment. Further qualitative and quantitative studies would be needed to strengthen these findings.
|
Aqueous habitats and carbon inputs shape the microscale geography and interaction ranges of soil bacteria | bef66680-c091-4c02-9509-8569549c524a | 10039866 | Microbiology[mh] | The focus on large-scale biogeographic patterns – has revealed important drivers for soil microbial abundance , and diversity , . Soil bacteria rank high in the global biomass distribution and provide crucial ecosystem functions , , . Interactions between soil properties, precipitation , temperature, and associated factors such as vegetation-derived primary productivity , exert significant control over macroscopic measures such as bacterial abundance or diversity. Yet, how bacterial populations are spatially organized in soil pore spaces and how they interact at small scales relevant to their life strategies – remain largely unknown. Elucidating typical ranges of bacterial interactions at the cell or colony scales as shaped by their physical environment can benefit the interpretation of measurements made at coarser scales (bulk samples or soil profiles) and provide mechanistic insights into the functioning of the soil microbiome in different biomes – . Evidence suggests that bacterial cells are spatially aggregated and exhibit highly localized activity in soil , and marine sediments . Complex pore spaces with a large specific surface area are characteristic of soil bacterial habitats , yet bacterial cells occupy less than 1% of the available surface area , . Soils of temperate regions host bacterial cell densities between 10 7 and 10 10 cells per gram of soil . Bacterial cells tend to aggregate near living plant roots or the active rhizosphere , and in the detritusphere around plant-derived particulate organic matter (POM). Bioturbation and root growth are the primary mechanisms for introducing POM into the soil. The soil aqueous phase connectivity shapes diffusive transport, cell dispersion, and access to patchy resources , . In water-replete environments, soil bacteria may attain cell densities similar to those found in biofilms , that host diverse communities with structural properties reminiscent of urbanization . A few dense bacterial “megacities” may constitute a substantial proportion of the bacterial biomass. The rest of the soil bacterial population is distributed across numerous “settlements” containing 10–100 cells each . Many standard soil microbiome analyses use centimeter-scale bulk samples that potentially mix spatially isolated and distinct populations , . This limits the attribution of soil microbiome functioning to properties such as bacterial biomass, diversity, and bulk activity. Although direct observations in soil are limited, we expect that the vast number of soil bacterial species and their complex biochemical functions , are spatially distributed and interact via dynamically connected aqueous microhabitats . Processes such as the development of nanotube infrastructure , electron transport , gene transfer , and cell–cell signaling require proximity among bacterial cells. The aqueous phase surrounding these cells facilitates diffusion of substances (e.g., metabolites, antibiotics, quorum sensing) and likely constrains the expression of various general traits , , . The spatial distribution of soil POM and average aqueous-phase connectivity jointly affect ecological processes that involve resource and metabolite exchanges between bacterial populations , . Characteristic diffusion times for mediating metabolic interactions can vary from seconds to months between wet and dry soils, respectively (Supplementary Note ). Thus, distances across which soil bacterial cells can effectively exchange metabolites and information modulates the functioning of spatially distributed bacterial communities in soil . Despite this critical aspect of bacterial life in soil, we lack understanding regarding of bacterial interactions and exchanges embedded within complex soil microarchitecture. Here, we introduce a quantitative framework that considers distinct levels of resource availability and soil aqueous phase connectivity to estimate effective spatial ranges of soil bacterial interactions. Limitations to direct observations of soil-bacterial distributions and interactions at the microscale, serve as the impetus to advance mechanistic biophysical models that bridge this knowledge gap and can provide insights into soil bacterial micro-geography. We employ a spatially explicit individual-based model (SIM) that simulates bacterial growth on hydrated soil surfaces to obtain spatial distributions of bacterial cells. The specific objectives of this study were: (i) to quantify the spatial variation of soil bacterial cell density based on biome-specific soil carrying capacity and aqueous phase connectivity; (ii) to link bacterial cell cluster size distribution to sample-scale (bulk) cell density; and (iii) to demonstrate how cell cluster size variations could affect bacterial interactions and soil ecosystem functioning. Supported by few observations and individual-based model results, we propose a bacterial interaction heuristic model (BIHM) that relies on well-established aggregation statistics of ecological populations – to predict the spatial distributions of soil bacterial communities as a function of auxiliary variables. The BIHM is an analytical formulation (Eqs. , , , and ) that links biome-specific bacterial cell density determined by carbon inputs with cell cluster size distributions across different climate conditions and soil types. Two important processes in unsaturated soil are embedded in the BIHM: the dependence of bulk cell density on soil carrying capacity where access to patchy resources is mediated by diffusive transport , and the natural spatial aggregation of bacterial cells due to local cell division and growth under constrained dispersal ranges . At the centimeter scale (macroscale), the templates that govern soil bacterial communities and their interactions vary with resource abundance that shapes bacterial bulk cell density across different biomes (Fig. ). We link bulk cell density with microscopic bacterial cell cluster size distributions at sub-millimeter scales using aggregation statistical laws to connect biome-specific bacterial micro-geography with soil microbiome functioning. An important implication of bacterial cell aggregation in soil is the emergence of large cell clusters at locations with limited cell dispersal and high growth. Near resource patches, clusters of sessile cells can grow into sizable colonies in which resource gradients establish due to cellular activity. For example, anoxic microsites can arise spontaneously where diffusion-limited oxygen fluxes are depleted in the core of dense and compact bacterial cell clusters . The amount of biomass associated with putative anoxic microsites may vary with soil type and bacterial population sizes , across environmental conditions and biomes.
Soil bacterial cell density and cell cluster size distribution linked to rainfall and vegetation Nearly 35% of a biome’s net primary productivity (NPP) is partitioned into new fine roots that contribute to soil POM’s annual turnover . Around one-quarter of this belowground NPP feeds soil bacterial biomass , at densities that decline with soil depth following the vertical distribution of plant roots and other conditions in soil profiles . On average the distance to a source of POM decreases with increasing NPP (Fig. ) while the length a small molecule could diffuse for one year increases with rainfall frequency (Fig. ). Hence, the number of bacterial cells maintained around sources of POM can be calculated for a range of rainfall frequencies using information on fine root fragments and estimated diffusive distances. The average distance across which bacterial cells can intercept diffusing nutrients depends on the aqueous phase connectivity of unsaturated soils shaped by the time interval between rainfall events. Using a biomes’ mean annual NPP, we assumed that local bacterial biomass decays exponentially around each POM nucleolus with a characteristic distance defined by the effective diffusive length for a given soil type and climatic condition (Eqs. – ). The resulting bacterial bulk cell densities are described as a function of average soil and climatic conditions using the BIHM (Fig. ). For comparison, previously reported microbial biomass carbon (topsoil, n = 429) has been converted to estimates of bacterial bulk cell density (Fig. ). Independent estimates of bacterial bulk cell densities were obtained from the SIM that simulates growth and dispersal of individual bacterial cells living on hydrated soil surfaces at the submillimeter scale . Both, numerical results (SIM) and global bacterial abundance data indicate a disproportionate increase of bacterial bulk cell density with increasing soil water content (Fig. ). This average (macroscopic) cell density shapes the cell cluster size distribution on soil surfaces (Fig. ). Model predictions of bacterial cell cluster size distributions in soil We present mechanistic simulations of bacterial populations using the SIM supplemented by examples of observed spatial cell distributions from a microcosm experiment conducted for this study. Bacterial cells were assigned to a cell cluster if they were located within five micrometers distance to neighboring cells . In the microcosm experiment we observed cell cluster size distributions across different hydration and nutrient conditions (Fig. ). Although, we did not observe effects of the treatment on the distribution of cell cluster sizes, the simulations by the SIM had substantial variation in bacterial cell cluster sizes with changes in hydration conditions and bacterial cell densities (Fig. ). The available information from simulation results, soil observations and our microcosm experiments indicate that soil bacterial cell cluster sizes could be described by an exponentially truncated power law. This observed pattern was not assumed in the BIHM a priori and was evidenced by simulations and experiments (Fig. ). Similar spatial aggregation patterns have been previously observed for bacteria and other organisms , – . Thus, we accepted this as a tentative representation and estimated the exponent [12pt]{minimal}
$$b$$ b and cutoff size [12pt]{minimal}
$${n}_{c}$$ n c (Eq. ) to quantify the size distribution of bacterial cell clusters. The observed cell cluster size curves collapse onto a single relationship by rescaling with the obtained parameters ( [12pt]{minimal}
$$b$$ b and [12pt]{minimal}
$${n}_{c}$$ n c ) thus lending support for the proposed cell cluster size distribution model (Eq. ) that describes experimental data and simulation results (Fig. ). We note that the SIM makes no assumptions regarding the bacterial cell cluster size distribution. Two examples of the observed spatial cell distributions are shown for the experiment and simulations (Fig. , respectively). The cell cluster size distributions vary with the total number of individual cells, as deduced from direct measurement of cell density , . This might explain why we could not observe clear differences between the experimental treatments, which resulted in similar average cell densities and cluster sizes (Fig. ). Hence, we used the microscopic cell cluster size distributions obtained from the SIM to parametrize the BIHM (Eqs. and ; Supplementary Note ) that estimates soil cell densities based on rainfall frequency, soil type and carbon input (Eqs. – ), and considers the spatial aggregation of bacterial cells as a function of bulk cell density (Eqs. – ). The resulting bacterial cell cluster size distributions are illustrated for high and low bulk cell densities (Fig. ). For comparison, we included data from a previous study using soil thin sections with low average cell density (around 500 cells per mm 2 ; n images = 341, n cells = 46,151) and our microcosm experiment using nutrient-rich garden soil with high average cell density (around 20,000 cells per mm 2 ; n images = 90, n cells = 640,100). These observed cell cluster size distributions lend support to the BIHM results and demonstrate the transition from an exponential to a power law distribution at high cell densities. We implemented the BIHM to estimate the proportion of bacterial biomass associated with small and larger cell clusters (>100 cells) across a range of climatic water contents assuming constant mean annual temperature (MAT) and mean NPP (Fig. ). Like the well-known aridity index, climatic water content is a proxy variable that considers soil water holding capacity, rainfall frequency and potential evaporation . The enhanced soil carrying capacity of humid environments with high carbon inputs and low temperatures supports the proliferation of large bacterial cell clusters. The distribution of bacterial cell cluster sizes shapes the strength of metabolic interactions in soil To quantify the strength (or ranges) of diffusion-mediated metabolic interactions among different soil bacterial communities, we estimated average distances between bacterial cell clusters emerging for different cluster sizes and climatic water contents (Fig. ). Congruent with the assumption of spatially uniform POM distribution within a thin slab of soil at a given soil depth, the distance between cell clusters in a soil is also assumed uniform. This simplification facilitates the use of volume-averaged macroscopic quantities such as effective nutrient diffusivities and carbon input fluxes while preserving microscopic variations in cell cluster sizes and numbers. The average distance between clusters containing at least two cells was about 100 μm and did not vary much with soil wetness considering average NPP, MAT, and soil type. However, the distance between larger clusters (>100 cells) increased rapidly with reduced soil water contents. To quantify the extent of temporal separation between cell clusters, we estimated how long it would take a small molecule to diffuse across the average inter-cluster separation distance (Fig. ). This timescale increased from hours to months as the soil became drier affecting the distribution of shared resources and limiting the ranges for cell–cell interactions . We have used the mechanistic individual-based SIM to study how soil bacterial interactions depend on the distribution and connectivity of aqueous habitats and carbon inputs. For simplicity, the SIM considers a conversion of three substrates ( [12pt]{minimal}
$$A B C$$ A → B → C ) by cells of different species ranging from specialists to generalists (that may use between one to three substrates, respectively). Metabolic interactions between distinct species are based on exchanging substrates via diffusion. Thus, interactions are suppressed under dry conditions with implications for bacterial community composition and diversity . The SIM considers that substrate [12pt]{minimal}
$$A$$ A is initially supplied from a point source and diffusive fluxes vary in space depending on aqueous phase connectivity. Simulation results show how conversion from the supplied substrate [12pt]{minimal}
$$A$$ A to the end-product [12pt]{minimal}
$$C$$ C depends on water contents (Fig. ). Although enhanced interactions and metabolite exchanges under wet conditions enabled higher bacterial cell densities and species richness (Fig. ), the Shannon index decreased towards higher water contents. This highlights how community dynamics are affected by spatial partitioning under predominantly competitive interactions . The predicted bacterial richness increased with the total number of cell clusters while community evenness decreased towards wet conditions associated with higher resource fluxes , and stronger interspecific interactions (Fig. ). Large bacterial cell clusters may induce anoxic microsites across a wide range of soil hydration conditions The distribution of cell cluster sizes follows a truncated power law that naturally emerges from the SIM, which makes no assumptions about the positioning of cells. The cell cluster size distribution is more skewed in wet soils, which have a higher carrying capacity and a higher cutoff size. This implies that the largest bacterial cell clusters (megacities) become bigger and smaller cell clusters (settlements) become more numerous towards humid biomes with high carbon inputs. Thus, the few bacterial megacities may disproportionally affect cell-cell interactions and ecosystem functioning. A potentially important implication of the spatial distribution of bacterial cell cluster sizes is the emergence of anoxic microsites within hotpots of bacterial activity. Activity within compact (and dense) cell clusters can result in the depletion of oxygen fluxes supplied by diffusion via soil pores whenever cell respiration rates exceed diffusive supply rates within a cell cluster (Fig. ). For simplicity, we considered fully packed (10 9 cells per mm 3 ) spherical cell clusters embedded in unsaturated soil to estimate the minimum cluster radius required to induce an anoxic core . Assuming oxygen concentrations in the soil liquid phase reflect equilibrium with atmospheric levels (a conservative assumption for many humid regions with considerable root respiration), we quantified the proportion of biomass associated with large cell clusters that could induce anoxic conditions as a function of cell density (Fig. ). Using spatially distributed soil properties , , climate attributes , , and vegetation carbon input at 0.1° resolution , we applied the BIHM to estimate the proportion of biomass associated with anoxic cell clusters (Fig. ). We removed permafrost soils from the analysis due to stringent limitations on substrate diffusion not currently considered. The dominant land cover type for 2019 was used to compare the amount of biomass associated with anoxic cell clusters for different biomes and land use (Fig. ). The number of anoxic cell clusters per gram of soil increased from 15 in bare soil to over 5500 in closed forests. The predicted amount of bacterial biomass in anoxic cell clusters tracked the bulk cell density and was highest for closed forests followed by herbaceous wetlands. For reference, we show previously reported cell counts of anaerobic species from soils of different biomes . The values ranged between 10 4 and 10 7 cells per gram of soil and were lower than our model estimates. This was expected since not all cells of an anoxic cluster must belong to an anaerobic species. Another study reported cell counts of anaerobic species between 10 8 and 10 10 per gram of soil associated with plant residues in a rice paddy field and marks an upper bound for very wet and organic matter rich soils. The importance of our estimate is in the ubiquity of anoxic microsites even in aerated unsaturated soils, and the link to bacterial cluster size distribution that depends on macroscopic properties of the soil and its microbiome.
Nearly 35% of a biome’s net primary productivity (NPP) is partitioned into new fine roots that contribute to soil POM’s annual turnover . Around one-quarter of this belowground NPP feeds soil bacterial biomass , at densities that decline with soil depth following the vertical distribution of plant roots and other conditions in soil profiles . On average the distance to a source of POM decreases with increasing NPP (Fig. ) while the length a small molecule could diffuse for one year increases with rainfall frequency (Fig. ). Hence, the number of bacterial cells maintained around sources of POM can be calculated for a range of rainfall frequencies using information on fine root fragments and estimated diffusive distances. The average distance across which bacterial cells can intercept diffusing nutrients depends on the aqueous phase connectivity of unsaturated soils shaped by the time interval between rainfall events. Using a biomes’ mean annual NPP, we assumed that local bacterial biomass decays exponentially around each POM nucleolus with a characteristic distance defined by the effective diffusive length for a given soil type and climatic condition (Eqs. – ). The resulting bacterial bulk cell densities are described as a function of average soil and climatic conditions using the BIHM (Fig. ). For comparison, previously reported microbial biomass carbon (topsoil, n = 429) has been converted to estimates of bacterial bulk cell density (Fig. ). Independent estimates of bacterial bulk cell densities were obtained from the SIM that simulates growth and dispersal of individual bacterial cells living on hydrated soil surfaces at the submillimeter scale . Both, numerical results (SIM) and global bacterial abundance data indicate a disproportionate increase of bacterial bulk cell density with increasing soil water content (Fig. ). This average (macroscopic) cell density shapes the cell cluster size distribution on soil surfaces (Fig. ).
We present mechanistic simulations of bacterial populations using the SIM supplemented by examples of observed spatial cell distributions from a microcosm experiment conducted for this study. Bacterial cells were assigned to a cell cluster if they were located within five micrometers distance to neighboring cells . In the microcosm experiment we observed cell cluster size distributions across different hydration and nutrient conditions (Fig. ). Although, we did not observe effects of the treatment on the distribution of cell cluster sizes, the simulations by the SIM had substantial variation in bacterial cell cluster sizes with changes in hydration conditions and bacterial cell densities (Fig. ). The available information from simulation results, soil observations and our microcosm experiments indicate that soil bacterial cell cluster sizes could be described by an exponentially truncated power law. This observed pattern was not assumed in the BIHM a priori and was evidenced by simulations and experiments (Fig. ). Similar spatial aggregation patterns have been previously observed for bacteria and other organisms , – . Thus, we accepted this as a tentative representation and estimated the exponent [12pt]{minimal}
$$b$$ b and cutoff size [12pt]{minimal}
$${n}_{c}$$ n c (Eq. ) to quantify the size distribution of bacterial cell clusters. The observed cell cluster size curves collapse onto a single relationship by rescaling with the obtained parameters ( [12pt]{minimal}
$$b$$ b and [12pt]{minimal}
$${n}_{c}$$ n c ) thus lending support for the proposed cell cluster size distribution model (Eq. ) that describes experimental data and simulation results (Fig. ). We note that the SIM makes no assumptions regarding the bacterial cell cluster size distribution. Two examples of the observed spatial cell distributions are shown for the experiment and simulations (Fig. , respectively). The cell cluster size distributions vary with the total number of individual cells, as deduced from direct measurement of cell density , . This might explain why we could not observe clear differences between the experimental treatments, which resulted in similar average cell densities and cluster sizes (Fig. ). Hence, we used the microscopic cell cluster size distributions obtained from the SIM to parametrize the BIHM (Eqs. and ; Supplementary Note ) that estimates soil cell densities based on rainfall frequency, soil type and carbon input (Eqs. – ), and considers the spatial aggregation of bacterial cells as a function of bulk cell density (Eqs. – ). The resulting bacterial cell cluster size distributions are illustrated for high and low bulk cell densities (Fig. ). For comparison, we included data from a previous study using soil thin sections with low average cell density (around 500 cells per mm 2 ; n images = 341, n cells = 46,151) and our microcosm experiment using nutrient-rich garden soil with high average cell density (around 20,000 cells per mm 2 ; n images = 90, n cells = 640,100). These observed cell cluster size distributions lend support to the BIHM results and demonstrate the transition from an exponential to a power law distribution at high cell densities. We implemented the BIHM to estimate the proportion of bacterial biomass associated with small and larger cell clusters (>100 cells) across a range of climatic water contents assuming constant mean annual temperature (MAT) and mean NPP (Fig. ). Like the well-known aridity index, climatic water content is a proxy variable that considers soil water holding capacity, rainfall frequency and potential evaporation . The enhanced soil carrying capacity of humid environments with high carbon inputs and low temperatures supports the proliferation of large bacterial cell clusters.
To quantify the strength (or ranges) of diffusion-mediated metabolic interactions among different soil bacterial communities, we estimated average distances between bacterial cell clusters emerging for different cluster sizes and climatic water contents (Fig. ). Congruent with the assumption of spatially uniform POM distribution within a thin slab of soil at a given soil depth, the distance between cell clusters in a soil is also assumed uniform. This simplification facilitates the use of volume-averaged macroscopic quantities such as effective nutrient diffusivities and carbon input fluxes while preserving microscopic variations in cell cluster sizes and numbers. The average distance between clusters containing at least two cells was about 100 μm and did not vary much with soil wetness considering average NPP, MAT, and soil type. However, the distance between larger clusters (>100 cells) increased rapidly with reduced soil water contents. To quantify the extent of temporal separation between cell clusters, we estimated how long it would take a small molecule to diffuse across the average inter-cluster separation distance (Fig. ). This timescale increased from hours to months as the soil became drier affecting the distribution of shared resources and limiting the ranges for cell–cell interactions . We have used the mechanistic individual-based SIM to study how soil bacterial interactions depend on the distribution and connectivity of aqueous habitats and carbon inputs. For simplicity, the SIM considers a conversion of three substrates ( [12pt]{minimal}
$$A B C$$ A → B → C ) by cells of different species ranging from specialists to generalists (that may use between one to three substrates, respectively). Metabolic interactions between distinct species are based on exchanging substrates via diffusion. Thus, interactions are suppressed under dry conditions with implications for bacterial community composition and diversity . The SIM considers that substrate [12pt]{minimal}
$$A$$ A is initially supplied from a point source and diffusive fluxes vary in space depending on aqueous phase connectivity. Simulation results show how conversion from the supplied substrate [12pt]{minimal}
$$A$$ A to the end-product [12pt]{minimal}
$$C$$ C depends on water contents (Fig. ). Although enhanced interactions and metabolite exchanges under wet conditions enabled higher bacterial cell densities and species richness (Fig. ), the Shannon index decreased towards higher water contents. This highlights how community dynamics are affected by spatial partitioning under predominantly competitive interactions . The predicted bacterial richness increased with the total number of cell clusters while community evenness decreased towards wet conditions associated with higher resource fluxes , and stronger interspecific interactions (Fig. ).
The distribution of cell cluster sizes follows a truncated power law that naturally emerges from the SIM, which makes no assumptions about the positioning of cells. The cell cluster size distribution is more skewed in wet soils, which have a higher carrying capacity and a higher cutoff size. This implies that the largest bacterial cell clusters (megacities) become bigger and smaller cell clusters (settlements) become more numerous towards humid biomes with high carbon inputs. Thus, the few bacterial megacities may disproportionally affect cell-cell interactions and ecosystem functioning. A potentially important implication of the spatial distribution of bacterial cell cluster sizes is the emergence of anoxic microsites within hotpots of bacterial activity. Activity within compact (and dense) cell clusters can result in the depletion of oxygen fluxes supplied by diffusion via soil pores whenever cell respiration rates exceed diffusive supply rates within a cell cluster (Fig. ). For simplicity, we considered fully packed (10 9 cells per mm 3 ) spherical cell clusters embedded in unsaturated soil to estimate the minimum cluster radius required to induce an anoxic core . Assuming oxygen concentrations in the soil liquid phase reflect equilibrium with atmospheric levels (a conservative assumption for many humid regions with considerable root respiration), we quantified the proportion of biomass associated with large cell clusters that could induce anoxic conditions as a function of cell density (Fig. ). Using spatially distributed soil properties , , climate attributes , , and vegetation carbon input at 0.1° resolution , we applied the BIHM to estimate the proportion of biomass associated with anoxic cell clusters (Fig. ). We removed permafrost soils from the analysis due to stringent limitations on substrate diffusion not currently considered. The dominant land cover type for 2019 was used to compare the amount of biomass associated with anoxic cell clusters for different biomes and land use (Fig. ). The number of anoxic cell clusters per gram of soil increased from 15 in bare soil to over 5500 in closed forests. The predicted amount of bacterial biomass in anoxic cell clusters tracked the bulk cell density and was highest for closed forests followed by herbaceous wetlands. For reference, we show previously reported cell counts of anaerobic species from soils of different biomes . The values ranged between 10 4 and 10 7 cells per gram of soil and were lower than our model estimates. This was expected since not all cells of an anoxic cluster must belong to an anaerobic species. Another study reported cell counts of anaerobic species between 10 8 and 10 10 per gram of soil associated with plant residues in a rice paddy field and marks an upper bound for very wet and organic matter rich soils. The importance of our estimate is in the ubiquity of anoxic microsites even in aerated unsaturated soils, and the link to bacterial cluster size distribution that depends on macroscopic properties of the soil and its microbiome.
The highly dynamic soil aqueous phase connectivity limits dispersal ranges that, in turn, affect the spatial arrangements of bacterial cells and their traits , . Irrespective of the specific mechanisms, bacterial biomass is not uniformly distributed in soil , with important implications for metabolite exchanges and the onset of anaerobic respiration , . The spatial distribution of soil bacterial cells emphasizes the localized nature of interactions relevant to soil ecological functioning , . The size of cell clusters marks the extent of contact-dependent interactions within bacterial communities including direct cell-cell signaling , gene transfer , electron transfer , and morphological development , . It is useful to note, however, that most of the bulk soil volume is inhabited by small bacterial cell clusters with unknown spatial separation and limited interactions with neighbors. The connectivity of water on soil pore surfaces and the size of associated bacterial habitats shape resource fluxes that affect community diversity , and functioning . Even in water-replete marine sediments, a large fraction of bacteria is attached to surfaces , with a few hot spots that display distinct species abundance distributions compared to those sampled from background communities . Across terrestrial biomes, the variations in bulk cell density , , define the spatial extent of bacterial neighborhoods that interact via exchanges of metabolites. The reliance on diffusion-mediated processes for signal and metabolite exchanges offers quantifiable links between the physical micro-geography and the functional capacity of soil bacterial communities that interact via water “highways” connecting richer and larger bacterial cities. Although more direct observations are needed to validate the proposed cell cluster size distributions and their relations with bulk cell density across soils of different biomes, the quantification of soil bacterial micro-geography is a critical step toward deciphering the complexity of bacterial habitats. By incorporating the microscale spatial context of soil microbiome functioning , , , our tentative spatial distributions quantify ecological interactions beyond descriptions of well-mixed bacterial communities. The assumption of spatially uniform abundance at the sample scale that often underlies inferences of species interactions and co-occurrence requires careful evaluation as it is likely to bias the picture of bacterial life under common soil conditions. The tentative estimates presented here provide a tractable modeling approach that is based on only few basic biophysical processes. It is noteworthy that the bacterial bulk cell densities presented in our heuristic model (BIHM) reflect average soil conditions and modifications for environments with different controls over connectivity and carbon supply could be introduced (e.g., for biocrusts or the immediate rhizosphere ). Refinements to the model might consider stoichiometric limitations on carrying capacity and the use of distributed bacterial trait values (e.g., oxygen uptake rates). Interactions with other soil microorganisms such as the competition with fungi could further affect bacterial cell density in ecosystems where fungi are most prevalent (e.g., in forests). The BIHM provides a parsimonious and general basis for considering interactions of soil bacteria with other organisms relevant to soil ecosystem functioning. We could interpret bacterial cell clusters as spatially distributed hosts and foraging grounds for bacteriophages and soil fauna, respectively. The perspective presented here offers a tentative, yet unifying, framework for linking soil bacterial cell cluster sizes to spatial , – and metabolic , , interactions based on sample-scale cell density. The few large cell clusters that might develop anoxic cores are restricted to densely populated resource patches that can be decoupled from oxygen (and redox) conditions of the bulk soil . A small increase in soil anoxic volume can greatly reduce carbon mineralization and affects soil gaseous emission from the microscale , . The large disparity in cell cluster size distributions predicts that there will be only a few anoxic cell clusters under a wide range of soil conditions that drive the persistence of soil carbon , and associated greenhouse gas emissions across biomes .
Average cell density based on diffusion and distance to particulate organic matter (POM) In the bacterial interactions heuristic model (BIHM), the maximal number of cells maintained in a soil volume (“soil carrying capacity“) is linked to carbon input by vegetation and cell-specific maintenance rate that is sensitive to temperature . Bulk cell density is estimated based on yearly averaged net primary productivity ( [12pt]{minimal}
$${NPP}$$ N P P ) that enters a section of the soil profile as new roots ( [12pt]{minimal}
$$$$ ξ = 0.35) and is available to soil bacteria , ( [12pt]{minimal}
$${NP}{P}_{b,z}$$ N P P b , z , with [12pt]{minimal}
$$$$ ϵ = 0.24). The vertical distribution of carbon [12pt]{minimal}
$$f(z)$$ f z to a maximum soil depth ( [12pt]{minimal}
$${d}_{{soil}}$$ d s o i l = 1 m) is described using a log-normal distribution with μ = 0.18 and σ = 1.00 as previously reported (Eq. ). We consider a homogeneous topsoil to a depth of 0.1 m for integration. 1 [12pt]{minimal}
$$NP{P}_{b,z}= { }_{soil}}{F}_{z}= { }_{soil}}{ }_{0}^{0.1}f(z)dz$$ N P P b , z = ξ ϵ N P P d s o i l F z = ξ ϵ N P P d s o i l ∫ 0 0.1 f ( z ) d z The bulk cell density at carrying capacity [12pt]{minimal}
$${ }_{{CC}}$$ ρ C C is estimated (Eq. ) by assuming bacterial cells with mass [12pt]{minimal}
$${M}_{c}$$ M c = 10 −13 gC, maintenance rate [12pt]{minimal}
$$m$$ m = 1.5 gC gC −1 y −1 and temperature sensitivity [12pt]{minimal}
$${f}_{T}$$ f T as previously described . 2 [12pt]{minimal}
$${ }_{CC}(z,T)=_{b,z}}{{f}_{T}m{M}_{c}}$$ ρ C C ( z , T ) = N P P b , z f T m M c We assume that the main source of carbon for soil bacteria is POM (and exudation) derived from fine roots with a turnover time of one year . The yearly average volume of POM ( [12pt]{minimal}
$${V}_{{POM}}$$ V P O M ) was estimated based on [12pt]{minimal}
$${NP}{P}_{b,z}$$ N P P b , z and the density of fine roots [12pt]{minimal}
$${ }_{{FR}}$$ ρ F R = 0.5 g cm −3 . The yearly number of POM fragments [12pt]{minimal}
$${N}_{{POM}}$$ N P O M is estimated based on a fine root diameter [12pt]{minimal}
$${d}_{{FR}}$$ d F R = 0.5 mm. Assuming POM has an uniform spatial distribution at the centimeter scale, we calculate an average distance to POM [12pt]{minimal}
$${ }_{{POM}}$$ δ P O M (Eqs. – ). 3 [12pt]{minimal}
$${V}_{POM} _{b,z}}{{}_{FR}}$$ V P O M ≅ N P P b , z ρ F R 4 [12pt]{minimal}
$${N}_{POM} = _{POM}}{{d}_{FR}^{3}}$$ N P O M = V P O M d F R 3 5 [12pt]{minimal}
$${ }_{POM}={(_{soil}}{{N}_{POM}})}^{}$$ δ P O M = V s o i l N P O M 1 3 The (climatic) soil water content [12pt]{minimal}
$$$$ θ is defined as previously reported (Eqs. ). The model assumes evaporation from the soil after drainage to field capacity [12pt]{minimal}
$${ }_{{FC}}$$ θ F C (approx. half of the water content at saturation [12pt]{minimal}
$${ }_{s}$$ θ s ). The soil is left for drying over given time [12pt]{minimal}
$$t$$ t with a constant rate [12pt]{minimal}
$$$$ α (estimated using potential evapotranspiration [12pt]{minimal}
$${PET}$$ P E T ). A climatic average timescale [12pt]{minimal}
$$$$ τ over which the soil dries can be estimated as the number of consecutive dry days using a precipitation time series . The time between rainfall events during which the soil is wet is used to calculate the average number of wetting cycles per year [12pt]{minimal}
$${N}_{{cyc}}$$ N c y c (Eq. ). 6 [12pt]{minimal}
$$ ={ }_{{FC}}{e}^{- t}\,{with}\, =}{{d}_{{soil}}{ }_{{FC}}}\,{and}\,{ }_{{FC}} _{s}}{2}$$ θ = θ F C e − α t w i t h α = P E T d s o i l θ F C a n d θ F C ≅ θ s 2 7 [12pt]{minimal}
$${N}_{{cyc}}=$$ N c y c = 365 τ The total distance a small molecule could travel during a year when released from a point source of POM is related to soil effective diffusivity (Eq. ). The area explored by a particle with bulk diffusivity [12pt]{minimal}
$${D}_{0}$$ D 0 is obtained by integration over a drying cycle [12pt]{minimal}
$$$$ τ . The average diffusive distance ( [12pt]{minimal}
$${ }_{D}$$ δ D ) is then obtained using the yearly number of wetting cycles [12pt]{minimal}
$${N}_{{cyc}}$$ N c y c (Eqs. and ). 8 [12pt]{minimal}
$${D}_{e}={D}_{0}^{}}{{ }_{s}^{2}}$$ D e = D 0 θ 10 3 θ s 2 9 [12pt]{minimal}
$${A}_{D }=4 { }_{0}^{ }{D}_{e}( ,t)dt=4 _{0}{ }_{FC}^{}}{{ }_{s}^{2}}(1-{e}^{ })$$ A D τ = 4 π ∫ 0 τ D e ( θ , t ) d t = 4 π D 0 θ F C 10 3 θ s 2 3 10 α ( 1 − e − 10 3 α τ ) 10 [12pt]{minimal}
$${ }_{D}=_{cyc}{A}_{D }}$$ δ D = N c y c A D τ The total number of bacterial cells sustained within the diffusive sphere around POM is estimated considering [12pt]{minimal}
$${N}_{{POM}}$$ N P O M and [12pt]{minimal}
$${ }_{D}$$ δ D . The population of bacterial cells around a point source (of POM) is assumed to decay radially with exponential rate [12pt]{minimal}
$${ }_{D}^{-1}$$ δ D − 1 . Integration over radius [12pt]{minimal}
$$r$$ r results in the expression for cell density [12pt]{minimal}
$${ }_{c}$$ ρ c that uses number of POM sources [12pt]{minimal}
$${N}_{{POM}}$$ N P O M and carrying capacity [12pt]{minimal}
$${ }_{{CC}}$$ ρ C C (Eqs. and ). 11 [12pt]{minimal}
$$_{c}}{{ }_{CC}}=4 {N}_{POM}{ }_{0}^{ }{e}^{_{D}}}{r}^{2}dr$$ ρ c ρ C C = 4 π N P O M ∫ 0 ∞ e − r δ D r 2 d r 12 [12pt]{minimal}
$${ }_{c}={ }_{CC}8 {N}_{POM}{ }_{D}^{3}$$ ρ c = ρ C C 8 π N P O M δ D 3 Conversion of cell densities using soil particle surface area Soil bacterial bulk cell density is estimated using soil microbial biomass carbon as previously described . Bacterial cells are mostly attached to particle surfaces in soil and sediments , . The specific soil-particle surface area ( [12pt]{minimal}
$${SSA}$$ S S A ) can be estimated using clay content [12pt]{minimal}
$${f}_{{clay}}$$ f c l a y and information on the dominant clay minerals. We consider proportions of kaolinite, illite, and smectite ( [12pt]{minimal}
$${f}_{K}$$ f K , [12pt]{minimal}
$${f}_{I}$$ f I , [12pt]{minimal}
$${f}_{S}$$ f S ) obtained from global maps that dominate most of the soil clay fraction and are each associated with different surface areas ( [12pt]{minimal}
$${SA}$$ S A = 60, 200, 590 m 2 g -1 , respectively). Only a fraction of the soil pore space is considered accessible to bacterial cells. We use 0.4% of the particle surface area ( [12pt]{minimal}
$$$$ η = 0.0038 ± 0.0005, n = 6). The following Eq. is used to estimate volumetric [12pt]{minimal}
$${SSA}$$ S S A ( [12pt]{minimal}
$${SS}{A}_{v}$$ S S A v ) using soil bulk density [12pt]{minimal}
$${ }_{{soil}}$$ ρ s o i l with [12pt]{minimal}
$${f}_{{dom}}={f}_{K}+{f}_{S}+{f}_{I}$$ f d o m = f K + f S + f I . 13 [12pt]{minimal}
$$SS{A}_{v}= { }_{soil}({f}_{clay}({f}_{K}S{A}_{K}+{f}_{S}S{A}_{S}+{f}_{I}S{A}_{I}+(1-{f}_{dom}))+1.1)$$ S S A v = η ρ s o i l ( f c l a y ( f K S A K + f S S A S + f I S A I + ( 1 − f d o m ) S A ¯ ) + 1.1 ) Spatially explicit individual-based model (SIM) of bacterial growth on soil grain surfaces A spatially explicit individual-based model (SIM) was previously implemented , and used here with few modifications. Cells of multiple species (characterized by two kinetic parameters) consume carbon sources and move continuously (active swimming and passive shoving) on hydrated heterogeneous soil surfaces. The heterogeneous domain describes pore surfaces within a thin slab of soil (1 by 1 mm) with 10 µm thickness and periodic boundary conditions. The domain is partitioned using a hexagonal grid resulting in grid cells of 100 µm 2 . The water holding capacity of each grid cell was specified by sampling from a random uniform distribution and assigning a total porosity of 0.5. For different (macroscopic) water contents the total volume of water was partitioned proportional to each grid cells water holding capacity while preserving the prescribed volume of water. For each grid cell the effective water film thickness was calculated (dividing the volume of water by the area of the grid cell) and was used to calculate local diffusive fluxes and the velocity of cellular motion as previously described . In the SIM, a bacterial species is defined by two Monod parameters for each of the three nutrients (2 × 3 parameters per species) prescribed from a range of values . Here, the number of species was reduced compared to previous implementations , by coarsening the discretization of the cell physiological parameter space that resulted in an initial cell number and richness of 504 cells, one of each species. The model considers a point source of diffusible carbon localized in the center of the warped domain representing 1 mm 2 of soil particle surface area with constant concentration boundary condition ( [12pt]{minimal}
$${C}_{A}$$ C A = 25 g m −3 ). In preliminary simulations, we have tested several concentrations (0.5× [12pt]{minimal}
$${C}_{A}$$ C A , [12pt]{minimal}
$${C}_{A}$$ C A , and 2x [12pt]{minimal}
$${C}_{A}$$ C A ) and selected the concentration, which resulted in cell densities comparable to observed values. Three carbon sources ( [12pt]{minimal}
$$A$$ A , [12pt]{minimal}
$$B,$$ B , and [12pt]{minimal}
$$C$$ C with assigned yields [12pt]{minimal}
$$Y$$ Y = 0.25, 0.5, and 0.75 g cell g -1 ) represent a metabolic cascade ( [12pt]{minimal}
$$A B C$$ A → B → C ). The carbon source [12pt]{minimal}
$$A$$ A with the lowest yield was initially provided and carbon mass was conserved by assigning the “leftover” carbon to the next carbon type (e. g. [12pt]{minimal}
$$A B$$ A → B with efficiency [12pt]{minimal}
$$1-{Y}_{A}$$ 1 − Y A ). The simulations were performed for a range of hydration conditions and a duration of eight days at a one-minute time step. The total cell density, spatial coordinates of the cells, and the species identity were recorded for further analysis. Spatial cell aggregation model—cell cluster size distribution In the BIHM, an exponentially truncated power law describes cell cluster sizes assuming that individuals tend to aggregate within a finite space . This general distribution can be applied to describe animal group sizes , aggregation patterns of bacterial cells , , and the distribution of soil bacterial aqueous habitats . The probability [12pt]{minimal}
$$P(n)$$ P n of having a group of [12pt]{minimal}
$$n$$ n individuals is described as in Eq. ( ) with exponent [12pt]{minimal}
$$b$$ b and cutoff size [12pt]{minimal}
$${n}_{c}$$ n c . The normalization constant [12pt]{minimal}
$$A$$ A is given by constraint 15 using the largest observed group size [12pt]{minimal}
$${n}_{ }$$ n max . 14 [12pt]{minimal}
$$P(n)=A{n}^{-b}{e}^{-_{c}}}$$ P ( n ) = A n − b e − n n c 15 [12pt]{minimal}
$$1={ }_{1}^{{n}_{max}}P(n)$$ 1 = ∑ 1 n m a x P ( n ) Similarly, the density fluctuations in growing bacterial colonies and the distribution of bacterial cluster sizes on leaf surfaces follow such cluster statistics where [12pt]{minimal}
$${n}_{c}$$ n c is related to the total number of cells in the system . Also, we can calculate the number of cell clusters [12pt]{minimal}
$$N(n)$$ N n and cells [12pt]{minimal}
$${N}_{c}(n)$$ N c n for every size class [12pt]{minimal}
$$n$$ n using cell density [12pt]{minimal}
$${ }_{C}$$ ρ C and Eqs. ( – ). 16 [12pt]{minimal}
$$p(n)=nP(n)=A{n}^{1-b}{e}^{-_{c}}}$$ p ( n ) = n P ( n ) = A n 1 − b e − n n c 17 [12pt]{minimal}
$$N(n)={ }_{C}^{}p(n)}$$ N ( n ) = ρ C p ( n ) ∑ p ( n ) 18 [12pt]{minimal}
$${N}_{c}(n)=nN(n)$$ N c ( n ) = n N ( n ) In soils, we expect [12pt]{minimal}
$$b$$ b and [12pt]{minimal}
$${n}_{c}$$ n c to vary with aqueous phase connectivity. The relation of both parameters with water contents and cell density is not known a priori and were determined using the simulated cell cluster size distributions obtained from the SIM under varying water contents. The dependency of [12pt]{minimal}
$${n}_{c}$$ n c on cell density [12pt]{minimal}
$${ }_{c}$$ ρ c is described using Eq. ( ) (with fitting parameters [12pt]{minimal}
$${a}_{n},{b}_{n}$$ a n , b n ) in agreement with model data (Fig. ). Parameters [12pt]{minimal}
$${n}_{c}$$ n c and [12pt]{minimal}
$$b$$ b are related and Eq. ( ) is used for [12pt]{minimal}
$$b$$ b (with fitting parameters [12pt]{minimal}
$$ , ,$$ α , β , γ ). These relations are used to parametrize the cell cluster size distribution of the BIHM using the SIM results. 19 [12pt]{minimal}
$${n}_{c}({ }_{c})={a}_{n}{ }_{c}^{{b}_{n}}$$ n c ( ρ c ) = a n ρ c b n 20 [12pt]{minimal}
$$b({ }_{c})=\{1,\,b < 1\\ {(_{c}}{{n}_{c}})}^{ }+ ,\,b 1$$ b ( ρ c ) = 1 , b < 1 α ( ρ c n c ) β + γ , b ≥ 1 We assume that cell clusters are spread uniformly on the accessible soil surface area to estimate the average distance between bacterial cell clusters. This assumption is justified for soil since the distances between cell clusters is much larger than the size of the clusters. Alternatively, biomass quantiles could be calculated based on the distance to POM by setting appropriate bounds for the integration of Eq. ( ) to distinguish, for example, bulk soil cell clusters from those inhabiting rhizosphere “hot spots” . Cell cluster size-induced anoxic microsites Cell clusters can spontaneously deplete oxygen in their core if cell numbers and cell densities are high enough as previously described . The oxygen diffusing trough a dense and compact cell cluster (10 9 cells per mm 3 ) can be intercepted by metabolizing cells. We assumed a constant biomass-specific oxygen uptake rate ( [12pt]{minimal}
$${k}_{{O}_{2}}$$ k O 2 = 46 g s − 1 m −3 ), oxygen diffusivity within a cell cluster ( [12pt]{minimal}
$${D}_{{O}_{2}}$$ D O 2 = 1.12 × 10 −9 m 2 s −1 ) and saturated oxygen concentration in the liquid phase ( [12pt]{minimal}
$${S}_{{O}_{2}}$$ S O 2 = 8.24 g m −3 ). For simplicity, spherical cell clusters were considered. Using Eq. ( ), we estimated the minimum cell cluster radius that could induce anoxic conditions at the core. Combined with the cluster size distribution obtained from the BIHM we estimated the proportion of biomass associated with such anoxic clusters. For the spatial mapping we use Eq. ( ) to estimate cell density and the associated cell cluster sizes. We note that this calculation is based solely on the cluster size distribution of the BIHM and no oxygen limitation was considered in the SIM microscale simulations. The consideration of anoxic microsites induced within sizable cell clusters illustrates cell cluster size effects on soil functioning and demonstrates the utility of collecting microscopic information on bacterial populations. 21 [12pt]{minimal}
$${R}_{min}=_{{O}_{2}}{S}_{{O}_{2}}}{{k}_{{O}_{2}}}}$$ R m i n = 6 D O 2 S O 2 k O 2 Soil microcosm experiment A nutrient-rich garden soil (on ETH campus, 47°22′43.8′′ N and 8°32′53.6′′E) was sampled between 5–10 cm depth in March 2018 and subsequently sieved (<2 mm) following air drying for 3 h. The soil was incubated for 4 days at 28 °C on the porous surface model that allowed for controlled hydration conditions. The experimental setup consisted of four hydrated ceramics with three holes drilled in each that were filled with soil (4 mm diameter and 3 mm depth, three replicate samples). Four treatments were applied by independently varying hydration conditions (−35 cm and −5 cm matric potential) and nutrient concentration (autoclaved tap water and tryptic soy broth, TSB). TSB is a general-purpose liquid enrichment media that supports a wide range of bacteria and is also used as a sterility test medium. It is used here because it can support growth of a wide range of (unknown) bacterial species. After incubation, soils were stained following the manufacturer’s guidelines using SYTO9 to label DNA (Thermofisher Scientific; 3 μl were applied to each sample with a concentration of 10 μM and incubated for 20 min). For image acquisition, an epifluorescence microscope was used with a GFP filter cube (EVOS FL Auto, Life Technologies, Zug, Switzerland). Imaging was done in situ at the soil surface since the microcosm was small enough to fit on the inverted microscope (turning the ceramic upside down). For each ceramic, at least 9 images (3–4 for each hole) were taken with the stock objective (AMG, 10X LPlan FL PH; AMEP-4681) covering an area of 1167 × 876 μm 2 at a resolution of 0.91 μm. Constant light settings were used throughout the experiment (light intensity = 10; exposure = 330 ms; gain = 0 dB). Images were taken at a single plane by maximizing the area in focus (autofocus). Staining and imaging were done under suction (−50 cm matric potential) to remove excess water from the soil surface. Measurements were obtained after two and four days. Image analysis for determination of cell locations The images were analyzed in Python using the SciPy stack . Greyscale images were normalized to the range of pixel intensity (max–min). Images were denoised using (approximately) shift-invariant wavelet denoising (cycle-spinning) as implemented in the skimage function “cycle_spin” with max_shifts = 9 and wavelet denoising implemented in “denoise_wavelet” using the Haar wavelet. Images taken at 10× resolution were used to localize individual cells. First, the area of focus was detected based on singular value decomposition with a window size of 29 and retaining the 3 most significant singular values. The resulting blur map was converted to a binary mask using cross-entropy thresholding (“threshold_li” in skimage ) and corresponds to the region in focus (effectively removing parts that contain no information). Holes and objects were removed from the mask if they were smaller than 25 pixels. Positions of cells were detected using the Laplacian of Gaussian method that is capable of sub-resolution edge detection and was implemented in skimage . A range of standard deviations was considered to detect local intensity peaks ( [12pt]{minimal}
$$$$ σ = 0.4–7.8 in 40 steps). The smallest object was represented by a standard deviation of 0.5 μm and the largest 10 μm. Coordinates of each cell were only used if they lied within the area of focus as determined by the blur detection. Total cell density was obtained by dividing the total number of cells by the area in focus. The individual steps of image processing are illustrated in Fig. . Images at increased resolution were used to confirm labeling of cells in the microcosm experiment (Fig. ). Clustering of proximal cells for estimation of cell cluster size distributions Cells within a Euclidean distance of 5 μm were assigned to the same cell cluster. Agglomerative clustering (single linkage) was used where computationally feasible. For cell numbers exceeding 30,000, HDBSCAN was used instead (with similar parameters: min_cluster_size = 3, min_samples = 3, cluster_selection_epsilon = 5 μm, cluster_selection_method = “eom”). The usage of HDBSCAN did not affect the clustering considerably. Data was pooled to a single cell cluster size distribution to increase the counts of large clusters (that are unlikely to be observed within small areas). Replicate simulations of the SIM were pooled for each water content. For a previous study and our microcosm experiment, images were pooled since no substantial differences in cell density were detected across samples and treatments. The spatial aggregation model (Eq. ) was fitted to the distribution of cell cluster sizes using maximum likelihood to obtain estimates of [12pt]{minimal}
$$b$$ b and [12pt]{minimal}
$${n}_{c}$$ n c (as implemented in “powerlaw” – with parameters discrete = True, discrete_approximation = “xmax” and xmin = 1). Reporting summary Further information on research design is available in the linked to this article.
In the bacterial interactions heuristic model (BIHM), the maximal number of cells maintained in a soil volume (“soil carrying capacity“) is linked to carbon input by vegetation and cell-specific maintenance rate that is sensitive to temperature . Bulk cell density is estimated based on yearly averaged net primary productivity ( [12pt]{minimal}
$${NPP}$$ N P P ) that enters a section of the soil profile as new roots ( [12pt]{minimal}
$$$$ ξ = 0.35) and is available to soil bacteria , ( [12pt]{minimal}
$${NP}{P}_{b,z}$$ N P P b , z , with [12pt]{minimal}
$$$$ ϵ = 0.24). The vertical distribution of carbon [12pt]{minimal}
$$f(z)$$ f z to a maximum soil depth ( [12pt]{minimal}
$${d}_{{soil}}$$ d s o i l = 1 m) is described using a log-normal distribution with μ = 0.18 and σ = 1.00 as previously reported (Eq. ). We consider a homogeneous topsoil to a depth of 0.1 m for integration. 1 [12pt]{minimal}
$$NP{P}_{b,z}= { }_{soil}}{F}_{z}= { }_{soil}}{ }_{0}^{0.1}f(z)dz$$ N P P b , z = ξ ϵ N P P d s o i l F z = ξ ϵ N P P d s o i l ∫ 0 0.1 f ( z ) d z The bulk cell density at carrying capacity [12pt]{minimal}
$${ }_{{CC}}$$ ρ C C is estimated (Eq. ) by assuming bacterial cells with mass [12pt]{minimal}
$${M}_{c}$$ M c = 10 −13 gC, maintenance rate [12pt]{minimal}
$$m$$ m = 1.5 gC gC −1 y −1 and temperature sensitivity [12pt]{minimal}
$${f}_{T}$$ f T as previously described . 2 [12pt]{minimal}
$${ }_{CC}(z,T)=_{b,z}}{{f}_{T}m{M}_{c}}$$ ρ C C ( z , T ) = N P P b , z f T m M c We assume that the main source of carbon for soil bacteria is POM (and exudation) derived from fine roots with a turnover time of one year . The yearly average volume of POM ( [12pt]{minimal}
$${V}_{{POM}}$$ V P O M ) was estimated based on [12pt]{minimal}
$${NP}{P}_{b,z}$$ N P P b , z and the density of fine roots [12pt]{minimal}
$${ }_{{FR}}$$ ρ F R = 0.5 g cm −3 . The yearly number of POM fragments [12pt]{minimal}
$${N}_{{POM}}$$ N P O M is estimated based on a fine root diameter [12pt]{minimal}
$${d}_{{FR}}$$ d F R = 0.5 mm. Assuming POM has an uniform spatial distribution at the centimeter scale, we calculate an average distance to POM [12pt]{minimal}
$${ }_{{POM}}$$ δ P O M (Eqs. – ). 3 [12pt]{minimal}
$${V}_{POM} _{b,z}}{{}_{FR}}$$ V P O M ≅ N P P b , z ρ F R 4 [12pt]{minimal}
$${N}_{POM} = _{POM}}{{d}_{FR}^{3}}$$ N P O M = V P O M d F R 3 5 [12pt]{minimal}
$${ }_{POM}={(_{soil}}{{N}_{POM}})}^{}$$ δ P O M = V s o i l N P O M 1 3 The (climatic) soil water content [12pt]{minimal}
$$$$ θ is defined as previously reported (Eqs. ). The model assumes evaporation from the soil after drainage to field capacity [12pt]{minimal}
$${ }_{{FC}}$$ θ F C (approx. half of the water content at saturation [12pt]{minimal}
$${ }_{s}$$ θ s ). The soil is left for drying over given time [12pt]{minimal}
$$t$$ t with a constant rate [12pt]{minimal}
$$$$ α (estimated using potential evapotranspiration [12pt]{minimal}
$${PET}$$ P E T ). A climatic average timescale [12pt]{minimal}
$$$$ τ over which the soil dries can be estimated as the number of consecutive dry days using a precipitation time series . The time between rainfall events during which the soil is wet is used to calculate the average number of wetting cycles per year [12pt]{minimal}
$${N}_{{cyc}}$$ N c y c (Eq. ). 6 [12pt]{minimal}
$$ ={ }_{{FC}}{e}^{- t}\,{with}\, =}{{d}_{{soil}}{ }_{{FC}}}\,{and}\,{ }_{{FC}} _{s}}{2}$$ θ = θ F C e − α t w i t h α = P E T d s o i l θ F C a n d θ F C ≅ θ s 2 7 [12pt]{minimal}
$${N}_{{cyc}}=$$ N c y c = 365 τ The total distance a small molecule could travel during a year when released from a point source of POM is related to soil effective diffusivity (Eq. ). The area explored by a particle with bulk diffusivity [12pt]{minimal}
$${D}_{0}$$ D 0 is obtained by integration over a drying cycle [12pt]{minimal}
$$$$ τ . The average diffusive distance ( [12pt]{minimal}
$${ }_{D}$$ δ D ) is then obtained using the yearly number of wetting cycles [12pt]{minimal}
$${N}_{{cyc}}$$ N c y c (Eqs. and ). 8 [12pt]{minimal}
$${D}_{e}={D}_{0}^{}}{{ }_{s}^{2}}$$ D e = D 0 θ 10 3 θ s 2 9 [12pt]{minimal}
$${A}_{D }=4 { }_{0}^{ }{D}_{e}( ,t)dt=4 _{0}{ }_{FC}^{}}{{ }_{s}^{2}}(1-{e}^{ })$$ A D τ = 4 π ∫ 0 τ D e ( θ , t ) d t = 4 π D 0 θ F C 10 3 θ s 2 3 10 α ( 1 − e − 10 3 α τ ) 10 [12pt]{minimal}
$${ }_{D}=_{cyc}{A}_{D }}$$ δ D = N c y c A D τ The total number of bacterial cells sustained within the diffusive sphere around POM is estimated considering [12pt]{minimal}
$${N}_{{POM}}$$ N P O M and [12pt]{minimal}
$${ }_{D}$$ δ D . The population of bacterial cells around a point source (of POM) is assumed to decay radially with exponential rate [12pt]{minimal}
$${ }_{D}^{-1}$$ δ D − 1 . Integration over radius [12pt]{minimal}
$$r$$ r results in the expression for cell density [12pt]{minimal}
$${ }_{c}$$ ρ c that uses number of POM sources [12pt]{minimal}
$${N}_{{POM}}$$ N P O M and carrying capacity [12pt]{minimal}
$${ }_{{CC}}$$ ρ C C (Eqs. and ). 11 [12pt]{minimal}
$$_{c}}{{ }_{CC}}=4 {N}_{POM}{ }_{0}^{ }{e}^{_{D}}}{r}^{2}dr$$ ρ c ρ C C = 4 π N P O M ∫ 0 ∞ e − r δ D r 2 d r 12 [12pt]{minimal}
$${ }_{c}={ }_{CC}8 {N}_{POM}{ }_{D}^{3}$$ ρ c = ρ C C 8 π N P O M δ D 3
Soil bacterial bulk cell density is estimated using soil microbial biomass carbon as previously described . Bacterial cells are mostly attached to particle surfaces in soil and sediments , . The specific soil-particle surface area ( [12pt]{minimal}
$${SSA}$$ S S A ) can be estimated using clay content [12pt]{minimal}
$${f}_{{clay}}$$ f c l a y and information on the dominant clay minerals. We consider proportions of kaolinite, illite, and smectite ( [12pt]{minimal}
$${f}_{K}$$ f K , [12pt]{minimal}
$${f}_{I}$$ f I , [12pt]{minimal}
$${f}_{S}$$ f S ) obtained from global maps that dominate most of the soil clay fraction and are each associated with different surface areas ( [12pt]{minimal}
$${SA}$$ S A = 60, 200, 590 m 2 g -1 , respectively). Only a fraction of the soil pore space is considered accessible to bacterial cells. We use 0.4% of the particle surface area ( [12pt]{minimal}
$$$$ η = 0.0038 ± 0.0005, n = 6). The following Eq. is used to estimate volumetric [12pt]{minimal}
$${SSA}$$ S S A ( [12pt]{minimal}
$${SS}{A}_{v}$$ S S A v ) using soil bulk density [12pt]{minimal}
$${ }_{{soil}}$$ ρ s o i l with [12pt]{minimal}
$${f}_{{dom}}={f}_{K}+{f}_{S}+{f}_{I}$$ f d o m = f K + f S + f I . 13 [12pt]{minimal}
$$SS{A}_{v}= { }_{soil}({f}_{clay}({f}_{K}S{A}_{K}+{f}_{S}S{A}_{S}+{f}_{I}S{A}_{I}+(1-{f}_{dom}))+1.1)$$ S S A v = η ρ s o i l ( f c l a y ( f K S A K + f S S A S + f I S A I + ( 1 − f d o m ) S A ¯ ) + 1.1 )
A spatially explicit individual-based model (SIM) was previously implemented , and used here with few modifications. Cells of multiple species (characterized by two kinetic parameters) consume carbon sources and move continuously (active swimming and passive shoving) on hydrated heterogeneous soil surfaces. The heterogeneous domain describes pore surfaces within a thin slab of soil (1 by 1 mm) with 10 µm thickness and periodic boundary conditions. The domain is partitioned using a hexagonal grid resulting in grid cells of 100 µm 2 . The water holding capacity of each grid cell was specified by sampling from a random uniform distribution and assigning a total porosity of 0.5. For different (macroscopic) water contents the total volume of water was partitioned proportional to each grid cells water holding capacity while preserving the prescribed volume of water. For each grid cell the effective water film thickness was calculated (dividing the volume of water by the area of the grid cell) and was used to calculate local diffusive fluxes and the velocity of cellular motion as previously described . In the SIM, a bacterial species is defined by two Monod parameters for each of the three nutrients (2 × 3 parameters per species) prescribed from a range of values . Here, the number of species was reduced compared to previous implementations , by coarsening the discretization of the cell physiological parameter space that resulted in an initial cell number and richness of 504 cells, one of each species. The model considers a point source of diffusible carbon localized in the center of the warped domain representing 1 mm 2 of soil particle surface area with constant concentration boundary condition ( [12pt]{minimal}
$${C}_{A}$$ C A = 25 g m −3 ). In preliminary simulations, we have tested several concentrations (0.5× [12pt]{minimal}
$${C}_{A}$$ C A , [12pt]{minimal}
$${C}_{A}$$ C A , and 2x [12pt]{minimal}
$${C}_{A}$$ C A ) and selected the concentration, which resulted in cell densities comparable to observed values. Three carbon sources ( [12pt]{minimal}
$$A$$ A , [12pt]{minimal}
$$B,$$ B , and [12pt]{minimal}
$$C$$ C with assigned yields [12pt]{minimal}
$$Y$$ Y = 0.25, 0.5, and 0.75 g cell g -1 ) represent a metabolic cascade ( [12pt]{minimal}
$$A B C$$ A → B → C ). The carbon source [12pt]{minimal}
$$A$$ A with the lowest yield was initially provided and carbon mass was conserved by assigning the “leftover” carbon to the next carbon type (e. g. [12pt]{minimal}
$$A B$$ A → B with efficiency [12pt]{minimal}
$$1-{Y}_{A}$$ 1 − Y A ). The simulations were performed for a range of hydration conditions and a duration of eight days at a one-minute time step. The total cell density, spatial coordinates of the cells, and the species identity were recorded for further analysis.
In the BIHM, an exponentially truncated power law describes cell cluster sizes assuming that individuals tend to aggregate within a finite space . This general distribution can be applied to describe animal group sizes , aggregation patterns of bacterial cells , , and the distribution of soil bacterial aqueous habitats . The probability [12pt]{minimal}
$$P(n)$$ P n of having a group of [12pt]{minimal}
$$n$$ n individuals is described as in Eq. ( ) with exponent [12pt]{minimal}
$$b$$ b and cutoff size [12pt]{minimal}
$${n}_{c}$$ n c . The normalization constant [12pt]{minimal}
$$A$$ A is given by constraint 15 using the largest observed group size [12pt]{minimal}
$${n}_{ }$$ n max . 14 [12pt]{minimal}
$$P(n)=A{n}^{-b}{e}^{-_{c}}}$$ P ( n ) = A n − b e − n n c 15 [12pt]{minimal}
$$1={ }_{1}^{{n}_{max}}P(n)$$ 1 = ∑ 1 n m a x P ( n ) Similarly, the density fluctuations in growing bacterial colonies and the distribution of bacterial cluster sizes on leaf surfaces follow such cluster statistics where [12pt]{minimal}
$${n}_{c}$$ n c is related to the total number of cells in the system . Also, we can calculate the number of cell clusters [12pt]{minimal}
$$N(n)$$ N n and cells [12pt]{minimal}
$${N}_{c}(n)$$ N c n for every size class [12pt]{minimal}
$$n$$ n using cell density [12pt]{minimal}
$${ }_{C}$$ ρ C and Eqs. ( – ). 16 [12pt]{minimal}
$$p(n)=nP(n)=A{n}^{1-b}{e}^{-_{c}}}$$ p ( n ) = n P ( n ) = A n 1 − b e − n n c 17 [12pt]{minimal}
$$N(n)={ }_{C}^{}p(n)}$$ N ( n ) = ρ C p ( n ) ∑ p ( n ) 18 [12pt]{minimal}
$${N}_{c}(n)=nN(n)$$ N c ( n ) = n N ( n ) In soils, we expect [12pt]{minimal}
$$b$$ b and [12pt]{minimal}
$${n}_{c}$$ n c to vary with aqueous phase connectivity. The relation of both parameters with water contents and cell density is not known a priori and were determined using the simulated cell cluster size distributions obtained from the SIM under varying water contents. The dependency of [12pt]{minimal}
$${n}_{c}$$ n c on cell density [12pt]{minimal}
$${ }_{c}$$ ρ c is described using Eq. ( ) (with fitting parameters [12pt]{minimal}
$${a}_{n},{b}_{n}$$ a n , b n ) in agreement with model data (Fig. ). Parameters [12pt]{minimal}
$${n}_{c}$$ n c and [12pt]{minimal}
$$b$$ b are related and Eq. ( ) is used for [12pt]{minimal}
$$b$$ b (with fitting parameters [12pt]{minimal}
$$ , ,$$ α , β , γ ). These relations are used to parametrize the cell cluster size distribution of the BIHM using the SIM results. 19 [12pt]{minimal}
$${n}_{c}({ }_{c})={a}_{n}{ }_{c}^{{b}_{n}}$$ n c ( ρ c ) = a n ρ c b n 20 [12pt]{minimal}
$$b({ }_{c})=\{1,\,b < 1\\ {(_{c}}{{n}_{c}})}^{ }+ ,\,b 1$$ b ( ρ c ) = 1 , b < 1 α ( ρ c n c ) β + γ , b ≥ 1 We assume that cell clusters are spread uniformly on the accessible soil surface area to estimate the average distance between bacterial cell clusters. This assumption is justified for soil since the distances between cell clusters is much larger than the size of the clusters. Alternatively, biomass quantiles could be calculated based on the distance to POM by setting appropriate bounds for the integration of Eq. ( ) to distinguish, for example, bulk soil cell clusters from those inhabiting rhizosphere “hot spots” .
Cell clusters can spontaneously deplete oxygen in their core if cell numbers and cell densities are high enough as previously described . The oxygen diffusing trough a dense and compact cell cluster (10 9 cells per mm 3 ) can be intercepted by metabolizing cells. We assumed a constant biomass-specific oxygen uptake rate ( [12pt]{minimal}
$${k}_{{O}_{2}}$$ k O 2 = 46 g s − 1 m −3 ), oxygen diffusivity within a cell cluster ( [12pt]{minimal}
$${D}_{{O}_{2}}$$ D O 2 = 1.12 × 10 −9 m 2 s −1 ) and saturated oxygen concentration in the liquid phase ( [12pt]{minimal}
$${S}_{{O}_{2}}$$ S O 2 = 8.24 g m −3 ). For simplicity, spherical cell clusters were considered. Using Eq. ( ), we estimated the minimum cell cluster radius that could induce anoxic conditions at the core. Combined with the cluster size distribution obtained from the BIHM we estimated the proportion of biomass associated with such anoxic clusters. For the spatial mapping we use Eq. ( ) to estimate cell density and the associated cell cluster sizes. We note that this calculation is based solely on the cluster size distribution of the BIHM and no oxygen limitation was considered in the SIM microscale simulations. The consideration of anoxic microsites induced within sizable cell clusters illustrates cell cluster size effects on soil functioning and demonstrates the utility of collecting microscopic information on bacterial populations. 21 [12pt]{minimal}
$${R}_{min}=_{{O}_{2}}{S}_{{O}_{2}}}{{k}_{{O}_{2}}}}$$ R m i n = 6 D O 2 S O 2 k O 2
A nutrient-rich garden soil (on ETH campus, 47°22′43.8′′ N and 8°32′53.6′′E) was sampled between 5–10 cm depth in March 2018 and subsequently sieved (<2 mm) following air drying for 3 h. The soil was incubated for 4 days at 28 °C on the porous surface model that allowed for controlled hydration conditions. The experimental setup consisted of four hydrated ceramics with three holes drilled in each that were filled with soil (4 mm diameter and 3 mm depth, three replicate samples). Four treatments were applied by independently varying hydration conditions (−35 cm and −5 cm matric potential) and nutrient concentration (autoclaved tap water and tryptic soy broth, TSB). TSB is a general-purpose liquid enrichment media that supports a wide range of bacteria and is also used as a sterility test medium. It is used here because it can support growth of a wide range of (unknown) bacterial species. After incubation, soils were stained following the manufacturer’s guidelines using SYTO9 to label DNA (Thermofisher Scientific; 3 μl were applied to each sample with a concentration of 10 μM and incubated for 20 min). For image acquisition, an epifluorescence microscope was used with a GFP filter cube (EVOS FL Auto, Life Technologies, Zug, Switzerland). Imaging was done in situ at the soil surface since the microcosm was small enough to fit on the inverted microscope (turning the ceramic upside down). For each ceramic, at least 9 images (3–4 for each hole) were taken with the stock objective (AMG, 10X LPlan FL PH; AMEP-4681) covering an area of 1167 × 876 μm 2 at a resolution of 0.91 μm. Constant light settings were used throughout the experiment (light intensity = 10; exposure = 330 ms; gain = 0 dB). Images were taken at a single plane by maximizing the area in focus (autofocus). Staining and imaging were done under suction (−50 cm matric potential) to remove excess water from the soil surface. Measurements were obtained after two and four days.
The images were analyzed in Python using the SciPy stack . Greyscale images were normalized to the range of pixel intensity (max–min). Images were denoised using (approximately) shift-invariant wavelet denoising (cycle-spinning) as implemented in the skimage function “cycle_spin” with max_shifts = 9 and wavelet denoising implemented in “denoise_wavelet” using the Haar wavelet. Images taken at 10× resolution were used to localize individual cells. First, the area of focus was detected based on singular value decomposition with a window size of 29 and retaining the 3 most significant singular values. The resulting blur map was converted to a binary mask using cross-entropy thresholding (“threshold_li” in skimage ) and corresponds to the region in focus (effectively removing parts that contain no information). Holes and objects were removed from the mask if they were smaller than 25 pixels. Positions of cells were detected using the Laplacian of Gaussian method that is capable of sub-resolution edge detection and was implemented in skimage . A range of standard deviations was considered to detect local intensity peaks ( [12pt]{minimal}
$$$$ σ = 0.4–7.8 in 40 steps). The smallest object was represented by a standard deviation of 0.5 μm and the largest 10 μm. Coordinates of each cell were only used if they lied within the area of focus as determined by the blur detection. Total cell density was obtained by dividing the total number of cells by the area in focus. The individual steps of image processing are illustrated in Fig. . Images at increased resolution were used to confirm labeling of cells in the microcosm experiment (Fig. ).
Cells within a Euclidean distance of 5 μm were assigned to the same cell cluster. Agglomerative clustering (single linkage) was used where computationally feasible. For cell numbers exceeding 30,000, HDBSCAN was used instead (with similar parameters: min_cluster_size = 3, min_samples = 3, cluster_selection_epsilon = 5 μm, cluster_selection_method = “eom”). The usage of HDBSCAN did not affect the clustering considerably. Data was pooled to a single cell cluster size distribution to increase the counts of large clusters (that are unlikely to be observed within small areas). Replicate simulations of the SIM were pooled for each water content. For a previous study and our microcosm experiment, images were pooled since no substantial differences in cell density were detected across samples and treatments. The spatial aggregation model (Eq. ) was fitted to the distribution of cell cluster sizes using maximum likelihood to obtain estimates of [12pt]{minimal}
$$b$$ b and [12pt]{minimal}
$${n}_{c}$$ n c (as implemented in “powerlaw” – with parameters discrete = True, discrete_approximation = “xmax” and xmin = 1).
Further information on research design is available in the linked to this article.
Peer Review File Supplementary Information Description of Additional Supplementary Files Supplementary Data 1 Reporting Summary
|
The effect of hyperbaric oxygen therapy on the clinical outcomes of necrotizing soft tissue infections: a systematic review and meta-analysis | 28019839-14ff-4e28-a944-42a239812351 | 10040118 | Debridement[mh] | Necrotizing soft tissue infections (NSTI), also known as necrotizing fasciitis (NF), are a rare but serious type of infection that can rapidly progress and lead to life-threatening consequences if not promptly and aggressively treated [ – ]. NSTI can be secondary to any skin injury or blood-borne transmission, such as postoperative skin biopsy, lacerated wounds, insect bites, pricking wounds, burns, surgical wounds, skin abscesses, herpes zoster, and venous ulceration . Due to the inconsistency between the early local symptoms and the systemic symptoms and the lack of specificity in the clinical presentation, NSTI is easily misdiagnosed in clinical practice. The early stages of NSTI may not be evident, but the condition can deteriorate rapidly within hours. The major systemic symptoms can include sustained fever, tachycardia, insufficient circulatory volume, hypoproteinemia, electrolyte disturbances, hyperglycemia, etc. If treatment is not timely, it can lead to septicemia, infectious shock, multiple organ dysfunction syndrome (MODS) or even death . Regardless of the underlying cause, NSTI demands prompt and comprehensive surgical removal of damaged tissue, antibiotics that are effective against a wide range of bacteria, and intense supportive care . NSTI differs from other soft tissue infections in that it can spread quickly through the subcutaneous tissue and fascia and has a high mortality rate, which has been estimated to be between 20 and 30%, or even higher [ – ]. Given the high mortality rate of NSTI, the use of effective adjuvant therapies to improve treatment outcomes is warranted. Hyperbaric oxygen therapy (HBO) is one of these modalities . HBO has been used to treat various conditions for over 50 years, starting with Brummelkamp's finding that hyperbaric oxygen conditions can suppress anaerobic infections . HBO has a bacteria-killing effect on anaerobic infections and has been demonstrated to improve tissue perfusion, promote angiogenesis, increase the oxygen level in tissues, and inhibit toxin production . It has also been used to treat mixed infections, including NSTI. The high-oxygen environment created by HBO can act as a barrier to prevent the spread of infection in NSTI . An expert consensus from China recommends HBO as an adjunctive therapy due to its ability to improve oxygen delivery to local tissues and increase survival rates, and provide favorable conditions for wound healing . However, some societies such as the Infectious Disease Society of America recommend against its use . An international multi-society document of skin and soft-tissue infections (SSTIs) in 2022 points that the role of HBO as an adjunctive treatment has been debated. There is currently no valid research evidence or published prospective randomized clinical trials (RCTs) that examine the impact of HBO on wound healing . Therefore, research progress on NSTI has become extremely significant, and close attention should be paid. Given the rarity and seriousness of NSTI, and the absence of evidence-based guidance on using HBO in its treatment, we carried out a systematic review and meta-analysis to assess the impact of HBO on the clinical outcomes of NSTI and provide evidence-based guidance for its use in this condition.
This systematic review and meta-analysis followed the Preferred Reporting Items for Systematic reviews and Meta-Analyses (PRISMA) guidelines (The PRISMA 2020 Checklist and PRISMA 2020 for Abstracts Checklist were showed in Additional file , ). The review protocol was registered in INPLASY register (INPLASY202320119). Search strategy A literature search was conducted using PubMed, Embase, Web of Science, and the Cochrane Database of Systematic Reviews from their inception to November 28, 2022 to identify relevant studies on the use of HBO in the treatment of NSTI, including NF and Fournier gangrene (FG). The search terms used were "necrotizing soft tissue infection," "necrotizing fasciitis," "Fournier gangrene," and "hyperbaric oxygen therapy." The language of the studies included in the review were restricted to English. The literature search strategy and full search string can be found in Additional file : Appendix A. Selection criteria The following criteria were used to determine eligibility for inclusion in this study: (1) Clinical trials and observational studies published before November 28, 2022; (2) Participants diagnosed with NSTI (or NF or FG); (3) Studies that compared the use of HBO with no use of HBO; (4) Studies that reported at least one outcome of interest. The following types of studies were excluded: (1) Conference abstracts, reviews, animal studies, case reports, editorials, letters, etc.; (2) Duplicate studies; (2) Full text unavailable; (3) Studies from which data could not be extracted; (4) Studies with inappropriate outcomes; (5) Studies with low sample sizes (total n < 10). Two reviewers (CH and YZ) independently reviewed candidate studies by screening title and abstract, and identified the studies which met the inclusion criteria. In the event of uncertainty, the eligibility of a study was discussed between the two reviewers (CH and YZ), and any disagreements were resolved depending on the third independent reviewer (BH). Data extraction The following data were extracted from included studies, if available: first author, year of publication, study design, country or region of the study, sample size, mean or median age, sex, body regions affected, confounders and the outcome of interest. Statistical analysis Qualitative synthesis Two reviewers independently evaluated the characteristics and quality of the included studies using the Newcastle Ottawa Scale . Any discrepancies were resolved through discussion and further review. Quantitative synthesis Relative risks (RRs) or standardized mean differences (SMDs) with 95% confidence intervals (CIs) were calculated for dichotomous and continuous outcomes, respectively. As clinical heterogeneity and methodological heterogeneity are inevitable at any time, we performed a meta-analysis using a random effect model. Sources of bias Publication bias was evaluated by visual inspection of funnel plots. Subgroup analyses Subgroup analyses were conducted based on pathological entity. Statistical software All statistical analyses were carried out using R software (version 4.0.2). A p value ≤ 0.05 was considered statistically significant. Evidence certainty The Grading of Recommendations Assessment, Development and Evaluation (GRADE) system was used to access the overall certainty of evidence. By GRADE system, the certainty of evidence derived from cohort studies receive an initial grade of low quality. The quality of evidence from cohort studies can be improved at larger effect sizes (RR ≥ 2 or ≤ 0. 5), dose–response gradients, or attenuation by plausible confounding after excluding various factors that could lead to downgrading. Finally, the evidence of outcomes can be graded as being of high, moderate, low, or very low.
A literature search was conducted using PubMed, Embase, Web of Science, and the Cochrane Database of Systematic Reviews from their inception to November 28, 2022 to identify relevant studies on the use of HBO in the treatment of NSTI, including NF and Fournier gangrene (FG). The search terms used were "necrotizing soft tissue infection," "necrotizing fasciitis," "Fournier gangrene," and "hyperbaric oxygen therapy." The language of the studies included in the review were restricted to English. The literature search strategy and full search string can be found in Additional file : Appendix A.
The following criteria were used to determine eligibility for inclusion in this study: (1) Clinical trials and observational studies published before November 28, 2022; (2) Participants diagnosed with NSTI (or NF or FG); (3) Studies that compared the use of HBO with no use of HBO; (4) Studies that reported at least one outcome of interest. The following types of studies were excluded: (1) Conference abstracts, reviews, animal studies, case reports, editorials, letters, etc.; (2) Duplicate studies; (2) Full text unavailable; (3) Studies from which data could not be extracted; (4) Studies with inappropriate outcomes; (5) Studies with low sample sizes (total n < 10). Two reviewers (CH and YZ) independently reviewed candidate studies by screening title and abstract, and identified the studies which met the inclusion criteria. In the event of uncertainty, the eligibility of a study was discussed between the two reviewers (CH and YZ), and any disagreements were resolved depending on the third independent reviewer (BH).
The following data were extracted from included studies, if available: first author, year of publication, study design, country or region of the study, sample size, mean or median age, sex, body regions affected, confounders and the outcome of interest.
Qualitative synthesis Two reviewers independently evaluated the characteristics and quality of the included studies using the Newcastle Ottawa Scale . Any discrepancies were resolved through discussion and further review. Quantitative synthesis Relative risks (RRs) or standardized mean differences (SMDs) with 95% confidence intervals (CIs) were calculated for dichotomous and continuous outcomes, respectively. As clinical heterogeneity and methodological heterogeneity are inevitable at any time, we performed a meta-analysis using a random effect model. Sources of bias Publication bias was evaluated by visual inspection of funnel plots. Subgroup analyses Subgroup analyses were conducted based on pathological entity. Statistical software All statistical analyses were carried out using R software (version 4.0.2). A p value ≤ 0.05 was considered statistically significant. Evidence certainty The Grading of Recommendations Assessment, Development and Evaluation (GRADE) system was used to access the overall certainty of evidence. By GRADE system, the certainty of evidence derived from cohort studies receive an initial grade of low quality. The quality of evidence from cohort studies can be improved at larger effect sizes (RR ≥ 2 or ≤ 0. 5), dose–response gradients, or attenuation by plausible confounding after excluding various factors that could lead to downgrading. Finally, the evidence of outcomes can be graded as being of high, moderate, low, or very low.
Two reviewers independently evaluated the characteristics and quality of the included studies using the Newcastle Ottawa Scale . Any discrepancies were resolved through discussion and further review.
Relative risks (RRs) or standardized mean differences (SMDs) with 95% confidence intervals (CIs) were calculated for dichotomous and continuous outcomes, respectively. As clinical heterogeneity and methodological heterogeneity are inevitable at any time, we performed a meta-analysis using a random effect model.
Publication bias was evaluated by visual inspection of funnel plots.
Subgroup analyses were conducted based on pathological entity.
All statistical analyses were carried out using R software (version 4.0.2). A p value ≤ 0.05 was considered statistically significant.
The Grading of Recommendations Assessment, Development and Evaluation (GRADE) system was used to access the overall certainty of evidence. By GRADE system, the certainty of evidence derived from cohort studies receive an initial grade of low quality. The quality of evidence from cohort studies can be improved at larger effect sizes (RR ≥ 2 or ≤ 0. 5), dose–response gradients, or attenuation by plausible confounding after excluding various factors that could lead to downgrading. Finally, the evidence of outcomes can be graded as being of high, moderate, low, or very low.
Data extraction and quality assessment Systematic review process A literature search identified a total of 2349 studies, of which 1508 were removed due to duplication or overlap. An additional 750 studies were excluded after screening titles and abstracts, leaving 91 full-text studies. Of these, 68 studies that did not meet the inclusion criteria were excluded, leaving 23 studies that were eligible for inclusion in the review. Figure shows a flow chart illustrating the process of selecting publications for inclusion. Quality assessment The Newcastle–Ottawa quality assessment scale was used to evaluate the quality of the evidence. According to this scale, all of the selected studies received at least 5 stars, indicating a low to moderate risk of bias (Table ). Characteristics of eligible studies All 23 studies [ , , – ] included were retrospective studies with a total of 49,152 patients, and 1448 patients were treated HBO, versus 47,704 patients treated without HBO. Ten studies came from North America (8 from USA, 2 from Canada). Six studies came from Europe (2 from Germany, 2 from Italy, 2 from Denmark). Four studies came from Asia (1 from China, 1 from Taiwan, 1 from Singapore, and 1 from Israel). Two studies came from South America (Brazil), and only 1 study came from Oceania (Australia). 65.2% of the patients were males. The body regions infected varied among the studies, and the major body regions affected were head and neck, truncal, perianal, anorectal, perineal and genital areas. The primary outcome included the mortality rate. The secondary outcomes included the amputation rate, the number of debridement and complications. The complications in this meta-analysis were sepsis, shock, myocardial infarction, pulmonary embolism, pneumonia and MODS. The characteristics of the included studies are summarized in Table . Evidence synthesis Primary outcomes As the primary outcome, the mortality rate was reported in all included studies. The mean mortality rate in the HBO group was 10.6% [95% CI (6.7, 14.5)] and the mean mortality rate in the non-HBO group was 25.6% [95% CI (19.5, 31.7)]. The study found that the mortality rate in the HBO group was significantly lower than that in the non-HBO group [RR = 0.522, 95% CI (0.403, 0.677), p < 0.05] (Fig. ). Secondary outcomes The number of debridements was reported in 8 studies [ , , , , , , , ]. The study found that the number of debridements in the HBO group was higher than in the non-HBO group [SMD = 0.611, 95% CI (0.012, 1.211), p < 0.05] (Fig. ). The amputation rate was reported in 6 studies [ , , – , ]. The study found no statistical significance in the amputation rate between the HBO group and non-HBO group [RR = 0.836, 95% CI (0.619, 1.129), p > 0.05] (Fig. ). Complications were reported in 5 studies [ , , , , ], including sepsis, shock, myocardial infarction, pulmonary embolism, pneumonia, and MODS. Data on the incidence of MODS was available for analysis from 2 studies . The study found that the incidence of MODS in the HBO group was lower than in the non-HBO group [RR = 0.205, 95% CI (0.164, 0.256), p < 0.05]. However, there was no statistical significance in the incidence of other complications, such as sepsis, shock, myocardial infarction, pulmonary embolism, and pneumonia, between the two groups ( p > 0.05) (Fig. ). Subgroup analyses We did a subgroup analysis with pathological entity into two categories: FG subgroup and Non-FG subgroup. The FG subgroup consisted of 10 studies [ , – , , , , , , ]. The Non-FG consisted of 13 studies [ , , , – , , , , – ]. The mortality rate was significantly lower in the HBO group compared to the non-HBO group in both non-FG [RR = 0.580, 95% CI (0.436, 0.770), p < 0.05] and FG subgroups [RR = 0.389, 95% CI (0.209, 0.723), p < 0.05]. The number of debridements in the HBO group was higher than in the non-HBO group [SMD = 0.614, 95% CI (0.453, 0.775), p < 0.05] in the non-FG subgroup, while there was no statistical significance of the number of debridements between the HBO group and non-HBO group [SMD = 0.340, 95% CI (− 3.032, 3.712), p > 0.05] in the FG subgroup. There was no statistical significance in the incidence rate of sepsis between the HBO group and non-HBO group [RR = 0.800, 95% CI (0.304, 2.108), p > 0.05] in the non-FG subgroup, as well as in the FG subgroup [RR = 1.319, 95% CI (0.943, 1.843), p > 0.05]. (Table ). The forest plots of subgroup analyses are showed in Additional file . Publication bias The funnel plot did not show significant publication bias for the mortality rate ( p = 0.086). The funnel plot is shown in Fig. . Evidence certainty The GRADE level of evidence is very low for mortality rate, amputation rate, very low for number of debridement. GRADE evidence certainty for the outcomes is shown in Table .
Systematic review process A literature search identified a total of 2349 studies, of which 1508 were removed due to duplication or overlap. An additional 750 studies were excluded after screening titles and abstracts, leaving 91 full-text studies. Of these, 68 studies that did not meet the inclusion criteria were excluded, leaving 23 studies that were eligible for inclusion in the review. Figure shows a flow chart illustrating the process of selecting publications for inclusion. Quality assessment The Newcastle–Ottawa quality assessment scale was used to evaluate the quality of the evidence. According to this scale, all of the selected studies received at least 5 stars, indicating a low to moderate risk of bias (Table ). Characteristics of eligible studies All 23 studies [ , , – ] included were retrospective studies with a total of 49,152 patients, and 1448 patients were treated HBO, versus 47,704 patients treated without HBO. Ten studies came from North America (8 from USA, 2 from Canada). Six studies came from Europe (2 from Germany, 2 from Italy, 2 from Denmark). Four studies came from Asia (1 from China, 1 from Taiwan, 1 from Singapore, and 1 from Israel). Two studies came from South America (Brazil), and only 1 study came from Oceania (Australia). 65.2% of the patients were males. The body regions infected varied among the studies, and the major body regions affected were head and neck, truncal, perianal, anorectal, perineal and genital areas. The primary outcome included the mortality rate. The secondary outcomes included the amputation rate, the number of debridement and complications. The complications in this meta-analysis were sepsis, shock, myocardial infarction, pulmonary embolism, pneumonia and MODS. The characteristics of the included studies are summarized in Table .
A literature search identified a total of 2349 studies, of which 1508 were removed due to duplication or overlap. An additional 750 studies were excluded after screening titles and abstracts, leaving 91 full-text studies. Of these, 68 studies that did not meet the inclusion criteria were excluded, leaving 23 studies that were eligible for inclusion in the review. Figure shows a flow chart illustrating the process of selecting publications for inclusion.
The Newcastle–Ottawa quality assessment scale was used to evaluate the quality of the evidence. According to this scale, all of the selected studies received at least 5 stars, indicating a low to moderate risk of bias (Table ).
All 23 studies [ , , – ] included were retrospective studies with a total of 49,152 patients, and 1448 patients were treated HBO, versus 47,704 patients treated without HBO. Ten studies came from North America (8 from USA, 2 from Canada). Six studies came from Europe (2 from Germany, 2 from Italy, 2 from Denmark). Four studies came from Asia (1 from China, 1 from Taiwan, 1 from Singapore, and 1 from Israel). Two studies came from South America (Brazil), and only 1 study came from Oceania (Australia). 65.2% of the patients were males. The body regions infected varied among the studies, and the major body regions affected were head and neck, truncal, perianal, anorectal, perineal and genital areas. The primary outcome included the mortality rate. The secondary outcomes included the amputation rate, the number of debridement and complications. The complications in this meta-analysis were sepsis, shock, myocardial infarction, pulmonary embolism, pneumonia and MODS. The characteristics of the included studies are summarized in Table .
Primary outcomes As the primary outcome, the mortality rate was reported in all included studies. The mean mortality rate in the HBO group was 10.6% [95% CI (6.7, 14.5)] and the mean mortality rate in the non-HBO group was 25.6% [95% CI (19.5, 31.7)]. The study found that the mortality rate in the HBO group was significantly lower than that in the non-HBO group [RR = 0.522, 95% CI (0.403, 0.677), p < 0.05] (Fig. ). Secondary outcomes The number of debridements was reported in 8 studies [ , , , , , , , ]. The study found that the number of debridements in the HBO group was higher than in the non-HBO group [SMD = 0.611, 95% CI (0.012, 1.211), p < 0.05] (Fig. ). The amputation rate was reported in 6 studies [ , , – , ]. The study found no statistical significance in the amputation rate between the HBO group and non-HBO group [RR = 0.836, 95% CI (0.619, 1.129), p > 0.05] (Fig. ). Complications were reported in 5 studies [ , , , , ], including sepsis, shock, myocardial infarction, pulmonary embolism, pneumonia, and MODS. Data on the incidence of MODS was available for analysis from 2 studies . The study found that the incidence of MODS in the HBO group was lower than in the non-HBO group [RR = 0.205, 95% CI (0.164, 0.256), p < 0.05]. However, there was no statistical significance in the incidence of other complications, such as sepsis, shock, myocardial infarction, pulmonary embolism, and pneumonia, between the two groups ( p > 0.05) (Fig. ). Subgroup analyses We did a subgroup analysis with pathological entity into two categories: FG subgroup and Non-FG subgroup. The FG subgroup consisted of 10 studies [ , – , , , , , , ]. The Non-FG consisted of 13 studies [ , , , – , , , , – ]. The mortality rate was significantly lower in the HBO group compared to the non-HBO group in both non-FG [RR = 0.580, 95% CI (0.436, 0.770), p < 0.05] and FG subgroups [RR = 0.389, 95% CI (0.209, 0.723), p < 0.05]. The number of debridements in the HBO group was higher than in the non-HBO group [SMD = 0.614, 95% CI (0.453, 0.775), p < 0.05] in the non-FG subgroup, while there was no statistical significance of the number of debridements between the HBO group and non-HBO group [SMD = 0.340, 95% CI (− 3.032, 3.712), p > 0.05] in the FG subgroup. There was no statistical significance in the incidence rate of sepsis between the HBO group and non-HBO group [RR = 0.800, 95% CI (0.304, 2.108), p > 0.05] in the non-FG subgroup, as well as in the FG subgroup [RR = 1.319, 95% CI (0.943, 1.843), p > 0.05]. (Table ). The forest plots of subgroup analyses are showed in Additional file . Publication bias The funnel plot did not show significant publication bias for the mortality rate ( p = 0.086). The funnel plot is shown in Fig. . Evidence certainty The GRADE level of evidence is very low for mortality rate, amputation rate, very low for number of debridement. GRADE evidence certainty for the outcomes is shown in Table .
As the primary outcome, the mortality rate was reported in all included studies. The mean mortality rate in the HBO group was 10.6% [95% CI (6.7, 14.5)] and the mean mortality rate in the non-HBO group was 25.6% [95% CI (19.5, 31.7)]. The study found that the mortality rate in the HBO group was significantly lower than that in the non-HBO group [RR = 0.522, 95% CI (0.403, 0.677), p < 0.05] (Fig. ).
The number of debridements was reported in 8 studies [ , , , , , , , ]. The study found that the number of debridements in the HBO group was higher than in the non-HBO group [SMD = 0.611, 95% CI (0.012, 1.211), p < 0.05] (Fig. ). The amputation rate was reported in 6 studies [ , , – , ]. The study found no statistical significance in the amputation rate between the HBO group and non-HBO group [RR = 0.836, 95% CI (0.619, 1.129), p > 0.05] (Fig. ). Complications were reported in 5 studies [ , , , , ], including sepsis, shock, myocardial infarction, pulmonary embolism, pneumonia, and MODS. Data on the incidence of MODS was available for analysis from 2 studies . The study found that the incidence of MODS in the HBO group was lower than in the non-HBO group [RR = 0.205, 95% CI (0.164, 0.256), p < 0.05]. However, there was no statistical significance in the incidence of other complications, such as sepsis, shock, myocardial infarction, pulmonary embolism, and pneumonia, between the two groups ( p > 0.05) (Fig. ).
We did a subgroup analysis with pathological entity into two categories: FG subgroup and Non-FG subgroup. The FG subgroup consisted of 10 studies [ , – , , , , , , ]. The Non-FG consisted of 13 studies [ , , , – , , , , – ]. The mortality rate was significantly lower in the HBO group compared to the non-HBO group in both non-FG [RR = 0.580, 95% CI (0.436, 0.770), p < 0.05] and FG subgroups [RR = 0.389, 95% CI (0.209, 0.723), p < 0.05]. The number of debridements in the HBO group was higher than in the non-HBO group [SMD = 0.614, 95% CI (0.453, 0.775), p < 0.05] in the non-FG subgroup, while there was no statistical significance of the number of debridements between the HBO group and non-HBO group [SMD = 0.340, 95% CI (− 3.032, 3.712), p > 0.05] in the FG subgroup. There was no statistical significance in the incidence rate of sepsis between the HBO group and non-HBO group [RR = 0.800, 95% CI (0.304, 2.108), p > 0.05] in the non-FG subgroup, as well as in the FG subgroup [RR = 1.319, 95% CI (0.943, 1.843), p > 0.05]. (Table ). The forest plots of subgroup analyses are showed in Additional file .
The funnel plot did not show significant publication bias for the mortality rate ( p = 0.086). The funnel plot is shown in Fig. .
The GRADE level of evidence is very low for mortality rate, amputation rate, very low for number of debridement. GRADE evidence certainty for the outcomes is shown in Table .
NSTI are a type of rapidly progressing infection that can be highly destructive to the skin, subcutaneous tissue, and superficial fascia . These infections involve the breakdown of tissues and fascia and can spread along tissue planes, sometimes resulting in myonecrosis and variable involvement of the skin above . The speed at which the necrotic area progresses is thought to be around 2–3 cm/h . NSTI are also known as NF or FG, and have been referred to as "flesh-eating bug disease." Clinical features of NSTI include grey necrotic tissue, swelling of the fascia, thin, watery, foul-smelling fluid, and blocked vessels . If NSTI is not diagnosed and treated promptly, it can have serious consequences such as limb loss or death . The mortality rate of NSTI has been historically reported to be as high as 76% . Despite advances in diagnostic approaches and treatment regimens, more recent literature has reported mortality rates of NSTI to be between 9 and 25%, or even higher . NSTI often result in local tissue hypoxia. The interaction between tissue hypoxia and infections, along with postoperative incision poor drainage or other factors, can lead to rapid spread of the infection. HBO is a comprehensive treatment that uses a medical hyperbaric oxygen chamber as a carrier and oxygen as the core. Under 0.2 MPa hyperbaric oxygen, physical dissolved oxygen can increase by 17 times, muscle oxygen partial pressure increases by 8 times, and tissue oxygen partial pressure increases by 4 times. Under high pressure, the effective diffusion radius of oxygen extends and the diffusion range expands. HBO has a direct effect on anaerobic bacteria through the formation of oxygen free radicals. During phagocytosis, neutrophil oxygen consumption increases, and HBO can enhance neutrophil activity. HBO can also promote the growth of fibroblasts and the formation of blood vessels, thus promoting wound healing. HBO can alleviate inflammation, reduce inflammatory immune cytokines, stimulate wound repair, maintain wound oxygenation, increase antioxidant enzymes, and treat tissue hypoxemia and radiation necrosis [ – ]. However, there has been ongoing controversy regarding the effectiveness of HBO in terms of mortality and other clinical outcomes in patients with NSTI . Some studies have shown that HBO is significantly beneficial in these patients, while others have found the opposite . To increase the body of evidence, we carried out a systematic review and meta-analysis to compare the outcomes of NSTI patients who were treated with or without HBO. In this study, 23 eligible retrospective studies were identified, with 65.2% of the patients being male. Previous research has shown that NSTI is more common among elderly males, with a mean age of over 50 years, which is consistent with the findings of this study . The literature reports on the relationship between the incidence of NSTI in patients and gender differences vary, and may be related to the number of cases collected and regional differences. The reason for the different incidence of NSTI between men and women is not yet clear. Zhang et al. reported that this sex difference and age feature may be associated with an increase in the number of conditions that can increase the risk of NSTI. In our study, almost all patients with NSTI had comorbidities, the most common of which were diabetes mellitus, hypertension, alcoholism, smoking, obesity, anorectal diseases, renal disease, malignancy, coronary artery disease, peripheral vascular disease, intravenous drug use, and immunosuppression. Among these predisposing diseases, diabetes mellitus was identified as the most common comorbidity associated with NSTI, which may accelerate bacterial infection progression and result in a poor prognosis, even increasing the risk of mortality . Consistent with the literature, over half of our participants suffered from this comorbidity. High blood sugar is itself a good culture medium for bacteria, and NSTI complicated with diabetes is the result of the combined action of various pathogenic bacteria (aerobic bacteria, anaerobic bacteria, and fungi) . Various pathogenic bacteria can invade the subcutaneous and fascia tissue through the wound. The irritative gases, such as H2, N2, H2S, and CH4, produced by bacteria accumulate in the soft tissue. At the same time, active substances, such as alidase and heparinase, are released to decompose and destroy the tissue, causing corresponding tissue edema and ischemic necrosis . Patients with diabetes are at a higher risk for developing NSTI due to their decreased immune function and increased presence of bacteria on the skin. Diabetic patients also have a decreased ability to phagocytize and a higher potential for local bacterial proliferation, which provides conditions for the proliferation of bacteria. Thus, it is crucial to closely monitor and control blood sugar levels during the treatment of NSTI. It is generally considered that keeping blood sugar levels below 10.0 mmol/L is beneficial for controlling wound infections and granulation growth . Due to the high mortality rate associated with NSTI, we considered mortality to be the primary outcome in this study. Hollabaugh et al. reported a mortality rate of 7% for the HBO group and 42% for the Non-HBO group. Creta et al. reported that mortality due to NSTI occurred in 32 (36.0%) of patients who did not undergo HBO and in 14 (19.4%) of patients who did undergo HBO ( p = 0.01). Some studies even reported that no patients in the HBO group died . According to the results of this meta-analysis, the mean mortality rate for the HBO group was 10.6% and the mean mortality rate for the Non-HBO group was 25.6%. The mortality rate for the HBO group was significantly lower than that of the Non-HBO group. It is believed that the use of HBO may contribute to this difference by increasing oxygen transport and diffusion to injured, oedematous, and infected hypoxic tissues and by creating a high pressure of oxygen around infected tissue, which can effectively prevent the invasion of microorganisms . Additionally, HBO may narrow the affected region, prevent the extension of necrosis, reduce systemic toxicity, and decrease the mortality rate when used in conjunction with surgical debridement and broad-spectrum antibiotic therapy. The results of this study also showed that the amputation rate was not significantly different between the HBO group and the Non-HBO group. However, it is worth noting that the number of debridements performed in the HBO group was higher than that in the Non-HBO group. Similarly, Tharakaram et al. also observed a higher number of surgical debridements in the HBO group. Usually, debridement surgery has three main purposes, including: to clearly define the extent of infection; to evaluate whether debridement or amputation is necessary; to obtain samples and stain and culture them for bacterial identification. In cases of highly suspected NSTI, performing an effective exploration and sending the sample for pathological and microbiological examination remains the direct evidence for establishing a diagnosis. Therefore, all suspected cases should be promptly explored to establish a clear diagnosis. One study suggested that tissue samples taken for microbiologic analysis were counted as debridement, which may have contributed to the higher number of surgeries in the HBO group due to improved survival . However, further evidence is needed to support this conclusion. It may be inappropriate to use the number of debridements as a measure of the efficacy of HBO. Given the poor prognosis and potential for relapse associated with this disease, the survival rate with HBO based on long-term follow-up should be considered a primary outcome in future meta-analyses. In terms of complications, the results of this study showed that the incidence of MODS was lower in the HBO group than in the Non-HBO group. However, there was no significant difference in the incidence of sepsis, shock, myocardial infarction, pulmonary embolism, or pneumonia between the two groups. These results suggest that the use of HBO is generally safe. HBO can cause oxygen poisoning, sinus barotrauma, middle ear barotrauma or pulmonary barotrauma and other adverse reactions, and induce claustrophobia in severe cases. As a consequence, we should closely monitor all adverse reactions and correct them timely during the treatment of HBO for NSTI. Based on the safety, it is essential to control and regulate the pressure value and time value of HBO, in order to avoid causing serious complications. As FG and NF diseases exhibit distinct epidemiology and pathological characteristics, we conducted a subgroup analysis by categorizing them into two pathological entities. We were limited by the amount of literature available and thus only examined the mortality rate, number of debridements, and incidence rate of sepsis. Among these variables, the number of debridements did not show a significant difference between the HBO and non-HBO groups in the FG subgroup. However, since only two studies were included, further verification is required to confirm the result. Nonetheless, the other subgroup analyses were consistent with previous findings, which suggests the stability of our results. There are several limitations to be considered in this systematic review and meta-analysis of the effectiveness of HBO in patients with NSTI. Firstly, the retrospective study design may result in inconsistent data quality and availability of certain clinical and laboratory parameters. While this is a limitation, it is worth noting that it would not be ethical to deprive patients of HBO treatment in many cases, and therefore it would not be feasible to conduct a prospective randomized controlled trial . Secondly, the duration and frequency of HBO treatment varied across studies, which could potentially affect the outcomes. It is important to establish unified therapy criteria for HBO in order to ensure consistent treatment. Additionally, the pooled results may be affected by the inclusion of data from different stages of treatment and different diagnostic criteria, courses of treatment, and lengths of follow-up. It is necessary to conduct independent systematic evaluation and analysis according to the different onset sites of NSTI to improve the reliability and stability of outcomes such as mortality rate, amputation rate, and survival rate. Compared with the number of debridement, we think that survival rate should be used as the main outcome indicator for NSTI. Generally, the use of antibiotics in the treatment of NSTI is commonly longer, the scope of debridement is wide, and the healing time is long. Therefore, long-term follow-up of survival patients with NSTI treated with HBO should be carried out to observe the recurrence rate, complications, and quality of life. This should be an important direction for future research. Research on the diagnosis and treatment of NSTI has made some progress in recent years, but the mortality rates and amputation rates of NSTI have not been significantly controlled yet. The main reason is that the progression of NSTI is rapid and early identification is difficult, and the diagnostic criteria are not clear. These are the key points for future research on NSTI.
The current evidence suggests that the use of HBO in the treatment of NSTI can significantly reduce the mortality rates and the incidence rates of complications. However, due to the retrospective nature of the studies, the evidence is weak, and further research is needed to establish its efficacy. It is also important to note that HBO is not available in all hospitals, and its use should be carefully considered based on the patient's individual circumstances. Additionally, it is still worthwhile to stress the significance of promptly evaluating surgical risks to prevent missing the optimal treatment time. Given the rarity of the disease, it is essential to continue producing high-quality research to provide guidance to clinicians.
Additional file 1. PRISMA 2020 Checklist. Additional file 2. PRISMA 2020 for Abstracts Checklist. Additional file 3. The literature search strategy. Additional file 4. Forest plots of subgroup analyses.
|
Editorial: Rising stars in cardiovascular endocrinology 2022 | d9bb4ff7-ebf7-4708-89c0-54c047cf70fc | 10042285 | Physiology[mh] | The author confirms being the sole contributor of this work and has approved it for publication. |
Value of the clinical pharmacist interventions in the application of the American College of Cardiology (ACC/AHA) 2018 guideline for cholesterol management | 2d071e88-bf5f-4e61-a809-cb71a6121cfd | 10042341 | Internal Medicine[mh] | Coronary heart disease (CHD) is considered one of the main causes of morbidity and mortality worldwide. One of the major risk factors that contributes to the development and progression of CHD is dyslipidemia . According to the World Health Organization (WHO), an estimated 17.9 million patients died each year because of cardiovascular diseases, of which more than 80.0% of the deaths were due to stroke and myocardial infarction . Ischemic heart diseases and stroke are among the top 3 leading causes of years of life lost (YLL) and mortality according to a global burden disease study . Another recent study in UAE showed that the overall prevalence of dyslipidemia among adults was 72.5%, where the total cholesterol and LDL-C levels were high in 42.8% and 38.6% of the participants, respectively . In addition, a study including an expatriate population in the UAE reported a high prevalence of either overweight or obesity (75.3%) as well as known associated risk factors for developing both metabolic syndrome and dyslipidemia . A large number of clinical trials have reported the benefits of lowering cholesterol levels, particularly LDL-C, in reducing the mortality rate among CHD patients. Based on that, the American College of Cardiology (ACC/AHA) published the 2013 blood cholesterol treatment guidelines to reduce atherosclerotic cardiovascular risk in adults. This guideline has been updated several times since then . The latest update of the ACC/AHA management of blood cholesterol guideline (2018) emphasizes on the importance of categorizing patients into the four statins benefit groups and on the importance of statin therapy using evidence-based intensity level (high- or moderate-intensity statins). Furthermore, it highlights the importance of adding PCSK9 inhibitor therapy after receiving the maximum tolerated statin therapy and ezetimibe to achieve LDL-C < 70 mg/dl or non-HDL-C < 100 mg/dl for some patients . The prevalence and treatment rates of dyslipidemia are high in the UAE and worldwide , however, it was found that a significant percentage of patients worldwide were not taking appropriate lipid-lowering agents or were taking statins but were not meeting the primary treatment goal . One of the reasons identified was the low adherence to the guideline recommendations. Clinical pharmacists play important role in individualizing patient treatment and improving adherence to guideline’s recommendations. The present study aims to examine the extent to which the updated ACC/AHA management of blood cholesterol guideline (2018) is implemented in practice and to assess the value of the clinical pharmacists’ interventions in improving physicians’ adherence to the guideline’s recommendations. In UAE, the ACC / AHA guidelines are the most commonly followed and recommended guidelines by the health authorities.
Subjects and settings The study was conducted on adult patients attending an internal medicine clinic at a large hospital in Al Ain City, UAE, from January to April 2019 (n = 647). Patients’ information had been collected through Hospital Information System (HIS). The study pharmacist evaluated the data for all patients attending the internal medicine clinic on daily basis during the study period to identify eligible patients. Patients aged ≥ 21 years who met the criteria of one of the statin benefit groups requiring high- or moderate-intensity statin therapy according to the 2018 ACC/AHA guidelines were included in this study if they had no of the below exclusion criteria (number of included patients = 272, 42%). As per 2018 ACC / AHA guideline recommendations, the following were the statins benefits groups: subjects with a history of ASCVD; a high intensity statin should be considered (Class 1 recommendation). subjects with a primary elevation of LDL cholesterol ≥190 mg/dL; a high intensity statin should be considered (Class 1 recommendation). subjects with diabetes. In adults 40 to 75 years of age with diabetes mellitus, regardless of estimated 10-year ASCVD risk, moderate-intensity statin therapy is indicated (Class 1 recommendation). In adults 40 to 75 years of age with diabetes mellitus who have multiple ASCVD risk factors, it is reasonable to prescribe high intensity statin therapy (Class 2A recommendation) In adults older than 75 years with diabetes mellitus, it may be reasonable to initiate statin therapy after a clinician–patient discussion of potential benefits and risks (Class 2B recommendation, Class 2A if already on statin) For diabetic patients 20–39 years old, statin may be considered in case of presence of multiple diabetes specific risk enhancers such as long disease duration, albuminuria ≥ 30 mcg albumin/mg creatinine, eGFR < 60 ml/min/1.73 m2, neuropathy, and retinopathy (Class 2B recommendation) subjects 40–75 years old without diabetes or ASCVD, with baseline LDL cholesterol levels of 70 to 189 mg/dL. Treatment recommendations are based on the 10-year ASCVD risk as follows: Borderline risk (10-year ASCVD risk 5–7.4%): consider a moderate intensity statin if risk enhancers present (Class 2B recommendation). Intermediate risk (10-year ASCVD risk 7.5–19.9%): consider moderate intensity statin (Class 1 recommendation). High risk (10-year ASCVD risk ≥20%): consider high intensity statin (Class 1 recommendation). In addition, the guideline recommends considering treatment with statins in individuals aged 20 to 39 years old with a family history of premature ASCVD disease and a high LDL of ≥ 160 mg/dL (Class 2B recommendation). The main exclusion criteria were: 1) history of statin-induced rhabdomyolysis or myopathy; 2) history of allergic reaction to statins; 3) current active liver disease; 4) creatine kinase levels >3 times the upper limit of normal; 5) Any contraindications to statins use; 6. Patients for whom a lipid profile was not available or who did not have a sufficient data to classify them into statin benefit groups or enough information for calculating the ASCVD risk score at the time the study was conducted were excluded as well. Ethical consideration This study was approved by the Hospital Research Ethics Committee (Ref. CR /2018/40). All methods were performed in accordance with relevant guidelines and regulations. The study was explained to patients and their consent was obtained before participation. Study design and data collection We utilized in this study an interventional before-after design. Demographic and clinical characteristics of the study sample were extracted from the hospital information system. Data collected did not include any personal or sensitive information such as patient’s identity or medical record number. represents the flowchart of the study design. Data collection before clinical pharmacists’ intervention Eligibility for statin therapy was evaluated based on 2018 ACC / AHA guideline recommendations as stated above. Adherence to guideline recommendations before clinical pharmacist’s intervention was measured by calculating the percentage of patients identified per each statin benefit group, the percentage of patients receiving statin therapy as per guideline recommendation, the type and intensity (moderate or high intensity) of statin therapy used, and the need for additional non-statin therapy. We calculated only adherence to class 1 and class 2A guideline’s recommendations. Clinical pharmacists’ interventions The clinical pharmacists received the medication order for each patient after the patients’ appointment with the physician. The clinical pharmacists evaluated all patients’ data and the appropriateness of their medication order and recommended therapy modifications to meet the 2018 ACC/AHA cholesterol management guideline recommendations. Physicians were automatically notified with the pharmacist’s intervention in the online system and responded accordingly by agreeing, modifying, or rejecting the pharmacist’s recommendation. Physician response and treatment plan changes due to clinical pharmacist interventions were extracted from the Hospital Information System. Data collection after clinical pharmacists’ intervention Adherence to guideline recommendations was measured by calculating the percentage of patients identified per each statin benefit group, the percentage of patients receiving statin therapy as per guideline recommendations, the type and intensity (moderate or high intensity) of statin therapy used, and the need for additional non-statin therapy. We calculated only adherence to class 1 and class 2A recommendations. Rejecting a recommendation that is based on a class 2B guideline recommendation was not considered as nonadherent. These class 2B recommendations are usually considered by the guideline as “may be reasonable” and “weak” where their benefit is ≥ risk. Study outcomes The following outcomes were examined in this study: Adherence to the 2018 ACC /AHA guideline recommendations was measured by determining the following: The number and percentage of statin benefit group patients who were prescribed a statin Appropriateness of statin dose (high intensity vs moderate intensity) The need for additional non-statin therapy. The impact of clinical pharmacist interventions on the application of guideline recommendations was determined by measuring the following: The number and type of recommendations attained by the clinical pharmacist. Physicians’ acceptance of recommendations Comparison of differences in adherence to 2018 ACC /AHA guidelines before and after clinical pharmacist interventions. Statistical analysis All data were entered and analyzed using Statistical Package for the Social Sciences (SPSS) version 22 (IBM SPSS Statistics for Windows, Version 24.0. Armonk, NY: IBM Corp.). Descriptive statistics were used to measure the frequencies and percentages. The chi-square test was used to compare adherence to guidelines before and after clinical pharmacist interventions. A p-value of 0.05 was considered statistically significant, using a 95.0% confidence interval for differences.
The study was conducted on adult patients attending an internal medicine clinic at a large hospital in Al Ain City, UAE, from January to April 2019 (n = 647). Patients’ information had been collected through Hospital Information System (HIS). The study pharmacist evaluated the data for all patients attending the internal medicine clinic on daily basis during the study period to identify eligible patients. Patients aged ≥ 21 years who met the criteria of one of the statin benefit groups requiring high- or moderate-intensity statin therapy according to the 2018 ACC/AHA guidelines were included in this study if they had no of the below exclusion criteria (number of included patients = 272, 42%). As per 2018 ACC / AHA guideline recommendations, the following were the statins benefits groups: subjects with a history of ASCVD; a high intensity statin should be considered (Class 1 recommendation). subjects with a primary elevation of LDL cholesterol ≥190 mg/dL; a high intensity statin should be considered (Class 1 recommendation). subjects with diabetes. In adults 40 to 75 years of age with diabetes mellitus, regardless of estimated 10-year ASCVD risk, moderate-intensity statin therapy is indicated (Class 1 recommendation). In adults 40 to 75 years of age with diabetes mellitus who have multiple ASCVD risk factors, it is reasonable to prescribe high intensity statin therapy (Class 2A recommendation) In adults older than 75 years with diabetes mellitus, it may be reasonable to initiate statin therapy after a clinician–patient discussion of potential benefits and risks (Class 2B recommendation, Class 2A if already on statin) For diabetic patients 20–39 years old, statin may be considered in case of presence of multiple diabetes specific risk enhancers such as long disease duration, albuminuria ≥ 30 mcg albumin/mg creatinine, eGFR < 60 ml/min/1.73 m2, neuropathy, and retinopathy (Class 2B recommendation) subjects 40–75 years old without diabetes or ASCVD, with baseline LDL cholesterol levels of 70 to 189 mg/dL. Treatment recommendations are based on the 10-year ASCVD risk as follows: Borderline risk (10-year ASCVD risk 5–7.4%): consider a moderate intensity statin if risk enhancers present (Class 2B recommendation). Intermediate risk (10-year ASCVD risk 7.5–19.9%): consider moderate intensity statin (Class 1 recommendation). High risk (10-year ASCVD risk ≥20%): consider high intensity statin (Class 1 recommendation). In addition, the guideline recommends considering treatment with statins in individuals aged 20 to 39 years old with a family history of premature ASCVD disease and a high LDL of ≥ 160 mg/dL (Class 2B recommendation). The main exclusion criteria were: 1) history of statin-induced rhabdomyolysis or myopathy; 2) history of allergic reaction to statins; 3) current active liver disease; 4) creatine kinase levels >3 times the upper limit of normal; 5) Any contraindications to statins use; 6. Patients for whom a lipid profile was not available or who did not have a sufficient data to classify them into statin benefit groups or enough information for calculating the ASCVD risk score at the time the study was conducted were excluded as well.
This study was approved by the Hospital Research Ethics Committee (Ref. CR /2018/40). All methods were performed in accordance with relevant guidelines and regulations. The study was explained to patients and their consent was obtained before participation.
We utilized in this study an interventional before-after design. Demographic and clinical characteristics of the study sample were extracted from the hospital information system. Data collected did not include any personal or sensitive information such as patient’s identity or medical record number. represents the flowchart of the study design.
Eligibility for statin therapy was evaluated based on 2018 ACC / AHA guideline recommendations as stated above. Adherence to guideline recommendations before clinical pharmacist’s intervention was measured by calculating the percentage of patients identified per each statin benefit group, the percentage of patients receiving statin therapy as per guideline recommendation, the type and intensity (moderate or high intensity) of statin therapy used, and the need for additional non-statin therapy. We calculated only adherence to class 1 and class 2A guideline’s recommendations.
The clinical pharmacists received the medication order for each patient after the patients’ appointment with the physician. The clinical pharmacists evaluated all patients’ data and the appropriateness of their medication order and recommended therapy modifications to meet the 2018 ACC/AHA cholesterol management guideline recommendations. Physicians were automatically notified with the pharmacist’s intervention in the online system and responded accordingly by agreeing, modifying, or rejecting the pharmacist’s recommendation. Physician response and treatment plan changes due to clinical pharmacist interventions were extracted from the Hospital Information System.
Adherence to guideline recommendations was measured by calculating the percentage of patients identified per each statin benefit group, the percentage of patients receiving statin therapy as per guideline recommendations, the type and intensity (moderate or high intensity) of statin therapy used, and the need for additional non-statin therapy. We calculated only adherence to class 1 and class 2A recommendations. Rejecting a recommendation that is based on a class 2B guideline recommendation was not considered as nonadherent. These class 2B recommendations are usually considered by the guideline as “may be reasonable” and “weak” where their benefit is ≥ risk.
The following outcomes were examined in this study: Adherence to the 2018 ACC /AHA guideline recommendations was measured by determining the following: The number and percentage of statin benefit group patients who were prescribed a statin Appropriateness of statin dose (high intensity vs moderate intensity) The need for additional non-statin therapy. The impact of clinical pharmacist interventions on the application of guideline recommendations was determined by measuring the following: The number and type of recommendations attained by the clinical pharmacist. Physicians’ acceptance of recommendations Comparison of differences in adherence to 2018 ACC /AHA guidelines before and after clinical pharmacist interventions.
All data were entered and analyzed using Statistical Package for the Social Sciences (SPSS) version 22 (IBM SPSS Statistics for Windows, Version 24.0. Armonk, NY: IBM Corp.). Descriptive statistics were used to measure the frequencies and percentages. The chi-square test was used to compare adherence to guidelines before and after clinical pharmacist interventions. A p-value of 0.05 was considered statistically significant, using a 95.0% confidence interval for differences.
Demographic and clinical characteristics of the study sample The demographic and clinical characteristics of the study sample are shown in . The mean age of the studied patients was 52.6 ±10.5, and 71.3% (n = 194) of them were males. Majority of the patients (95.6%, n = 260) had previous illness or chronic disease. Out of these patients, 178 patients (65.4%) had hypertension, 150 patients (55.1%) had dyslipidemia, and 196 patients (72.1%) had diabetes mellitus. Nevertheless, 20 patients (7.4%) had LDL-C levels less than 70 mg/dl, 236 patients (86.8%) had LDL-C levels between 70–189 mg/dl, and 16 patients (5.9%) had LDL-C levels equal to or more than 190. Additionally, 52 patients (19.1%) had a history of clinical ASCVD. Of these, 38 patients (14.0%) were at very high risk of recurrent CVD. Out of all participants without a history of clinical ASCVD, 50 patients (18.4%) had an estimated 10-year CVD risk less than 7.5%, 110 patients (40.5%) had an estimated 10-year CVD risk greater than or equal to 7.5% but less than 20.0%, and 60 patients (22.1%) had an estimated 10-year CVD risk equal to or greater than 20.0%. The adherence with the 2018 ACC/AHA guideline recommendations for the management of cholesterol in adults before clinical pharmacist interventions Adherence with the 2018 ACC/AHA guideline recommendations for the management of cholesterol in adults is shown in . Based on the inclusion criteria, all the patients who were enrolled (100.0%, n = 272) were identified as statin benefit groups according to the 2018 ACC/AHA guideline recommendations. Of these, only 60.3% (n = 164) were initiated on statin therapy. Out of those who were on statin therapy, 9.8% (n = 16) were on low intensity statin (e.g., simvastatin 10 mg and pitavastatin 1 mg), 52.4% (n = 86) were on moderate intensity statin (e.g., simvastatin 40 mg, rosuvastatin 10 mg, atorvastatin 20 mg, pitavastatin 2 mg and 4 mg) and 37.8% (n = 62) were on high intensity statin (e.g., atorvastatin 40 mg and 80 mg and rosuvastatin 20 mg and 40 mg). Adherence to the recommend level of statins intensity was identified in only 47.6% of patients (n = 78). The addition of non-statin therapies to achieve LDL-C goals was also assessed, and ezetimibe was required for 51.2% (n = 84) of those who were on statin therapy. While it was initiated for only 8.5% (n = 14). PCSK9 inhibitors were required for 3.7% (n = 6) of those who were on statin and ezetimibe therapies. However, such treatment was not initiated in any patient. Other lipid-lowering agents, such as fibric acid derivatives (fenofibrate 145 mg and 300 mg and gemfibrozil 600 mg), were initiated on 14.6% (n = 24) of those who were on statin therapy. The value of the clinical pharmacist’s interventions on applying the 2018 ACC/AHA guideline recommendations The impact of the clinical pharmacist interventions on applying the 2018 ACC/AHA guideline recommendations is shown in . In patients with LDL-C<70 mg/dl, 18 recommendations were made, ranging from adding moderate- or high-intensity statins for those who were not initiated on statins (need additional therapy–class I and IIa recommendations as per 2018 ACC/AHA guideline definition of recommendation class), changing to moderate- or high-intensity statin agents for those who were on lower-intensity statin agents and stopping other lipid-lowering agents that may not help in achieving LDL-C goals (dose adjustment/stop unnecessary medications—class I and IIa recommendations). However, the physicians’ acceptance of the aforementioned recommendations was only 22.2%. In patients with LDL-C between 70–189 mg/dl, 262 recommendations were carried out ranged from adding ezetimibe and stopping other ineffective LDL-C lowering agents for those who were in maximum tolerated dose of statin (need additional therapy/stop the unnecessary medication–class I and IIa recommendations), adding moderate or high intensity statin for those with who were not initiated on statin (need additional therapy–class I and IIa recommendations) and changing to high intensity statin dose or drug for those who were on lower intensity statin agents and stopping other ineffective LDL-C lowering agents (dose adjustment/change drug/stop the unnecessary medications—class I and IIa recommendations). The physicians’ acceptance of these recommendations was 79.4%. In patients with LDL-C ≥190 mg/dl, 30 recommendations were submitted, ranging from adding ezetimibe, PCSK9 inhibitor and stopping gemfibrozil for those with very high LDL-C results and requiring a more than 25.0% reduction in LDL-C levels despite the use of high-intensity statins (requiring additional therapy–class I recommendation), adding high-intensity statins together with ezetimibe for those who were not on statins (requiring additional therapies–class I recommendation), adding PCSK9 inhibitors for those with high LDL-C levels, although they were on high-intensity statins together with ezetimibes (requiring additional therapy–class I recommendation) and changing to high-intensity statins and adding ezetimibe for those on moderate-intensity statins even though the LDL-C level was more than or equal to 190 mg/dl (dose adjustment/requiring additional therapy–class I recommendation). Interestingly, the physicians’ acceptance of these recommendations was 93.2%. summarizes the number of recommendations and type of interventions performed by the clinical pharmacist to achieve the desired outcomes. Adherence with the 2018 ACC/AHA guideline after clinical pharmacist’s interventions Adherence with the 2018 ACC/AHA guideline for the management of cholesterol in adults after clinical pharmacist interventions is shown in . Accordingly, the number of patients who were initiated on statin therapy increased significantly up to 92.6% (n = 252) after the clinical pharmacist interventions were implemented ( X 2 (df = 1, n = 272) = 79.1, p = 0.0001). Consequently, the number of patients who were on low- or moderate-intensity statins decreased to 2.4% (n = 6) and 17.9% (n = 45), respectively. However, the number of patients who were on high-intensity statins potentially increased to 79.8% (n = 201). Based on that, adherence with the recommendations regarding the level of statin intensity used was significantly improved to 94.4% (n = 238) after the clinical pharmacist interventions ( X 2 (df = 1, n = 252) = 72.5, p = 0.0001). The use of ezetimibe as an add-on nonstatin therapy was encouraged and effectively added to the treatment plan to achieve LDL-C goals. The number of patients who were initiated ezetimibe increased significantly to 91.7% (n = 77) after the clinical pharmacist interventions ( X 2 (df = 1, n = 84) = 95, p < 0.0001). Interestingly, for those who were on statin and ezetimibe therapies and required PCSK9 inhibitors to achieve LDL-C goals, adherence with the recommendations was effectively improved to 66.7% (n = 4); ( X 2 (df = 1, n = 6) = 6, p = 0.014). The use of other lipid-lowering agents, such as fibrates, was markedly reduced to 3.2% (n = 8) for those who were on statin therapy after the clinical pharmacist interventions ( X 2 (df = 1, n = 208) = 19.2, p < 0.0001). shows comparison of adherence with the 2018 ACC/AHA guideline recommendations for the management of cholesterol before and after clinical pharmacist interventions regarding the initiation of statins, the proper use of moderate- or high-intensity statins, evidence-based addition of ezetimibe and PCSK9 inhibitors and minimization of other lipid-lowering agent abuse.
The demographic and clinical characteristics of the study sample are shown in . The mean age of the studied patients was 52.6 ±10.5, and 71.3% (n = 194) of them were males. Majority of the patients (95.6%, n = 260) had previous illness or chronic disease. Out of these patients, 178 patients (65.4%) had hypertension, 150 patients (55.1%) had dyslipidemia, and 196 patients (72.1%) had diabetes mellitus. Nevertheless, 20 patients (7.4%) had LDL-C levels less than 70 mg/dl, 236 patients (86.8%) had LDL-C levels between 70–189 mg/dl, and 16 patients (5.9%) had LDL-C levels equal to or more than 190. Additionally, 52 patients (19.1%) had a history of clinical ASCVD. Of these, 38 patients (14.0%) were at very high risk of recurrent CVD. Out of all participants without a history of clinical ASCVD, 50 patients (18.4%) had an estimated 10-year CVD risk less than 7.5%, 110 patients (40.5%) had an estimated 10-year CVD risk greater than or equal to 7.5% but less than 20.0%, and 60 patients (22.1%) had an estimated 10-year CVD risk equal to or greater than 20.0%.
Adherence with the 2018 ACC/AHA guideline recommendations for the management of cholesterol in adults is shown in . Based on the inclusion criteria, all the patients who were enrolled (100.0%, n = 272) were identified as statin benefit groups according to the 2018 ACC/AHA guideline recommendations. Of these, only 60.3% (n = 164) were initiated on statin therapy. Out of those who were on statin therapy, 9.8% (n = 16) were on low intensity statin (e.g., simvastatin 10 mg and pitavastatin 1 mg), 52.4% (n = 86) were on moderate intensity statin (e.g., simvastatin 40 mg, rosuvastatin 10 mg, atorvastatin 20 mg, pitavastatin 2 mg and 4 mg) and 37.8% (n = 62) were on high intensity statin (e.g., atorvastatin 40 mg and 80 mg and rosuvastatin 20 mg and 40 mg). Adherence to the recommend level of statins intensity was identified in only 47.6% of patients (n = 78). The addition of non-statin therapies to achieve LDL-C goals was also assessed, and ezetimibe was required for 51.2% (n = 84) of those who were on statin therapy. While it was initiated for only 8.5% (n = 14). PCSK9 inhibitors were required for 3.7% (n = 6) of those who were on statin and ezetimibe therapies. However, such treatment was not initiated in any patient. Other lipid-lowering agents, such as fibric acid derivatives (fenofibrate 145 mg and 300 mg and gemfibrozil 600 mg), were initiated on 14.6% (n = 24) of those who were on statin therapy.
The impact of the clinical pharmacist interventions on applying the 2018 ACC/AHA guideline recommendations is shown in . In patients with LDL-C<70 mg/dl, 18 recommendations were made, ranging from adding moderate- or high-intensity statins for those who were not initiated on statins (need additional therapy–class I and IIa recommendations as per 2018 ACC/AHA guideline definition of recommendation class), changing to moderate- or high-intensity statin agents for those who were on lower-intensity statin agents and stopping other lipid-lowering agents that may not help in achieving LDL-C goals (dose adjustment/stop unnecessary medications—class I and IIa recommendations). However, the physicians’ acceptance of the aforementioned recommendations was only 22.2%. In patients with LDL-C between 70–189 mg/dl, 262 recommendations were carried out ranged from adding ezetimibe and stopping other ineffective LDL-C lowering agents for those who were in maximum tolerated dose of statin (need additional therapy/stop the unnecessary medication–class I and IIa recommendations), adding moderate or high intensity statin for those with who were not initiated on statin (need additional therapy–class I and IIa recommendations) and changing to high intensity statin dose or drug for those who were on lower intensity statin agents and stopping other ineffective LDL-C lowering agents (dose adjustment/change drug/stop the unnecessary medications—class I and IIa recommendations). The physicians’ acceptance of these recommendations was 79.4%. In patients with LDL-C ≥190 mg/dl, 30 recommendations were submitted, ranging from adding ezetimibe, PCSK9 inhibitor and stopping gemfibrozil for those with very high LDL-C results and requiring a more than 25.0% reduction in LDL-C levels despite the use of high-intensity statins (requiring additional therapy–class I recommendation), adding high-intensity statins together with ezetimibe for those who were not on statins (requiring additional therapies–class I recommendation), adding PCSK9 inhibitors for those with high LDL-C levels, although they were on high-intensity statins together with ezetimibes (requiring additional therapy–class I recommendation) and changing to high-intensity statins and adding ezetimibe for those on moderate-intensity statins even though the LDL-C level was more than or equal to 190 mg/dl (dose adjustment/requiring additional therapy–class I recommendation). Interestingly, the physicians’ acceptance of these recommendations was 93.2%. summarizes the number of recommendations and type of interventions performed by the clinical pharmacist to achieve the desired outcomes.
Adherence with the 2018 ACC/AHA guideline for the management of cholesterol in adults after clinical pharmacist interventions is shown in . Accordingly, the number of patients who were initiated on statin therapy increased significantly up to 92.6% (n = 252) after the clinical pharmacist interventions were implemented ( X 2 (df = 1, n = 272) = 79.1, p = 0.0001). Consequently, the number of patients who were on low- or moderate-intensity statins decreased to 2.4% (n = 6) and 17.9% (n = 45), respectively. However, the number of patients who were on high-intensity statins potentially increased to 79.8% (n = 201). Based on that, adherence with the recommendations regarding the level of statin intensity used was significantly improved to 94.4% (n = 238) after the clinical pharmacist interventions ( X 2 (df = 1, n = 252) = 72.5, p = 0.0001). The use of ezetimibe as an add-on nonstatin therapy was encouraged and effectively added to the treatment plan to achieve LDL-C goals. The number of patients who were initiated ezetimibe increased significantly to 91.7% (n = 77) after the clinical pharmacist interventions ( X 2 (df = 1, n = 84) = 95, p < 0.0001). Interestingly, for those who were on statin and ezetimibe therapies and required PCSK9 inhibitors to achieve LDL-C goals, adherence with the recommendations was effectively improved to 66.7% (n = 4); ( X 2 (df = 1, n = 6) = 6, p = 0.014). The use of other lipid-lowering agents, such as fibrates, was markedly reduced to 3.2% (n = 8) for those who were on statin therapy after the clinical pharmacist interventions ( X 2 (df = 1, n = 208) = 19.2, p < 0.0001). shows comparison of adherence with the 2018 ACC/AHA guideline recommendations for the management of cholesterol before and after clinical pharmacist interventions regarding the initiation of statins, the proper use of moderate- or high-intensity statins, evidence-based addition of ezetimibe and PCSK9 inhibitors and minimization of other lipid-lowering agent abuse.
Based on this study, adherence with the 2018 ACC/AHA guideline recommendation for the management of cholesterol in adult patients before clinical pharmacist interventions was 60.3% for the initiation of statins therapy and 47.6% for adherence to proper intensity statin therapy. Accordingly, the initiation of statins, particularly high-intensity statins, is prescribed to far fewer patients than recommended. Consequently, the use of non-statin therapies such as ezetimibe and PCSK9 inhibitors was nearly diminished, taking into consideration that several studies highlighted the importance of pharmacist intervention on cholesterol risk management and revealed the treatment gap between research evidence and clinical practice . According to our findings, the clinical pharmacist plays a crucial role in the management of cholesterol levels by recommending new therapies, adjusting or increasing drug doses and stopping or changing medications. Furthermore, systematic reviews and meta-analyses of randomized trials conducted by Machado et al. and Santschi et al. emphasized the importance of pharmaceutical care interventions in the management of CVDs . Pharmacist interventions achieved greater reductions in systolic and diastolic blood pressure (BP), total cholesterol (TC), and LDL-C and in the risk of smoking compared with the usual care group . Nevertheless, various clinical trials have illustrated great benefits of statin use, such as pleiotropic effects, which could be beneficial for the treatment and management of several comorbidities . In this study, adherence with the 2018 ACC/AHA guideline to achieve the required LDL-C goals was significantly improved after clinical pharmacist interventions and the implementation of the appropriate recommendations. Consistently, Bozovich et al. 2000 and Tahaineh et al. 2011 showed significant improvement in achieving LDL-C goals when clinical pharmacists managed lipid clinics or through clinical pharmacy services under the supervision of cardiologists . The same was achieved by Tsuyuki RT et al. (2016) . In the current study, physicians’ acceptance of the clinical pharmacist’s recommendation according to the guidelines was variable based on patients’ LDL-C levels. For instance, physicians” acceptance of clinical pharmacist interventions was high among patients with LDL-C ≥70 mg/dl. Subsequently, this resulted in greater improvement of LDL-C levels and improvement in health outcomes. Likewise, recent studies reported that primary healthcare physicians significantly relied on clinical pharmacists in assessing and improving patients’ adherence to their medications as well as in educating and counseling the patients to avoid clinical malpractice and achieve better health outcomes . Several studies presented the major explanations for statin refractoriness reported by healthcare practitioners, and patients were concerned about adverse events . Rosenson, R. S 2016 stated that evaluation of potential adverse events requires validated tools to differentiate between statin-associated adverse events versus nonspecific complaints. Additionally, treatment options for statin-intolerant patients include the use of different statins, often at a lower dose or frequency. To lower LDL cholesterol, lower doses of statins may be combined with ezetimibe or bile acid sequestrants . Newer treatment options for patients with statin-associated muscle symptoms may include proprotein convertase subtilisin kexin 9 (PCSK9) inhibitors . There are many reasons that contribute to non-adherence with the guidelines or rejecting the pharmacists’ recommendations such as the availability of specific drugs and patient’s reluctant for treatment initiation or dose escalation. The results of the current study indicated that many physicians are reluctant to prescribe high intensity statins due the worry about side effects and myopathy. In some cases, physicians stop or change the dose of statins when patients’ report intolerance of such medications or due to the high cost. Another reason for rejecting the pharmacists’ recommendations in this study was low bassline LDL-C level for some patients despite the patient being categorized as a statin benefit group. In clinical trials, statin-associated adverse events showed no differences between participants assigned to statins or placebo . However, it is important to know that these trials select patients with better tolerability and lower risk for myopathy based on their ages, absence of musculoskeletal complaints, normal renal function and less concomitant medications that may alter the pharmacokinetic pathways . One of the solutions to overcome the problem is to switch to the fully human monoclonal antibodies proprotein convertase subtilisin/kexin type 9 (PCSK9) inhibitors (alirocumab and evolocumab) that caused fewer muscle symptoms based on clinical trials and were no more often than when ezetimibe was used . However, the cost of such treatment is still one of the main barriers. Limitations This study is an interventional before after design. This design is usually used in circumstances where it is not possible to use a control group for ethical or practical issues. Although this design is suitable for the current study, the lack of control group makes this design prone to bias and many confounders. Therefore, the outcome can instead be related to any changes that occurred around the same time as the intervention. Although the clinical pharmacists and physicians who treated the patients in the studied clinics remained the same during the study period, other unknown confounders could have occurred. Another limitation is that it is not an easy task to initiate statin therapy for those are low or intermediate risk for ASCVD. It involves looking for wide range of risk-enhancing factors which could favor the initiation of statin therapy for the low or intermediate risk group, for example the presence of premature CVD in the family. These risk-enhancing factors can be missed during patient assessment or during data extraction. on the other hand, the patient with low or intermediate ASCVD risk is allowed to option for not taking statins for primary ASCVD prevention. Therefore, it is difficult to precisely judge physicians for adherence to guideline for initiating lipid-lowering agents. In addition, the estimate ASCVD risk is more imprecise in some patients when cholesterol levels were used after treatment.
This study is an interventional before after design. This design is usually used in circumstances where it is not possible to use a control group for ethical or practical issues. Although this design is suitable for the current study, the lack of control group makes this design prone to bias and many confounders. Therefore, the outcome can instead be related to any changes that occurred around the same time as the intervention. Although the clinical pharmacists and physicians who treated the patients in the studied clinics remained the same during the study period, other unknown confounders could have occurred. Another limitation is that it is not an easy task to initiate statin therapy for those are low or intermediate risk for ASCVD. It involves looking for wide range of risk-enhancing factors which could favor the initiation of statin therapy for the low or intermediate risk group, for example the presence of premature CVD in the family. These risk-enhancing factors can be missed during patient assessment or during data extraction. on the other hand, the patient with low or intermediate ASCVD risk is allowed to option for not taking statins for primary ASCVD prevention. Therefore, it is difficult to precisely judge physicians for adherence to guideline for initiating lipid-lowering agents. In addition, the estimate ASCVD risk is more imprecise in some patients when cholesterol levels were used after treatment.
The clinical pharmacist has a key role in improving the management of blood cholesterol by recommending therapies, adjusting doses and stopping or changing medications. Furthermore, adherence with the latest updated guideline recommendations to achieve the desired treatment goals was notably enhanced after the clinical pharmacist interventions and the implementation of the appropriate recommendations. This study illustrates how collaboration between physicians and clinical pharmacists can be crucial strategy to improve patients’ treatment and hence, achieve better health outcomes among patients suffering from dyslipidemia.
S1 Data (XLSX) Click here for additional data file.
|
Evaluation of a tool to improve the quality of preceptor written feedback for family medicine residents: training and use of a CanMEDS-MF competency-based criterion guide | 7a01c8dd-448d-4dff-8788-6be2f022e76a | 10042787 | Family Medicine[mh] | Written feedback is an essential element in teaching residents, but preceptors are not always equipped to provide relevant feedback. Family medicine residency programs are aligned with a CanMEDS-Family Medicine (CanMEDS-FM) 2017 seven-competency framework to address patient needs, developed by the College of Family Physicians of Canada. Its integration is based on a formative evaluation that promotes residents’ professional development , but also depends on quality feedback, which remains a challenge for preceptors. , In the Montfort unit of the University of Ottawa, we have observed a lack of quality feedback, which is often absent or non-specific. Yet relevant feedback allows for the development of competent and independent clinicians , – by: clarifying expected performance; providing specific information on progress; encouraging the learner with constructive comments; identifying and explaining difficulties; offering solutions to achieve expected performance. – However, the literature remains limited on the tools available to support preceptors in this function. To address these gaps, we hypothesized that preceptors need tools developed according to the CanMEDS-FM competencies, including preconstructed sentences to facilitate writing. We developed a criterion-referenced guide, then trained preceptors to focus their comments according to this guide using an evaluation sheet (field note). This study evaluates the effectiveness of multi-episodic training and the use of the criterion-referenced guide for written feedback among family medicine preceptors. The goal is to expand the use of this model to all the department’s units if it is valid.
This article was written according to the standards and guidelines of Standards for Quality Improvement Reporting Excellence (SQUIRE 2.0) and Standards for Quality Improvement Reporting Excellence in Education (SQUIRE-EDU). REB submission and review The project was submitted to the Montfort Hospital Research Ethics Board (August 13, 2020). Ethics approval was not deemed necessary (File No. 20-21-06-016). During the team meetings, interested participants reviewed the project details and their involvement based on written informed consent. Study design The target population, made up of 32 preceptors in the Montfort unit of the University of Ottawa Family Medicine Residency Program, was invited by email and at team meetings. Data collection was carried out for three months from June 1 to September 30, 2020 (pre-test), then three months after the intervention (post-test) from January 1 to March 31, 2021. Volunteer preceptors, from novice to highly experienced, participated in a series of four monthly one-hour training sessions (October 2020 to Janury 2021), for which they received compensation of $150. The trainings were organized in discussion groups and followed an initial presentation of the tool by a moderator, also an experienced preceptor in the Montfort unit, who had participated in designing the educational tool. The training sessions focused on understanding the CanMEDS-FM framework and relevant feedback according to the educational tool, a criterion-referenced guide previously developed by the research team (Appendix A, ). Preceptors were put in the situation of writing quality feedback based on their own evaluations noted in the previous month. Then, their feedback was recorded on new interactions over the next few months using a field note (Appendix A, ) for the roles of Communicator, Collaborator, Expert, Professional, Health Advocate, Scholar and Leader based on in-office observations made during patient/resident encounters. The content of these field notes was extracted for analysis. This field note has been in use throughout the department since 2013. We note that previous trainings took place in 2013-2014 when field notes were introduced and primarily described the CanMEDS-FM roles and not how to write feedback. Sample size The required sample size was not calculated a priori. The convenience sample included 23 preceptors who had participated in training. However, it was not possible to get more than 16 of the 23 participants to perform the pre-test. We explain this discrepancy by the fact that many are not diligent in completing the field notes in a timely manner and some new preceptors had not yet been exposed to this function. We also believe that our trainings motivated people to complete the field notes, but it is impossible to subtract the data of the 7 additional post-test participants because the data were anonymised as per the consent to participate. The analysis was continued with this limitation in mind. Outcome measures Analysis of the field notes content, carried out by a member of the research team and reviewed for consensus by a second, was based on the primary outcome measures describing their completion and the quality of feedback. The field notes were classified according to complete and adequate use including a performance level evaluation and additional written feedback on the observable roles; incomplete use with no feedback for the observable roles; inadequate use not corresponding to the educational tool. The quality of feedback was classified as specific (personalized to the resident’s performance with examples or constructive criticism), non-specific (not personalized and using, for example, a CanMEDS-FM framework citation with no preceptor feedback) or absent. The feedback rate was then recorded by CanMEDS-FM role as a secondary outcome measure. Data analysis An independent repeated measures analysis was carried out between the pre-test and post-test using the two proportion Z-test with Excel software (Excel 2018, Microsoft Corporation) following an accounting of rates by percentage.
The project was submitted to the Montfort Hospital Research Ethics Board (August 13, 2020). Ethics approval was not deemed necessary (File No. 20-21-06-016). During the team meetings, interested participants reviewed the project details and their involvement based on written informed consent.
The target population, made up of 32 preceptors in the Montfort unit of the University of Ottawa Family Medicine Residency Program, was invited by email and at team meetings. Data collection was carried out for three months from June 1 to September 30, 2020 (pre-test), then three months after the intervention (post-test) from January 1 to March 31, 2021. Volunteer preceptors, from novice to highly experienced, participated in a series of four monthly one-hour training sessions (October 2020 to Janury 2021), for which they received compensation of $150. The trainings were organized in discussion groups and followed an initial presentation of the tool by a moderator, also an experienced preceptor in the Montfort unit, who had participated in designing the educational tool. The training sessions focused on understanding the CanMEDS-FM framework and relevant feedback according to the educational tool, a criterion-referenced guide previously developed by the research team (Appendix A, ). Preceptors were put in the situation of writing quality feedback based on their own evaluations noted in the previous month. Then, their feedback was recorded on new interactions over the next few months using a field note (Appendix A, ) for the roles of Communicator, Collaborator, Expert, Professional, Health Advocate, Scholar and Leader based on in-office observations made during patient/resident encounters. The content of these field notes was extracted for analysis. This field note has been in use throughout the department since 2013. We note that previous trainings took place in 2013-2014 when field notes were introduced and primarily described the CanMEDS-FM roles and not how to write feedback.
The required sample size was not calculated a priori. The convenience sample included 23 preceptors who had participated in training. However, it was not possible to get more than 16 of the 23 participants to perform the pre-test. We explain this discrepancy by the fact that many are not diligent in completing the field notes in a timely manner and some new preceptors had not yet been exposed to this function. We also believe that our trainings motivated people to complete the field notes, but it is impossible to subtract the data of the 7 additional post-test participants because the data were anonymised as per the consent to participate. The analysis was continued with this limitation in mind.
Analysis of the field notes content, carried out by a member of the research team and reviewed for consensus by a second, was based on the primary outcome measures describing their completion and the quality of feedback. The field notes were classified according to complete and adequate use including a performance level evaluation and additional written feedback on the observable roles; incomplete use with no feedback for the observable roles; inadequate use not corresponding to the educational tool. The quality of feedback was classified as specific (personalized to the resident’s performance with examples or constructive criticism), non-specific (not personalized and using, for example, a CanMEDS-FM framework citation with no preceptor feedback) or absent. The feedback rate was then recorded by CanMEDS-FM role as a secondary outcome measure.
An independent repeated measures analysis was carried out between the pre-test and post-test using the two proportion Z-test with Excel software (Excel 2018, Microsoft Corporation) following an accounting of rates by percentage.
We collected 70 pre-test field notes from 16 preceptors and 138 post-test field notes from 23 preceptors. The primary outcome measures and secondary outcome measure are analyzed using the p value as a significant difference ( p < 0.05).
This study supports the hypothesis that preceptors need tools to improve feedback given in written form. Evidence shows that a clear understanding and elaboration of competencies facilitates specific and concise written feedback , , which we observe by the increase in complete field notes (40% vs 92%, z = 3.51, p = 0.0005) and specific feedback (59% vs 92%, z = 2.47, p = 0.0137). Preceptors expressed an active commitment for a tool allowing them to better focus their comments in the absence of training since 2015. Despite lower participation ( n = 23) than a minimum sample ( n = 30) for a confidence interval, we are certain that the results would be similar with higher participation considering the level of engagement observed. We also believe, from their enthusiasm, that compensation is not essential and that the tool would find takers without specific remuneration. In contrast, we note a variability in feedback rates by CanMEDS-FM role during the study with greater popularity for communication and expertise. This variability is observed in other comparable studies where communication often sees the most improvement. , It is argued that the variability is partly due to the recent operationalization of certain concepts in medical education including collaboration, leadership and professionalism. Some roles, including professionalism, are difficult to implement, as educational tools are influenced by a limited and dated curriculum for the teaching of professional commitment. In the context of this study, we add that it is difficult to evaluate collaboration, leadership, and scholarship in a practice setting. Communication and expertise are more applicable to numerous contexts during the residency program. However, despite the fact that equal use of the competencies is questionable , all CanMEDS-FM roles are essential to patient needs. , The analysis of the field notes makes it possible to identify educational gaps and the ongoing need for training , as well as the need to expand the tool’s applicability to different clinical settings to reach a greater number of these roles. Limitations This study has certain limitations including its implementation during the pandemic. With preceptor burnout and health restrictions, the number of pre-test participants was lower and less than 30 for a confidence interval, impacting the validity of the results over a short period. We can only hypothesize that educational tools lead to an improvement. However, it is also possible that gradual preceptor burnout minimized the impact of the intervention on certain outcomes by discouraging their participation. And, the results do not include the residents’ perspective, creating a bias in perceived improvement. Further study of educational tools has the potential for better understanding in the use of the CanMEDS-FM framework. There is an opportunity to validate the tools with preceptors and residents in order to implement their use in medical education in a longitudinal study in all units of the department.
This study has certain limitations including its implementation during the pandemic. With preceptor burnout and health restrictions, the number of pre-test participants was lower and less than 30 for a confidence interval, impacting the validity of the results over a short period. We can only hypothesize that educational tools lead to an improvement. However, it is also possible that gradual preceptor burnout minimized the impact of the intervention on certain outcomes by discouraging their participation. And, the results do not include the residents’ perspective, creating a bias in perceived improvement. Further study of educational tools has the potential for better understanding in the use of the CanMEDS-FM framework. There is an opportunity to validate the tools with preceptors and residents in order to implement their use in medical education in a longitudinal study in all units of the department.
The development of multi-episodic training and a criterion-referenced guide, created according to the CanMEDS-FM framework, suggests an improvement in complete and specific written feedback by preceptors. Further studies are needed to validate these tools in family medicine and medical education.
|
Soil contamination in nearby natural areas mirrors that in urban greenspaces worldwide | ffcc3821-9f0c-4e9a-816c-d53c57aff37d | 10042830 | Microbiology[mh] | Soil contamination challenges many of the United Nations Sustainable Development Goals such as good health and wellbeing, sustainable ecosystems and cities, and climate change regulation , . Environmental stress associated with soil contamination, either from natural or anthropogenic origins, can directly affect biodiversity and ecosystem functions , and further compromise the resistance and resilience of ecosystems to climate change and natural disasters . Moreover, soil contamination in urban areas can negatively influence the health of city residents through cross-media migration-induced risk (e.g., toxic metal(loid)s affecting drinking water quality and vapor intrusion of organic contaminants) , . In addition, urban greenspaces are important recreational places where people have direct contact with soil. Currently, soil contamination is associated with vehicle emissions, industrial processes, weed, and plant disease treatment, as well as poor waste management , . Thus, urban greenspaces are expected to be more influenced by contaminants than natural ecosystems, which are geographically removed from anthropogenic activities. However, studies have shown that contaminants of concern such as metal(loid)s, pesticides, microplastics, and antibiotic resistance genes (ARGs) can be dispersed through aerial transport, uncontrolled waste disposal/littering, and runoff – , and thus may have impacts on adjacent natural ecosystems. Moreover, some potential soil contaminants have natural origins (e.g., high levels of soil metal(loid)s and ARGs) and can also affect surrounding managed ecosystems. Previous studies have demonstrated significant dispersion of certain contaminants at local and regional scales; , – however, these efforts mainly focused on single contaminants across a relatively narrow range of climate and environmental gradients, and many times lack a direct comparison of the most widespread contaminants simultaneously between urban and natural ecosystems. Thus, a global and multidimensional assessment of soil contamination in urban greenspaces is urgently needed to understand the distribution and range of human-driven soil contamination. Furthermore, the importance of human and natural factors in explaining soil contamination across contrasting ecosystems remains virtually unknown. A global assessment of soil contamination is a huge challenge, as different soil types and climates may have contributed considerable uncertainty to contaminant effects. Yet, any attempt to estimate patterns and consequences of soil contamination requires multi-ecosystem and multidimensional approaches simultaneously across a broad range of environmental and socio-economic conditions . Moreover, it is imperative to identify how soil functional activities respond to environmental stress typically associated with natural or anthropogenic contaminants. Soil microbes perform critical functions and ecosystem services , and are known to respond rapidly to contamination, with changes in the proportion of fundamental microbial traits , . Because of their environmental sensitivity and widespread distribution, we posit that soil microbes can be used as global indicators of multiple dimensions of natural and anthropogenic soil contaminants (i.e., from microplastics to metal(loid)s). This is vital to better understand the impacts of soil contaminants and will help design strategies to improve ecosystem conservation and human health. We conducted a global standardized field survey including surface soils collected from 56 paired urban greenspaces and adjacent natural areas (i.e., unmanaged natural/seminatural ecosystems) across six continents (Supplementary. Fig. , Table ). Our paired design allows the direct comparison of urban and natural areas while allowing for biogeographic and macroclimatic patterns, which is a major novelty compared with previous work. We aimed to (i) compare the levels of multiple soil contaminants of concern including metal(loid)s, pesticides, microplastics, and ARGs in urban greenspaces and adjacent natural areas (Supplementary Table ); (ii) and explore environmental factors associated with soil contaminants in paired areas; and (iii) examine the potential influence of soil contaminants on functional microbial traits associated with soil health such as stress resistance, nutrient cycling, and pathogenesis. Finally, to better understand the importance and implications of soil contamination for the conservation of natural ecosystems, we compared the level of soil contaminants in urban greenspaces with those found in three remote ecosystems from maritime Antarctica (Supplementary. Fig. ). Our global survey focused on four main groups of soil contaminants of wide concern including eight heavy metals and metalloids (metal(loid)s, hereafter): arsenic (As), cadmium (Cd), chromium (Cr), copper (Cu), lead (Pb), mercury (Hg), nickel (Ni) and zinc (Zn), 46 pesticide residues, the shape and polymer type of microplastics, and 285 ARGs (Supplementary Table ). In the case of metal(loid)s and ARGs, we acknowledge that they are also naturally enriched in certain soils, irrespective of contamination. Therefore, to interpret the degree to which values of these contaminants can be attributed to anthropogenic activities, we identified the associations between natural (climate, plants and soil) and human factors (population size and density, human development index (HDI), gross domestic product (GDP), and management practices of urban greenspaces) and soil contaminants. Further, we explored the links between the co-occurrence of multiple soil contaminants and microbial functional attributes (e.g., genes associated with biological stress), which remain unexplored and are critically important to predict potential effects of different widespread contaminants on soil functions.
Our survey demonstrates that, irrespective of their origins (geochemical background vs. anthropogenic), soil contaminants (i.e., metal(loid)s, pesticides, microplastics, and ARGs) are widely distributed in terrestrial ecosystems. In particular, we found similar levels of multiple dimensions of soil contamination in urban greenspaces and adjacent natural areas across continents (Supplementary Fig. and Supplementary Table ). The same results were obtained when using a response ratio of soil contaminants (urban vs. natural; Fig. ). Although the level of individual contaminants varied greatly across locations, we detected significant correlations among each type of soil contaminants studied (Supplementary Fig. ). Our data further provide a snapshot of multiple soil contaminants in natural areas, comparing with the adjacent urban greenspaces worldwide (Supplementary Table and Supplementary Fig. ). We also found significant levels of these contaminants in the soil of remote Antarctica, confirming earlier local reports of atmospheric transport of specific contaminants including pesticides and microplastics , . Moreover, we provide evidence that human-associated factors are essential to explain the distribution of multiple soil contaminants across contrasting ecosystems (Fig. ), supporting an important link between anthropogenic activity and soil contamination that goes beyond the natural origin of particular contaminants, such as the case of metal(loid)s. Finally, we found that levels of soil contaminants were significantly correlated with the relative abundance of important functional genes associated with stress resistance (i.e., the resistance to stresses of metal(loid), drugs, and pathogens), nutrient cycling, pathogenesis, and microbial metabolism (Fig. ). Importantly, we found that an increasing number of soil contaminants can have a greater contribution in explaining the proportion of microbial functional traits (see Supplementary Fig. ). Our work is relevant because it provides evidence of a quantitative comparison of soil contaminants in urban greenspaces and adjacent natural areas across six continents. The results revealed that adequate conservation and functioning of terrestrial ecosystems requires an assessment of risk associated with multiple dimensions of soil contamination. Multiple dimensions of soil contaminants in paired urban and natural areas We found similar levels of metal(loid)s such as Pb, Ni, Cd, and As in soils of urban greenspaces and adjacent natural areas, but urban greenspaces had higher contents of soil Hg, Zn, Cu, and Cr than those in natural ecosystems (Supplementary Fig. ). Differences in the accumulation of metal(loid)s in urban greenspaces compared with natural areas could be related to geographic variations in the extent of metal(loid) emissions from human activities, though the exact mechanisms remain unknown . The broad spectrum of anthropogenic impacts considered in this study (e.g., Pb and Hg) captures, at least partially, the potential anthropogenic sources of these elements. However, the natural sources and geochemical background of these trace metals are also critical in explaining levels of these metal(loid)s . Metal(loid)s in the topsoils of urban greenspaces may affect soil organisms and ecosystem processes when they exceed a specific threshold concentration. For instance, according to, the guidelines of the Finnish legislation for soil contamination , standards that are widely used to assess soil contamination, 42% of urban greenspace sites and 36% of sites in natural areas exceeded the lower limit for As contamination in soil (Supplementary Table ). As expected, however, average levels of these metal(loid)s in the remote Antarctica ecosystem were below the legislation limit (Supplementary Figs. and ). These findings advance our knowledge of the typical levels of toxic metal(loid)s found in urban greenspaces and adjacent natural areas worldwide. Accumulated contaminants in surface soils of greenspaces do not necessarily increase environmental risk because the bioavailability of soil contaminants depends on soil properties such as pH, texture, and organic carbon content . Similarly, pesticide residues were widely detected in many urban greenspaces and natural ecosystems at a global scale. We detected the presence of pesticide residues, including fungicides, herbicides, and insecticides, in 63% of the surveyed natural ecosystems (Supplementary Fig. ; Supplementary Table ). Our analyses likely underestimate these levels as we did not measure several widely used pesticides, such as glyphosate, glufosinate, paraquat, and 2,4-Dichlorophenoxyacetic acid. We did not detect a difference in the occurrence of pesticides between urban and natural greenspaces (Supplementary Fig. ). This is also evident when analyzing response ratios (urban vs. natural; Fig. ) or accounting for our paired design (Supplementary Fig. ). The application of persistent pesticides in urban areas may influence adjacent natural areas, for example through surface runoff and atmospheric transport and deposition . In addition, agricultural pesticide application may contribute to the reported levels of pesticide residues in adjacent urban greenspaces and natural areas by deposition. Microplastics, typical anthropogenic-driven contaminants, were also ubiquitous in soils of both urban greenspaces and natural ecosystems worldwide. Strikingly, we found similar contents of soil microplastics in the paired ecosystems worldwide (Fig. ; Supplementary Fig. ). The average level of microplastics was 916.2 items kg −1 soil (range from 166.7 to 3482.3 items kg −1 ), comparable to previously reported values , . The detected microplastics consisted predominately of polypropylene and polyester, with fibers as the dominant shape (Supplementary Fig. ), consistent with reports of microplastic levels in the ocean and terrestrial ecosystems . We also found similar proportions of the shape and polymer type of microplastics in natural areas and urban greenspaces (Fig. ; Supplementary Fig. ), which further supports the idea of a spillover of anthropogenic contaminants across ecosystems. These microplastics, which typically originate from cities, can also reach distant areas by atmospheric transport, with fibers being the main shape of airborne plastic particles in the atmosphere of cities such as Paris , London and Dongguan (China) . These fibers generally consist of polyester and polypropylene, which can originate from synthetic fabric, rope, and nets. In addition, semi-natural ecosystems can also accumulate microplastics due to human activities, leading to waste disposal /littering, runoff, and transport through the atmosphere . Unexpectedly, our results showed that soils in remote Antarctica had similar signatures of microplastics (in both concentration and type) in surface soils to our surveyed urban greenspaces worldwide (Supplementary Fig. ). This could be related to microplastic dispersion from Antarctic research stations and other continents by sea and air, though this would depend on wind strength and direction . For example, recent studies showed that more than 60% of microplastics detected in Antarctic samples were <50 µm, and these tiny-sized plastics can be transported over long distances before deposition , . Other activities, such as, tourism may also contribute to the accumulation of microplastics in the soils of Antarctica sites . Future studies should provide more detailed measurements of the characteristics of microplastics in order to track their distribution and fate in the environment. On the whole, our findings provide evidence that the level and characteristics of microplastics in natural areas match those in urban city parks and gardens, with multiple conservation and ethical implications for microplastic contamination of terrestrial ecosystems worldwide. We then characterized the richness and abundance of 285 ARGs in the soils from the 112 ecosystems studied (Fig. ; Supplementary Fig. ). Despite their natural origin (soil-borne bacteria), soil ARGs are emerging as potential biological contaminants threatening human health , , and the presence of some ARGs in soil has been linked to the widespread use of antibiotics in medical and agricultural industries (e.g., veterinary medicine) . The intense presence of pet excreta, atmospheric deposition, and population movement may be the main transfer routes of ARGs to the soil of urban greenspaces , . Thus we consider them an indirect measure of antibiotic influences in highly disturbed environments such as urban greenspaces. Our results show the ubiquity of ARGs in both urban greenspaces and natural areas, though the proportion of specific ARG types varied across locations (Supplementary Figs. and ). The similarity in ARGs between natural and urban soils might be related to the physical mobility of bacterial cells and plasmids carrying ARGs from urban to natural environments . We show that soils from urban greenspaces supported a greater diversity (i.e., richness) of ARGs than natural ecosystems, suggesting that urban areas may receive more diverse sources of ARGs from anthropogenic sources. These analyses advance our current understanding of ARGs by providing quantitative PCR data to investigate the diversity of ARGs in urban environments , though our analyses may not include those ARGs that are below detection limits. In addition, the significantly lower richness of ARGs in remote Antarctica (Supplementary Fig. ) further suggests that the presence of ARGs could be linked to human antibiotic use. Human and natural factors associated with soil contamination across the globe To provide further insights into their origin and anthropogenic contribution to soil contamination in urban and natural ecosystems, we investigated the association among socio-economic and environmental factors and soil contaminants. Structural equation modelling (SEM) and linear mixed-effects models consistently revealed that regional population density was the most important socio-economic factor associated with microplastics (Fig. ; Supplementary Figs. and ). This finding accords with local studies showing significant links between population density and environmental contamination , . Densely populated areas are typically associated with greater microplastic inputs from traffic, industrial emissions, domestic waste and the release of synthetic fibers , , or sewage sludge and compost used as soil fertilizers . We also found negative associations between city wealth (reflected in GDP and HDI, Supplementary Fig. ) and soil microplastics and metal(loid)s (Fig. ; Supplementary Fig. ). Significant associations between response ratios of soil contaminants and population density further underline the importance of human factors in the accumulation of soil contaminants, particularly microplastics in urban ecosystems (Supplementary Table ). According to our modelling analyses, human factors such as soil fertilization played more important roles than natural factors in influencing the abundance of pesticides and ARGs worldwide (Fig. ). For instance, despite the natural origin of ARGs, fertilization was positively associated with the abundance of ARGs (Fig. ). The widespread use of sewage sludge and animal manure containing antibiotic residues may have increased the presence of ARGs in fertilized urban sites . Taken together, we posit that human activities are causing a wide spillover of multiple soil contaminants across ecosystems, transporting contaminants between urban greenspaces and natural areas by atmospheric deposition and surface runoff. Soil contamination, however, is a complex process, and may be affected by other undocumented factors, many of which act at a local scale. Our statistical analyses are also limited in that we cannot assume causal relationships among these factors. Long-term field monitoring and controlled experiments are needed to strengthen our understanding of the diffusion and accumulation of soil contaminants across contrasting ecosystems. Our work points to the global magnitude of these processes and their links to human activities. Associations among soil contaminants and microbial functional traits Finally, we used metagenomic data to establish potential associations among multiple soil contaminants and biological responses associated with key functional aspects maintaining ecosystem sustainability . We found, for example, that greater contents of soil Cr, Cu, Hg and Zn in urban greenspaces than in natural areas matched the larger proportion of genes associated with biological resistance (e.g., multidrug resistance) in urban environments (Fig. ; Supplementary Fig. , Table ). Moreover, the proportion of genes associated with resistance to streptothricin (an antibiotic) was positively correlated with elevated concentrations of Hg across soils (Supplementary Fig. ). These findings suggest that the soil functional microbiome is linked to metal(loid) concentrations. In addition, other well-known genes associated with stress resistance, nutrient cycling, pathogenesis and microbial metabolism were associated with a greater level of multiple soil contaminants when considered simultaneously (multi-contamination calculated as the averaged of standardized concentration of contaminants, see Methods) (Fig. ). For example, the soil multi-contamination index was positively correlated with the proportion of the genes associated with pathogenic Listeria prophages and multidrug resistance, but negatively correlated with the genes involved in P metabolism, Fe acquisition and transport, and DNA repair (Fig. ). These results suggest that the accumulation of contaminants potentially increased the pathogenicity of soil microbes and affected their functions. We also found negative associations among pesticides and genes associated with nutrient (e.g., N, P, and Fe) cycling and basic metabolisms (e.g., ABC transporter and 4Fe-4S ferredoxin) (Fig. ; Supplementary Figs. and ). Importantly, we found that the number of soil contaminants over certain thresholds was significantly correlated with the proportions of the selected functional genes (Supplementary Fig. ). These results indicate significant impacts of multiple contaminants on the soil functional microbiome, which is critical to supporting ecosystem services . However, gene abundance does not necessarily reflect functional activity , and further experiments are needed to improve our understanding of soil contamination impacts on functional attributes supported by the soil microbiome. Experimental studies manipulating soil contaminants and testing how they influence soil health are important to test further links between soil contaminations and functional gene activities observed here. Such knowledge will help to protect ecosystem functions from human-driven contamination across the globe efficiently. Implications Our study provides the cross-continental-scale assessment of the patterns and biological responses associated with soil contaminants in natural and urban areas. We provide evidence that soil contamination in natural areas surrounding cities mirrors that in urban greenspaces worldwide. This highlights the growing global contaminant footprint, which has reached even remote locations such as Antarctica. We also found that socio-economic factors are significantly associated with the occurrence of soil contaminants, with positive associations between population density and the levels of metal(loid)s and microplastics, highlighting the contribution of humans to soil contamination regardless of their origins. We further reveal the potential influence of soil contamination on key microbial traits such as stress resistance, nutrient cycling, and pathogenesis. These findings pave the way to the formulation of hypotheses that can be tested using controlled laboratory- and field-based studies. Together, our work demonstrates that soils in nearby natural areas are as contaminated as our urban greenspaces at a large-spatial scale. Future studies on a wider spectrum of contaminants (e.g., pharmaceuticals) in paired ecosystems across more specific regions should consolidate our understanding of soil contamination and its environmental risks.
We found similar levels of metal(loid)s such as Pb, Ni, Cd, and As in soils of urban greenspaces and adjacent natural areas, but urban greenspaces had higher contents of soil Hg, Zn, Cu, and Cr than those in natural ecosystems (Supplementary Fig. ). Differences in the accumulation of metal(loid)s in urban greenspaces compared with natural areas could be related to geographic variations in the extent of metal(loid) emissions from human activities, though the exact mechanisms remain unknown . The broad spectrum of anthropogenic impacts considered in this study (e.g., Pb and Hg) captures, at least partially, the potential anthropogenic sources of these elements. However, the natural sources and geochemical background of these trace metals are also critical in explaining levels of these metal(loid)s . Metal(loid)s in the topsoils of urban greenspaces may affect soil organisms and ecosystem processes when they exceed a specific threshold concentration. For instance, according to, the guidelines of the Finnish legislation for soil contamination , standards that are widely used to assess soil contamination, 42% of urban greenspace sites and 36% of sites in natural areas exceeded the lower limit for As contamination in soil (Supplementary Table ). As expected, however, average levels of these metal(loid)s in the remote Antarctica ecosystem were below the legislation limit (Supplementary Figs. and ). These findings advance our knowledge of the typical levels of toxic metal(loid)s found in urban greenspaces and adjacent natural areas worldwide. Accumulated contaminants in surface soils of greenspaces do not necessarily increase environmental risk because the bioavailability of soil contaminants depends on soil properties such as pH, texture, and organic carbon content . Similarly, pesticide residues were widely detected in many urban greenspaces and natural ecosystems at a global scale. We detected the presence of pesticide residues, including fungicides, herbicides, and insecticides, in 63% of the surveyed natural ecosystems (Supplementary Fig. ; Supplementary Table ). Our analyses likely underestimate these levels as we did not measure several widely used pesticides, such as glyphosate, glufosinate, paraquat, and 2,4-Dichlorophenoxyacetic acid. We did not detect a difference in the occurrence of pesticides between urban and natural greenspaces (Supplementary Fig. ). This is also evident when analyzing response ratios (urban vs. natural; Fig. ) or accounting for our paired design (Supplementary Fig. ). The application of persistent pesticides in urban areas may influence adjacent natural areas, for example through surface runoff and atmospheric transport and deposition . In addition, agricultural pesticide application may contribute to the reported levels of pesticide residues in adjacent urban greenspaces and natural areas by deposition. Microplastics, typical anthropogenic-driven contaminants, were also ubiquitous in soils of both urban greenspaces and natural ecosystems worldwide. Strikingly, we found similar contents of soil microplastics in the paired ecosystems worldwide (Fig. ; Supplementary Fig. ). The average level of microplastics was 916.2 items kg −1 soil (range from 166.7 to 3482.3 items kg −1 ), comparable to previously reported values , . The detected microplastics consisted predominately of polypropylene and polyester, with fibers as the dominant shape (Supplementary Fig. ), consistent with reports of microplastic levels in the ocean and terrestrial ecosystems . We also found similar proportions of the shape and polymer type of microplastics in natural areas and urban greenspaces (Fig. ; Supplementary Fig. ), which further supports the idea of a spillover of anthropogenic contaminants across ecosystems. These microplastics, which typically originate from cities, can also reach distant areas by atmospheric transport, with fibers being the main shape of airborne plastic particles in the atmosphere of cities such as Paris , London and Dongguan (China) . These fibers generally consist of polyester and polypropylene, which can originate from synthetic fabric, rope, and nets. In addition, semi-natural ecosystems can also accumulate microplastics due to human activities, leading to waste disposal /littering, runoff, and transport through the atmosphere . Unexpectedly, our results showed that soils in remote Antarctica had similar signatures of microplastics (in both concentration and type) in surface soils to our surveyed urban greenspaces worldwide (Supplementary Fig. ). This could be related to microplastic dispersion from Antarctic research stations and other continents by sea and air, though this would depend on wind strength and direction . For example, recent studies showed that more than 60% of microplastics detected in Antarctic samples were <50 µm, and these tiny-sized plastics can be transported over long distances before deposition , . Other activities, such as, tourism may also contribute to the accumulation of microplastics in the soils of Antarctica sites . Future studies should provide more detailed measurements of the characteristics of microplastics in order to track their distribution and fate in the environment. On the whole, our findings provide evidence that the level and characteristics of microplastics in natural areas match those in urban city parks and gardens, with multiple conservation and ethical implications for microplastic contamination of terrestrial ecosystems worldwide. We then characterized the richness and abundance of 285 ARGs in the soils from the 112 ecosystems studied (Fig. ; Supplementary Fig. ). Despite their natural origin (soil-borne bacteria), soil ARGs are emerging as potential biological contaminants threatening human health , , and the presence of some ARGs in soil has been linked to the widespread use of antibiotics in medical and agricultural industries (e.g., veterinary medicine) . The intense presence of pet excreta, atmospheric deposition, and population movement may be the main transfer routes of ARGs to the soil of urban greenspaces , . Thus we consider them an indirect measure of antibiotic influences in highly disturbed environments such as urban greenspaces. Our results show the ubiquity of ARGs in both urban greenspaces and natural areas, though the proportion of specific ARG types varied across locations (Supplementary Figs. and ). The similarity in ARGs between natural and urban soils might be related to the physical mobility of bacterial cells and plasmids carrying ARGs from urban to natural environments . We show that soils from urban greenspaces supported a greater diversity (i.e., richness) of ARGs than natural ecosystems, suggesting that urban areas may receive more diverse sources of ARGs from anthropogenic sources. These analyses advance our current understanding of ARGs by providing quantitative PCR data to investigate the diversity of ARGs in urban environments , though our analyses may not include those ARGs that are below detection limits. In addition, the significantly lower richness of ARGs in remote Antarctica (Supplementary Fig. ) further suggests that the presence of ARGs could be linked to human antibiotic use.
To provide further insights into their origin and anthropogenic contribution to soil contamination in urban and natural ecosystems, we investigated the association among socio-economic and environmental factors and soil contaminants. Structural equation modelling (SEM) and linear mixed-effects models consistently revealed that regional population density was the most important socio-economic factor associated with microplastics (Fig. ; Supplementary Figs. and ). This finding accords with local studies showing significant links between population density and environmental contamination , . Densely populated areas are typically associated with greater microplastic inputs from traffic, industrial emissions, domestic waste and the release of synthetic fibers , , or sewage sludge and compost used as soil fertilizers . We also found negative associations between city wealth (reflected in GDP and HDI, Supplementary Fig. ) and soil microplastics and metal(loid)s (Fig. ; Supplementary Fig. ). Significant associations between response ratios of soil contaminants and population density further underline the importance of human factors in the accumulation of soil contaminants, particularly microplastics in urban ecosystems (Supplementary Table ). According to our modelling analyses, human factors such as soil fertilization played more important roles than natural factors in influencing the abundance of pesticides and ARGs worldwide (Fig. ). For instance, despite the natural origin of ARGs, fertilization was positively associated with the abundance of ARGs (Fig. ). The widespread use of sewage sludge and animal manure containing antibiotic residues may have increased the presence of ARGs in fertilized urban sites . Taken together, we posit that human activities are causing a wide spillover of multiple soil contaminants across ecosystems, transporting contaminants between urban greenspaces and natural areas by atmospheric deposition and surface runoff. Soil contamination, however, is a complex process, and may be affected by other undocumented factors, many of which act at a local scale. Our statistical analyses are also limited in that we cannot assume causal relationships among these factors. Long-term field monitoring and controlled experiments are needed to strengthen our understanding of the diffusion and accumulation of soil contaminants across contrasting ecosystems. Our work points to the global magnitude of these processes and their links to human activities.
Finally, we used metagenomic data to establish potential associations among multiple soil contaminants and biological responses associated with key functional aspects maintaining ecosystem sustainability . We found, for example, that greater contents of soil Cr, Cu, Hg and Zn in urban greenspaces than in natural areas matched the larger proportion of genes associated with biological resistance (e.g., multidrug resistance) in urban environments (Fig. ; Supplementary Fig. , Table ). Moreover, the proportion of genes associated with resistance to streptothricin (an antibiotic) was positively correlated with elevated concentrations of Hg across soils (Supplementary Fig. ). These findings suggest that the soil functional microbiome is linked to metal(loid) concentrations. In addition, other well-known genes associated with stress resistance, nutrient cycling, pathogenesis and microbial metabolism were associated with a greater level of multiple soil contaminants when considered simultaneously (multi-contamination calculated as the averaged of standardized concentration of contaminants, see Methods) (Fig. ). For example, the soil multi-contamination index was positively correlated with the proportion of the genes associated with pathogenic Listeria prophages and multidrug resistance, but negatively correlated with the genes involved in P metabolism, Fe acquisition and transport, and DNA repair (Fig. ). These results suggest that the accumulation of contaminants potentially increased the pathogenicity of soil microbes and affected their functions. We also found negative associations among pesticides and genes associated with nutrient (e.g., N, P, and Fe) cycling and basic metabolisms (e.g., ABC transporter and 4Fe-4S ferredoxin) (Fig. ; Supplementary Figs. and ). Importantly, we found that the number of soil contaminants over certain thresholds was significantly correlated with the proportions of the selected functional genes (Supplementary Fig. ). These results indicate significant impacts of multiple contaminants on the soil functional microbiome, which is critical to supporting ecosystem services . However, gene abundance does not necessarily reflect functional activity , and further experiments are needed to improve our understanding of soil contamination impacts on functional attributes supported by the soil microbiome. Experimental studies manipulating soil contaminants and testing how they influence soil health are important to test further links between soil contaminations and functional gene activities observed here. Such knowledge will help to protect ecosystem functions from human-driven contamination across the globe efficiently.
Our study provides the cross-continental-scale assessment of the patterns and biological responses associated with soil contaminants in natural and urban areas. We provide evidence that soil contamination in natural areas surrounding cities mirrors that in urban greenspaces worldwide. This highlights the growing global contaminant footprint, which has reached even remote locations such as Antarctica. We also found that socio-economic factors are significantly associated with the occurrence of soil contaminants, with positive associations between population density and the levels of metal(loid)s and microplastics, highlighting the contribution of humans to soil contamination regardless of their origins. We further reveal the potential influence of soil contamination on key microbial traits such as stress resistance, nutrient cycling, and pathogenesis. These findings pave the way to the formulation of hypotheses that can be tested using controlled laboratory- and field-based studies. Together, our work demonstrates that soils in nearby natural areas are as contaminated as our urban greenspaces at a large-spatial scale. Future studies on a wider spectrum of contaminants (e.g., pharmaceuticals) in paired ecosystems across more specific regions should consolidate our understanding of soil contamination and its environmental risks.
Field survey and soil sampling Soil samples were collected from paired urban greenspaces and adjacent natural areas (i.e., natural and semi-natural ecosystems near cities that are not subjected to management) from 56 municipalities of 17 countries and six continents (Supplementary Fig. and Supplementary Table ). Natural areas were about 20 km from the urban greenspaces, which include forests, grasslands, shrublands, or relict forests with their original vegetation. Urban greenspaces were mainly public parks and large residential gardens, and comprised a mixture of open areas with lawns, scattered trees, patches of shrubs, and associated flowerbeds. All the selected urban greenspaces were well-established and many decades old. Our findings indicate that soils in urban and natural ecosystems were similar in their levels of contamination. A total of 112 ecosystems (56 paired urban vs natural areas) were investigated in this study. The mean annual precipitation and temperature ranged from 210 to 1577 mm and 3.1 °C to 26.4 °C, respectively. These natural areas are locations with no historically documented or evident human impact. The selected cities have a wide range of socio-economic development indices, representing different development levels of cities worldwide. For example, region population density and human development index (HDI, a summary measure of average achievement in key dimensions of human development) ranged from 7.7 to 16 people/km 2 and 0 to 0.89, respectively. We did not sample agricultural soils, and the selected sites were remote from any mining areas. We also considered soil contamination attributes and sampling feasibility when choosing sampling sites across such a broad range of environmental gradients worldwide. A 30 m × 30 m plot (900 m 2 ) consisting of three parallel transects of equal length was surveyed at each location . These plots were selected to represent the most common environments within ecosystems (e.g., a grass lawn or an urban forest in a city park). We then collected surface soils (top 5 cm) from the 112 ecosystems. To account for spatial heterogeneity, we collected three sampling points for a composite sample under the most common environments (vegetation and open areas between plant canopies covered by bare soils and nonvascular plants) found at each site (Supplementary Fig. ). Thus we obtained 336 composite soil samples in total. For comparison, we collected three additional composite soil samples (from three sampling points) from Antarctica, which is far from human activities. After field collection, soils were sieved (<2 mm) for downstream analyses. The sieves were carefully washed, and cleaned with 90% ethanol between samples to avoid cross- contamination. One sub-sample was air-dried for chemical (i.e., pH, organic carbon, total phosphorus, and nitrogen, metal(loid)s, pesticide residues, and microplastics) analyses, and the other soil fraction was frozen at −20 °C for the analyses of microbiome including ARGs. Due to sample availability and resource limitation, we did not analyze all soils for each category of contaminant. Many of the analyses included in this study (e.g., polymer identification of microplastics and 46 pesticide residues) are highly costly and time-consuming, which require a relatively large amount of sample for contaminant extraction (see Methods below). We thus focused on the subset of sites potentially impacted by specific contaminants according to advice from local greenspace managers and environmental scientists. Therefore, we attempted to cover the entire gradient of conditions by focusing on a subset of the samples. Thus, we measured metal(loid)s and ARGs in all the 336 composite samples from 112 plots, microplastics in 64 composite samples from the selected 64 plots and pesticide residues in 54 composite samples from the selected 54 plots (See Supplementary Fig. , Table ). For comparability, we standardized within-site replicates for metal(loid)s and ARGs (e.g., 112 plots × 3 composite samples = 336 samples). In all cases, the subsets of samples were selected to cover the entire biogeographic range along with a broad range of environmental gradients (Supplementary Fig. and Table ). To the best of our knowledge, this is the largest paired sampling dataset of soil contamination worldwide. We acknowledge, however, the potential limitations of an unbalanced number of samples for the analyses of the four categories of soil contaminants. Environmental factors Climatic information (i.e., mean annual precipitation and temperature) were extracted from the WorldClim database ( https://www.worldclim.org/data/index.html ), and averaged values of climatic parameters from recent years (2013–2021) were used in this study. Plant cover was determined at each site based on three transects across 30 m × 30 m plots. Information on the population was obtained from the latest available city censuses using official national statistical sources . GDP and HDI per capita (over 1990–2015) in the regions for the cities surveyed were extracted from the reported dataset , providing information on the economic activity and key dimensions of human development for each location. Moreover, we collected information about mowing, irrigation, and fertilization treatments by surveying the managers of urban greenspaces . Analyses of soil chemical properties Soil pH was determined on a 1: 5 soil/water extract using a pH electrode. Total carbon and nitrogen in the soil were analyzed using an elemental analyzer (C/N Flash EA 112 Series-Leco Truspec). Soil organic carbon was measured using the same elemental analyzer after fumigation with HCl . Total phosphorus was determined using an inductively coupled plasma optical emission spectrometer (ICAP6500 DUO; Thermo-Scientific) after digestion using nitric-perchloric acid. Analyses of soil contaminants For the analyses of Cu, Pb, Cd, Zn, Ni, Cr, and As, 0.25 g of each finely ground (<0.149 mm) soil was digested by a MARS microwave digestion system using mixed acids (9.0 mL of HNO3 and 3.0 mL of HF) . The concentrations of metal(loid)s in the digestion were subsequently determined by Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES) with Thermo ICP 6500 Duo equipment (Thermo Fisher Scientific, Waltham, MA, USA). In parallel, 0.20 g of each finely ground soil was digested with 3:1 aqua regia for total Hg analysis, and the amount of Hg in the digestion was measured using cold vapor atomic fluorescence spectrometry (CVAFS) . Analytical blanks were included in all analyses for quality control. The limit of detection in for each element was (in µg g −1 ): 0.0008 for As, 0.0006 for Cd, 0.0011 for Cr, 0.0012 for Cu, 0.0011 for Ni, 0.0008 for Pb, 0.0014 for Zn and 0.0001 for Hg. The pesticide residues in soils were extracted and measured as a previous method . Briefly, Accelerated Solvent Extraction (Dionex ASE 350, Thermo Scientific) was used to extract pesticide residues from 6 g of soil. The extraction contained two steps. In the first step, an organic mixture of acetone, methanol, and acetonitrile at a ratio of 65:10:25 (% v/v) was used. In the second, pesticide residues in soil were extracted with a mixture of acetone and 1% phosphoric acid in Millipore water at a ratio of 70:30 (% v/v). The extracts were processed, and the pesticide residues were analyzed by high-performance liquid chromatography coupled to a triple quadrupole tandem mass spectrometer (HPLC-MS/MS). Reversed-phase HPLC using water and methanol as mobile phase was used to perform the chromatographic separation. Detection was performed with a triple quadrupole mass spectrometer (QTrap 5500, Sciex) and all detected concentrations (µg L −1 ) were converted into µg per kg of dry soil. A total of 46 different pesticide residues were assessed. The limit of quantification ranged from 0.064 μg kg −1 to 36 μg kg −1 depending on the substance. We want to acknowledge that the analytical method was not designed to cover all pesticide residues potentially present in soil, but still provide a good overall view of the soil contamination associated with pesticides. Specifically, out of the 20 worldwide most applied active pesticides ingredients, our method can detect up to 30% of these substances, including the quantification of the frequently used active pesticide ingredients such as acetochlor, alachlor, atrazine, azoxystrobin, chlorpyrifos, clothianidin, diuron, imidacloprid propiconazole, S-metolachlor tebuconazole, and trifloxystrobin. Microplastics (i.e., plastics <5 mm in diameter) in soil were pretreated using Fenton’s oxidization method, while maintaining the integrity of the plastic particles. , The Fenton’s reagent was added slowly in small quantities to minimize the potential effect on the recovery of plastics. Ice bath was used to maintain temperatures below 40 °C. Samples were pretreated with Fenton’s reagent to oxidize organic matter for 5 hours. After that, the wall of the beakers was rinsed with ultrapure water to detach any microplastics and left to dry (at 45 °C) for 1 day. Microplastics were then extracted through density separation with a saturated NaCl solution, and the soil was stirred with a glass rod for 3 min and allowed to settle overnight. To improve the recovery rate of microplastics, we repeated the extraction of each soil sample four times and used calcium bicarbonate to promote flotation via the release of CO 2 bubbles . The supernatant of samples was subsequently decanted, filtered and stored for the next analysis . Filters were analyzed and counted visually using a binocular microscope glass (MOTIC SMZ 143 N2GG, Wetzlar, Germany) to find all the tentative microplastic particles, ranging from 5 mm to 20 µm. The shape (i.e., fiber, fragment, and film) of the particles was assessed according to a previous method . Noted that this method may have underestimated actual microplastic levels due to limitations in the extraction method. A representative number of particles were subsequently analyzed for the identification of polymer type by RAMAN spectroscopy (NRS-5100, Jasco, Madrid, Spain) equipped with LMU-20X-UVB lens. The laser excitation frequency and intensity used was 784.79 nm and 11.8 mW, respectively. Raman spectra were obtained between 162 and 1886 cm −1 with a spectral resolution of 2.47 cm −1 , and were recorded with a charge-coupled device camera (UV-NIR range, 1024 × 255 pixels) electrically cooled to −70 °C. The RAMAN spectra of the analyzed particles were finally compared to reference polymers from the spectral library Open Specy . Microplastics extracted from the soils were checked to ensure they were not contaminated by plastic bags used in the collection and storage processes. We determined the spectroscopic signature of all the bags used for the sampling, which did not match any of the signature of microplastics found in the samples. To minimize microplastic contamination of samples throughout the laboratory procedures, the plastic material was avoided as much as possible, using glass material that had been cleaned with ultrapure water, cotton laboratory coats and, whenever possible, clothes that were made of natural fibers. The NaCl solution was always previously filtered to remove any possible microplastics in the salt . Additionally, containers holding the samples were covered to prevent airborne contamination of microplastics. To control for possible cross-contamination, procedural blanks were used during all the procedures in the laboratory and no microplastics were detected in them . Genomic DNA from soils of urban greenspaces and natural areas was extracted using the PowerSoil DNA Isolation Kit (QIAGEN Inc., Germany) following the manufacturer’s instruction. A high-throughput quantitative PCR (HT-qPCR) based chip was used to quantify ARGs on the Wafergen SmartChip Real-Time PCR System (Fremont, CA, USA). The ARG chip consisted of 285 primer sets targeting ARGs conferring resistance to all major classes of antibiotics, including aminoglycoside, beta-lactams, FCA (fluoroquinolone, quinolone, florfenicol, chloramphenicol and amphenicol), macrolide-lincosamide-streptogramin B (MLSB), multidrug, sulfonamide, tetracycline and vancomycin (Supplementary Table ). Meanwhile, the 16S ribosomal RNA gene was included as a reference gene. All HT-qPCR reactions were performed in technical triplicates with a negative control. Analyses of microbial functional traits In total, 54 composite topsoil samples (one composite sample per plot) from 27 typical paired urban and natural areas were selected for metagenomic analyses. Sufficient amounts (roughly 500 ng) of microbial DNA were extracted using the DNeasy PowerSoil DNA Isolation Kit (QIAGEN Inc., Germany) according to the manufacturer’s protocol. Shotgun sequencing was performed using an Illumina HiSeq (Illumina Inc., USA) at Majorbio in Shanghai, China. Raw reads (PE150, 150 bp pairedend reads) were trimmed to remove low quality reads. The SeqPrep software ( https://github.com/jstjohn/SeqPrep ) was used to remove adapter sequences. Then, the library sickle ( https://github.com/najoshi/sickle ) was used to trim the reads from the 5′ end to 3′end using a sliding window (size 50 bp, 1 bp step). All reads below the quality threshold of 20 were trimmed. The resultant reads below 50 bp or containing N (ambiguous bases) were discarded. The trimmed reads were then mapped against the protein sequence of the BacMet ( http://bacmet.biomedicine.gu.se/ ) and Subsystem Technology (MG-RAST; https://www.mg-rast.org ) , respectively. MG-RAST generates taxonomic assignments based on the SEED subsystem database by DIAMOND software (version 0.9.32) by best-hit classification with a maximum E-value of 1e −5 , a minimum identity of 60%, and a minimum alignment length of 25 amino acids for proteins and functional categories. The resulting table was parsed at SEED Subsystem Level 3 by the software SUPER-FOCUS. The relative abundances of annotated genes were calculated using the number of reads mapping to genes for each sample. We focused on the genes associated with ecosystem functions such as the resistance to environmental stress, pathogenesis, nutrient cycling and microbial metabolism. The resistance traits included the resistant genes to stress of metal(loid)s, drugs, and pathogens. Detailed information on microbial traits was assessed by the gene abundances in the soil microbiome is shown in Supplementary Table . Statistical analyses We first calculated a soil contaminant index for each category by averaging standardized values of the individual soil contaminants. To further provide an overall assessment of soil contaminants, we calculated the multi-contamination index by averaging standardized values of four categories of examined soil contaminants including metal(loid)s, ARGs, pesticide residues and microplastics. We used the following standardization approach so that all variables contribute similarly to the multi-contamination indices (1): 1 [12pt]{minimal}
$${standardization}=\,{content}- (\,{variable}\,{content})]}{[ (\,{variable}\,{content})- ({variable}\,{content})]}$$ s t a n d a r d i z a t i o n = v a r i a b l e c o n t e n t − min v a r i a b l e c o n t e n t [ max v a r i a b l e c o n t e n t − min ( v a r i a b l e c o n t e n t ) ] variable is the level of the examined soil contaminant from a specific site, and the min (variable content) and max (variable content) represent the minimum and maximum level of each soil contaminant among all the examined sites. Thus, each transformed variable had a minimum value of zero and a maximum value of one. This approach is similar to the methods used for calculating ecosystem multifunctionality , . It is noted that only the sites simultaneously having data for four categories of soil contaminants were selected to calculate the multi-contamination index ( n = 48 sites). Levels of metal(loid)s and ARGs derived from three composite soil samples per plot were averaged to obtain plot-level estimations prior to statistical analyses , . This allowed us to relate the spatial heterogeneity within plots to the metrics of soil contaminants. We calculated response ratios of contaminants in the urban greenspaces vs. adjacent natural ecosystems. Response ratios were used to assess the enrichment of examined soil contaminants in urban greenspaces by comparing them with the adjacent natural ecosystems as follows (2): 2 [12pt]{minimal}
$${{{{{}}}}}\; {{{{{}}}}}=}}}}}\; {{{{{}}}}}\; {{{{{}}}}}\; {{{{{}}}}}-{{{{{}}}}}\; {{{{{}}}}}\; {{{{{}}}}}\; {{{{{}}}}}}{{{{{{}}}}}\; {{{{{}}}}}\; {{{{{}}}}}\; {{{{{}}}}}} 100$$ Response ratio = Value variable A urban − Value variable A natural Value variable A natural × 100 We also used a blocked design to test differences in contents of contaminants (i.e., metal(loid)s pesticides, microplastics, and ARGs) between urban greenspaces and the adjacent natural ecosystems accounting for paired locations (i.e., we fixed location in the strata), providing complementary statistical support to the response ratios. These analyses were performed using nested PERMANOVA (permutations = 999) in the R package “Vegan” (“adonis” routine) in R 4.0.3. We used structural equation modelling (SEM) to identify the relative importance of socio-economic vs. environmental factors in explaining the accumulation of four major types of contaminants (i.e., metal(loid)s, microplastics, pesticide residues and ARGs) from the natural and urban ecosystems. An a priori model was established based on our current knowledge regarding environmental impacts on soil contamination (Supplementary Fig. ). These factors include socio-economic (i.e., population size and density, HDI and GDP), soil factors (i.e., pH, and organic carbon, total phosphorus and nitrogen), plant cover, climate (mean annual precipitation and temperature) and land management (i.e., fertilization, irrigation and mowing of urban greenspaces). We assume that socio-economic factors, land management and climate generally influence plant and soil variables, and therefore directly and indirectly affect the accumulation of soil contaminants. For example, countries with a high GDP generally invest more to waste management and have more restrictive policies in terms of environment protection, compared with countries with low GDP. The management practices were included in our SEM as categorical variables with two levels: 1 (a given management practice) and 0 (no management practice). Information on variables included in the model is either measured in situ or in the laboratory, or by collecting data from the literature as explained above. The SEMs allowed us to determine the direct and indirect effects of multiple environmental factors on soil contaminants. The probability that a path coefficient differs from zero was tested using bootstrap resampling. Doing this instead of assessing significance with frequentist methods allowed us to overcome limitations related to the data matching a particular theoretical distribution. The data matrix was fitted to the model using the maximum-likelihood estimation method. There is no single universally accepted test of the overall goodness of fit for SEM. Thus, we used the Chi-square test (χ 2 ; the model has a good fit when 0 ≤ χ 2 /d.o.f ≤ 2 and 0.05 < P ≤ 1.00) and the root mean square error of approximation (the model has a good fit when 0 ≤ RMSEA ≤ 0.05 and 0.10 < P ≤ 1.00). The SEM analyses were performed using AMOS 21.0 (SPSS Inc., Chicago, IL, USA). In parallel, we evaluated the relative importance of individual natural and human factors on the four types of soil contaminants by a mixed-effects model using the “glmulti” package in R. The importance of each predictor was expressed as the sum of Akaike weights for models. We used a cutoff of 0.8 to differentiate between essential and nonessential predictors as done in a previous study . We also used Spearman’s correlation analyses to evaluate associations between selected factors and soil contaminants and the response ratios. We explored associations between potential soil contaminants and the selected functional genes, and observational correlations are essential to understand ecosystems and develop new hypotheses . For simplicity and research focus, we conducted ordinary least square linear regressions between soil multi-contamination index and the proportions of the genes associated with environmental stress, pathogenesis, nutrient cycling and basic metabolisms. To strengthen our understanding of soil contaminant impacts, we analyzed the relationships between the number of soil contaminants over four thresholds of maximum contaminant levels and the relative abundances of selected functional genes (see Supplementary Method for details). In addition, we used a co-occurrence network approach to visualize the associations between the soil contaminants (i.e., metal(loid)s, pesticides, microplastics and ARGs) and selected functional genes. To achieve this, Spearman correlations between soil contaminants and these functional genes were calculated using R package “psych”, and we considered P < 0.05 as statistical significance. The network was then visualized using the Cytoscape software. We also evaluated the relationships between specific soil contaminants and the proportions of these functional genes. Reporting summary Further information on research design is available in the linked to this article.
Soil samples were collected from paired urban greenspaces and adjacent natural areas (i.e., natural and semi-natural ecosystems near cities that are not subjected to management) from 56 municipalities of 17 countries and six continents (Supplementary Fig. and Supplementary Table ). Natural areas were about 20 km from the urban greenspaces, which include forests, grasslands, shrublands, or relict forests with their original vegetation. Urban greenspaces were mainly public parks and large residential gardens, and comprised a mixture of open areas with lawns, scattered trees, patches of shrubs, and associated flowerbeds. All the selected urban greenspaces were well-established and many decades old. Our findings indicate that soils in urban and natural ecosystems were similar in their levels of contamination. A total of 112 ecosystems (56 paired urban vs natural areas) were investigated in this study. The mean annual precipitation and temperature ranged from 210 to 1577 mm and 3.1 °C to 26.4 °C, respectively. These natural areas are locations with no historically documented or evident human impact. The selected cities have a wide range of socio-economic development indices, representing different development levels of cities worldwide. For example, region population density and human development index (HDI, a summary measure of average achievement in key dimensions of human development) ranged from 7.7 to 16 people/km 2 and 0 to 0.89, respectively. We did not sample agricultural soils, and the selected sites were remote from any mining areas. We also considered soil contamination attributes and sampling feasibility when choosing sampling sites across such a broad range of environmental gradients worldwide. A 30 m × 30 m plot (900 m 2 ) consisting of three parallel transects of equal length was surveyed at each location . These plots were selected to represent the most common environments within ecosystems (e.g., a grass lawn or an urban forest in a city park). We then collected surface soils (top 5 cm) from the 112 ecosystems. To account for spatial heterogeneity, we collected three sampling points for a composite sample under the most common environments (vegetation and open areas between plant canopies covered by bare soils and nonvascular plants) found at each site (Supplementary Fig. ). Thus we obtained 336 composite soil samples in total. For comparison, we collected three additional composite soil samples (from three sampling points) from Antarctica, which is far from human activities. After field collection, soils were sieved (<2 mm) for downstream analyses. The sieves were carefully washed, and cleaned with 90% ethanol between samples to avoid cross- contamination. One sub-sample was air-dried for chemical (i.e., pH, organic carbon, total phosphorus, and nitrogen, metal(loid)s, pesticide residues, and microplastics) analyses, and the other soil fraction was frozen at −20 °C for the analyses of microbiome including ARGs. Due to sample availability and resource limitation, we did not analyze all soils for each category of contaminant. Many of the analyses included in this study (e.g., polymer identification of microplastics and 46 pesticide residues) are highly costly and time-consuming, which require a relatively large amount of sample for contaminant extraction (see Methods below). We thus focused on the subset of sites potentially impacted by specific contaminants according to advice from local greenspace managers and environmental scientists. Therefore, we attempted to cover the entire gradient of conditions by focusing on a subset of the samples. Thus, we measured metal(loid)s and ARGs in all the 336 composite samples from 112 plots, microplastics in 64 composite samples from the selected 64 plots and pesticide residues in 54 composite samples from the selected 54 plots (See Supplementary Fig. , Table ). For comparability, we standardized within-site replicates for metal(loid)s and ARGs (e.g., 112 plots × 3 composite samples = 336 samples). In all cases, the subsets of samples were selected to cover the entire biogeographic range along with a broad range of environmental gradients (Supplementary Fig. and Table ). To the best of our knowledge, this is the largest paired sampling dataset of soil contamination worldwide. We acknowledge, however, the potential limitations of an unbalanced number of samples for the analyses of the four categories of soil contaminants.
Climatic information (i.e., mean annual precipitation and temperature) were extracted from the WorldClim database ( https://www.worldclim.org/data/index.html ), and averaged values of climatic parameters from recent years (2013–2021) were used in this study. Plant cover was determined at each site based on three transects across 30 m × 30 m plots. Information on the population was obtained from the latest available city censuses using official national statistical sources . GDP and HDI per capita (over 1990–2015) in the regions for the cities surveyed were extracted from the reported dataset , providing information on the economic activity and key dimensions of human development for each location. Moreover, we collected information about mowing, irrigation, and fertilization treatments by surveying the managers of urban greenspaces .
Soil pH was determined on a 1: 5 soil/water extract using a pH electrode. Total carbon and nitrogen in the soil were analyzed using an elemental analyzer (C/N Flash EA 112 Series-Leco Truspec). Soil organic carbon was measured using the same elemental analyzer after fumigation with HCl . Total phosphorus was determined using an inductively coupled plasma optical emission spectrometer (ICAP6500 DUO; Thermo-Scientific) after digestion using nitric-perchloric acid.
For the analyses of Cu, Pb, Cd, Zn, Ni, Cr, and As, 0.25 g of each finely ground (<0.149 mm) soil was digested by a MARS microwave digestion system using mixed acids (9.0 mL of HNO3 and 3.0 mL of HF) . The concentrations of metal(loid)s in the digestion were subsequently determined by Inductively Coupled Plasma-Optical Emission Spectrometry (ICP-OES) with Thermo ICP 6500 Duo equipment (Thermo Fisher Scientific, Waltham, MA, USA). In parallel, 0.20 g of each finely ground soil was digested with 3:1 aqua regia for total Hg analysis, and the amount of Hg in the digestion was measured using cold vapor atomic fluorescence spectrometry (CVAFS) . Analytical blanks were included in all analyses for quality control. The limit of detection in for each element was (in µg g −1 ): 0.0008 for As, 0.0006 for Cd, 0.0011 for Cr, 0.0012 for Cu, 0.0011 for Ni, 0.0008 for Pb, 0.0014 for Zn and 0.0001 for Hg. The pesticide residues in soils were extracted and measured as a previous method . Briefly, Accelerated Solvent Extraction (Dionex ASE 350, Thermo Scientific) was used to extract pesticide residues from 6 g of soil. The extraction contained two steps. In the first step, an organic mixture of acetone, methanol, and acetonitrile at a ratio of 65:10:25 (% v/v) was used. In the second, pesticide residues in soil were extracted with a mixture of acetone and 1% phosphoric acid in Millipore water at a ratio of 70:30 (% v/v). The extracts were processed, and the pesticide residues were analyzed by high-performance liquid chromatography coupled to a triple quadrupole tandem mass spectrometer (HPLC-MS/MS). Reversed-phase HPLC using water and methanol as mobile phase was used to perform the chromatographic separation. Detection was performed with a triple quadrupole mass spectrometer (QTrap 5500, Sciex) and all detected concentrations (µg L −1 ) were converted into µg per kg of dry soil. A total of 46 different pesticide residues were assessed. The limit of quantification ranged from 0.064 μg kg −1 to 36 μg kg −1 depending on the substance. We want to acknowledge that the analytical method was not designed to cover all pesticide residues potentially present in soil, but still provide a good overall view of the soil contamination associated with pesticides. Specifically, out of the 20 worldwide most applied active pesticides ingredients, our method can detect up to 30% of these substances, including the quantification of the frequently used active pesticide ingredients such as acetochlor, alachlor, atrazine, azoxystrobin, chlorpyrifos, clothianidin, diuron, imidacloprid propiconazole, S-metolachlor tebuconazole, and trifloxystrobin. Microplastics (i.e., plastics <5 mm in diameter) in soil were pretreated using Fenton’s oxidization method, while maintaining the integrity of the plastic particles. , The Fenton’s reagent was added slowly in small quantities to minimize the potential effect on the recovery of plastics. Ice bath was used to maintain temperatures below 40 °C. Samples were pretreated with Fenton’s reagent to oxidize organic matter for 5 hours. After that, the wall of the beakers was rinsed with ultrapure water to detach any microplastics and left to dry (at 45 °C) for 1 day. Microplastics were then extracted through density separation with a saturated NaCl solution, and the soil was stirred with a glass rod for 3 min and allowed to settle overnight. To improve the recovery rate of microplastics, we repeated the extraction of each soil sample four times and used calcium bicarbonate to promote flotation via the release of CO 2 bubbles . The supernatant of samples was subsequently decanted, filtered and stored for the next analysis . Filters were analyzed and counted visually using a binocular microscope glass (MOTIC SMZ 143 N2GG, Wetzlar, Germany) to find all the tentative microplastic particles, ranging from 5 mm to 20 µm. The shape (i.e., fiber, fragment, and film) of the particles was assessed according to a previous method . Noted that this method may have underestimated actual microplastic levels due to limitations in the extraction method. A representative number of particles were subsequently analyzed for the identification of polymer type by RAMAN spectroscopy (NRS-5100, Jasco, Madrid, Spain) equipped with LMU-20X-UVB lens. The laser excitation frequency and intensity used was 784.79 nm and 11.8 mW, respectively. Raman spectra were obtained between 162 and 1886 cm −1 with a spectral resolution of 2.47 cm −1 , and were recorded with a charge-coupled device camera (UV-NIR range, 1024 × 255 pixels) electrically cooled to −70 °C. The RAMAN spectra of the analyzed particles were finally compared to reference polymers from the spectral library Open Specy . Microplastics extracted from the soils were checked to ensure they were not contaminated by plastic bags used in the collection and storage processes. We determined the spectroscopic signature of all the bags used for the sampling, which did not match any of the signature of microplastics found in the samples. To minimize microplastic contamination of samples throughout the laboratory procedures, the plastic material was avoided as much as possible, using glass material that had been cleaned with ultrapure water, cotton laboratory coats and, whenever possible, clothes that were made of natural fibers. The NaCl solution was always previously filtered to remove any possible microplastics in the salt . Additionally, containers holding the samples were covered to prevent airborne contamination of microplastics. To control for possible cross-contamination, procedural blanks were used during all the procedures in the laboratory and no microplastics were detected in them . Genomic DNA from soils of urban greenspaces and natural areas was extracted using the PowerSoil DNA Isolation Kit (QIAGEN Inc., Germany) following the manufacturer’s instruction. A high-throughput quantitative PCR (HT-qPCR) based chip was used to quantify ARGs on the Wafergen SmartChip Real-Time PCR System (Fremont, CA, USA). The ARG chip consisted of 285 primer sets targeting ARGs conferring resistance to all major classes of antibiotics, including aminoglycoside, beta-lactams, FCA (fluoroquinolone, quinolone, florfenicol, chloramphenicol and amphenicol), macrolide-lincosamide-streptogramin B (MLSB), multidrug, sulfonamide, tetracycline and vancomycin (Supplementary Table ). Meanwhile, the 16S ribosomal RNA gene was included as a reference gene. All HT-qPCR reactions were performed in technical triplicates with a negative control.
In total, 54 composite topsoil samples (one composite sample per plot) from 27 typical paired urban and natural areas were selected for metagenomic analyses. Sufficient amounts (roughly 500 ng) of microbial DNA were extracted using the DNeasy PowerSoil DNA Isolation Kit (QIAGEN Inc., Germany) according to the manufacturer’s protocol. Shotgun sequencing was performed using an Illumina HiSeq (Illumina Inc., USA) at Majorbio in Shanghai, China. Raw reads (PE150, 150 bp pairedend reads) were trimmed to remove low quality reads. The SeqPrep software ( https://github.com/jstjohn/SeqPrep ) was used to remove adapter sequences. Then, the library sickle ( https://github.com/najoshi/sickle ) was used to trim the reads from the 5′ end to 3′end using a sliding window (size 50 bp, 1 bp step). All reads below the quality threshold of 20 were trimmed. The resultant reads below 50 bp or containing N (ambiguous bases) were discarded. The trimmed reads were then mapped against the protein sequence of the BacMet ( http://bacmet.biomedicine.gu.se/ ) and Subsystem Technology (MG-RAST; https://www.mg-rast.org ) , respectively. MG-RAST generates taxonomic assignments based on the SEED subsystem database by DIAMOND software (version 0.9.32) by best-hit classification with a maximum E-value of 1e −5 , a minimum identity of 60%, and a minimum alignment length of 25 amino acids for proteins and functional categories. The resulting table was parsed at SEED Subsystem Level 3 by the software SUPER-FOCUS. The relative abundances of annotated genes were calculated using the number of reads mapping to genes for each sample. We focused on the genes associated with ecosystem functions such as the resistance to environmental stress, pathogenesis, nutrient cycling and microbial metabolism. The resistance traits included the resistant genes to stress of metal(loid)s, drugs, and pathogens. Detailed information on microbial traits was assessed by the gene abundances in the soil microbiome is shown in Supplementary Table .
We first calculated a soil contaminant index for each category by averaging standardized values of the individual soil contaminants. To further provide an overall assessment of soil contaminants, we calculated the multi-contamination index by averaging standardized values of four categories of examined soil contaminants including metal(loid)s, ARGs, pesticide residues and microplastics. We used the following standardization approach so that all variables contribute similarly to the multi-contamination indices (1): 1 [12pt]{minimal}
$${standardization}=\,{content}- (\,{variable}\,{content})]}{[ (\,{variable}\,{content})- ({variable}\,{content})]}$$ s t a n d a r d i z a t i o n = v a r i a b l e c o n t e n t − min v a r i a b l e c o n t e n t [ max v a r i a b l e c o n t e n t − min ( v a r i a b l e c o n t e n t ) ] variable is the level of the examined soil contaminant from a specific site, and the min (variable content) and max (variable content) represent the minimum and maximum level of each soil contaminant among all the examined sites. Thus, each transformed variable had a minimum value of zero and a maximum value of one. This approach is similar to the methods used for calculating ecosystem multifunctionality , . It is noted that only the sites simultaneously having data for four categories of soil contaminants were selected to calculate the multi-contamination index ( n = 48 sites). Levels of metal(loid)s and ARGs derived from three composite soil samples per plot were averaged to obtain plot-level estimations prior to statistical analyses , . This allowed us to relate the spatial heterogeneity within plots to the metrics of soil contaminants. We calculated response ratios of contaminants in the urban greenspaces vs. adjacent natural ecosystems. Response ratios were used to assess the enrichment of examined soil contaminants in urban greenspaces by comparing them with the adjacent natural ecosystems as follows (2): 2 [12pt]{minimal}
$${{{{{}}}}}\; {{{{{}}}}}=}}}}}\; {{{{{}}}}}\; {{{{{}}}}}\; {{{{{}}}}}-{{{{{}}}}}\; {{{{{}}}}}\; {{{{{}}}}}\; {{{{{}}}}}}{{{{{{}}}}}\; {{{{{}}}}}\; {{{{{}}}}}\; {{{{{}}}}}} 100$$ Response ratio = Value variable A urban − Value variable A natural Value variable A natural × 100 We also used a blocked design to test differences in contents of contaminants (i.e., metal(loid)s pesticides, microplastics, and ARGs) between urban greenspaces and the adjacent natural ecosystems accounting for paired locations (i.e., we fixed location in the strata), providing complementary statistical support to the response ratios. These analyses were performed using nested PERMANOVA (permutations = 999) in the R package “Vegan” (“adonis” routine) in R 4.0.3. We used structural equation modelling (SEM) to identify the relative importance of socio-economic vs. environmental factors in explaining the accumulation of four major types of contaminants (i.e., metal(loid)s, microplastics, pesticide residues and ARGs) from the natural and urban ecosystems. An a priori model was established based on our current knowledge regarding environmental impacts on soil contamination (Supplementary Fig. ). These factors include socio-economic (i.e., population size and density, HDI and GDP), soil factors (i.e., pH, and organic carbon, total phosphorus and nitrogen), plant cover, climate (mean annual precipitation and temperature) and land management (i.e., fertilization, irrigation and mowing of urban greenspaces). We assume that socio-economic factors, land management and climate generally influence plant and soil variables, and therefore directly and indirectly affect the accumulation of soil contaminants. For example, countries with a high GDP generally invest more to waste management and have more restrictive policies in terms of environment protection, compared with countries with low GDP. The management practices were included in our SEM as categorical variables with two levels: 1 (a given management practice) and 0 (no management practice). Information on variables included in the model is either measured in situ or in the laboratory, or by collecting data from the literature as explained above. The SEMs allowed us to determine the direct and indirect effects of multiple environmental factors on soil contaminants. The probability that a path coefficient differs from zero was tested using bootstrap resampling. Doing this instead of assessing significance with frequentist methods allowed us to overcome limitations related to the data matching a particular theoretical distribution. The data matrix was fitted to the model using the maximum-likelihood estimation method. There is no single universally accepted test of the overall goodness of fit for SEM. Thus, we used the Chi-square test (χ 2 ; the model has a good fit when 0 ≤ χ 2 /d.o.f ≤ 2 and 0.05 < P ≤ 1.00) and the root mean square error of approximation (the model has a good fit when 0 ≤ RMSEA ≤ 0.05 and 0.10 < P ≤ 1.00). The SEM analyses were performed using AMOS 21.0 (SPSS Inc., Chicago, IL, USA). In parallel, we evaluated the relative importance of individual natural and human factors on the four types of soil contaminants by a mixed-effects model using the “glmulti” package in R. The importance of each predictor was expressed as the sum of Akaike weights for models. We used a cutoff of 0.8 to differentiate between essential and nonessential predictors as done in a previous study . We also used Spearman’s correlation analyses to evaluate associations between selected factors and soil contaminants and the response ratios. We explored associations between potential soil contaminants and the selected functional genes, and observational correlations are essential to understand ecosystems and develop new hypotheses . For simplicity and research focus, we conducted ordinary least square linear regressions between soil multi-contamination index and the proportions of the genes associated with environmental stress, pathogenesis, nutrient cycling and basic metabolisms. To strengthen our understanding of soil contaminant impacts, we analyzed the relationships between the number of soil contaminants over four thresholds of maximum contaminant levels and the relative abundances of selected functional genes (see Supplementary Method for details). In addition, we used a co-occurrence network approach to visualize the associations between the soil contaminants (i.e., metal(loid)s, pesticides, microplastics and ARGs) and selected functional genes. To achieve this, Spearman correlations between soil contaminants and these functional genes were calculated using R package “psych”, and we considered P < 0.05 as statistical significance. The network was then visualized using the Cytoscape software. We also evaluated the relationships between specific soil contaminants and the proportions of these functional genes.
Further information on research design is available in the linked to this article.
Supplementary Information Peer Review File Reporting Summary
|
A neurophysiological perspective on the integration between incidental learning and cognitive control | d73a0b01-3523-4583-b83b-14eb5b90c9fd | 10042851 | Physiology[mh] | Changing the course of action to meet our goal requires effort, therefore, it is paramount to recognise when it is necessary to do so – . A collection of functions, often labelled as cognitive control, can be recruited to meet one’s goal , . For instance, when stimulus properties trigger incompatible stimulus-response (S-R) representations or response tendencies, detection of the cognitive conflict typically evokes higher-order cortical processes to redirect behaviour in a goal-directed manner , , . Cognitive control’s main function is to adjust the engagement of effortful, task-related mechanisms in accordance with the mentally held task set . Monitoring cognitive conflict is not an isolated process, as it is also influenced by contextual cues , – . Through repetition, regularities in the context can be learnt and subsequently used to build predictions on the upcoming demand for cognitive control , , . Here we investigated how incidental learning influences cognitive control at the levels of behavioural adaptation and neurophysiological signatures. The current study aimed to connect current trends in the fields of cognitive control and incidental learning, and by doing so we promote an integrated view between the implications of intentional and incidental forms of adaptive behaviour . Imagine that you learn a new dance with a partner. As you gain more expertise, your moves follow a predefined sequence of steps and executing them will not require much attention. At the same time, you also need to anticipate when your partner would like to change directions and adapt your steps accordingly. If you are preoccupied with how and when to make changes, the dance will never be smooth. If you do not notice the changes at the right time, one of yours might leave the dance floor with an aching foot. In a lab environment without a dance floor, a typical paradigm to study cognitive control is the colour-word Stroop task . Here, participants are asked to respond to the colour of the stimulus while ignoring the meaning of the word. The congruency between the perceptual and semantic information (i.e., the word “blue” written in blue colour) facilitates response selection compared to a neutral condition (i.e., blue colour target without a semantic association). However, incongruency between the task-relevant colour and task-irrelevant semantic stimulus dimensions (i.e., the word blue written in yellow colour) leads to more errors and slower responses. Contextual cues can modulate performance in a Stroop task. For instance, the list-wide proportion congruence (LWPC) effect , is observed through the manipulation of the overall ratio between congruent and incongruent trials. If the trials are predominantly incongruent (high level of conflict), conflict-related response cost is reduced in comparison to a predominantly congruent set of trials (low level of conflict) , . That is, distributional statistics of the stimuli influence cognitive control. An example of trial-by-trial contextual adaptation is the congruency sequence effect (CSE). Smaller conflict-related response cost is observed after a high conflict trial (incongruent) than after a low conflict event (congruent) , . That is, recent memory on the conflict level has predictive power on subsequent conflict detection and response control . The robustness of CSE raises the question of whether transitions between Stroop trials can be learnt and if yes, does learning of conditional probabilities contribute to cognitive control? In a modified Stroop task, when sequential probabilities predicted the next stimulus’ colour, a smaller congruency effect was detected compared to a random series of trials . Thus, incidental acquisition of transitional probabilities and cognitive control may engage in a cooperative fashion. Consequently, both distributional and conditional probability information could influence cognitive control by building predictions on conflict demand to reduce the related response cost. This cooperative mode has been put forward by the notion of control state associations . It has been proposed that cooperation between cognitive conflict and incidental learning processes occurs through abstract categories or “control states”. Control states (C) represent the allocation of top-down resources, such as visual attention or response inhibition . A control state can integrate (i) internal models of the task goal, (ii) implicit memories associated with the current S/R/C features, and (iii) pre-existing biases (i.e., habits, stereotypical responses, etc.) . Control states are represented in an associative S-C network similar to the network of S-R associations , , . Through these networks, a specific stimulus can activate a context-appropriate control state . Whenever predictions based on S-R properties are sufficient to guide the behaviour, the control state representation does not need to be involved. However, if the original (lower level) predictions fail, the association between the appropriate control state and stimulus contingency has to be used – , . These bindings between contextual information (stimulus) and control state are the so-called S-C event files or “episode files” . Incidental learning of S-C contingencies can potentially predict the need for top-down engagement and alleviate response adaptation. Conflict detection can also serve as a teaching signal to promote learning by modifying the focus of attention or prioritising the sensory input , , . In sum, S-C representations allow cooperation between incidental learning and intentional control processes. In a recent study , Stroop trials were presented according to either colour-based (S-R) or congruency-based (S-C) sequential regularities. Interestingly, the two groups showed comparable effects of conflict. However, predictability facilitated responses only in the S-R group, and this effect was independent of congruency. Thus, the study did not support cooperation between learning and control through S-C predictability. However, it is possible that the brief training did not allow the participants to form statistical memories that are more complex than the distribution of congruent and incongruent conditions. Therefore, the current study followed training protocols in which participants were able to learn transitional relations incidentally , . Another consideration was that conditional probabilities can define relations not only between adjacent sequential items (first-order transitions) but also between non-adjacent ones (second-order transitions) – . The two forms of learning also involve different neurocognitive mechanisms: since the processing of non-adjacent relations necessitates the suppression of the interleaving item, it has been thought that learning of non-adjacent dependencies requires an engagement of top-down processes . In contrast, learning of the more simple adjacent dependencies is a purely automatic, bottom-up mechanism , , . In the current study, S-C relations were defined as non-adjacent, second-order transitions in contrast to the adjacent dependencies used by Jimenez et al. . Figure provides details about the experimental approach. Multiple coding of regularities: statistical and rule-based learning Sequence learning is not a unitary construct, but involves different processes that detect, encode and retrieve sequentially presented regularities , , – . A prominent distinction is between the acquisition of probabilistic information (statistical learning) and sequential order , , , . This latter process is called higher-order sequence learning , , order-based learning or rule-based learning , . Within sequence learning, statistical learning can be operationalized as the differentiation between high- and low-probability elements that otherwise carry indistinguishable information about the serial order . In contrast, rule-based learning is the differentiation between elements that are defined by the serial order and those that are presented randomly, while the probability of occurrence remains the same , . While statistical learning enables the acquisition of distributions and probabilistic interdependence in the stimulus stream, rule-based learning controls and integrates learning results into higher-order regulations . These learning functions operate in overlapping time windows, however, they have different time courses , : fast estimation of statistics reaches its plateau early in contrast to the more gradual curve of rule-based learning. In other words, a larger rule-based learning effect can be expected at the end of the task than at the beginning. In contrast, statistical learning shows less variability over time. Moreover, the two learning processes have different interdependence on other cognitive functions , , , they are related to different neural sources , and are thought to generate differentiable expectations about upcoming events . Statistical learning is considered to be a more bottom-up process that develops fast in the presence of recurring patterns in the sensory experience , , . In contrast, rule-based learning operates with the gradual accumulation of sequential rules, and the learnt associations are also available for top-down processes , , . Considering the simultaneous nature of statistical and rule-based learning, the current study will investigate the potential contribution of both processes to cognitive control. The two forms of sequence learning may have different capacities to interact with cognitive control. Due to their simultaneity, it is often hard to disentangle the two processes at the behavioural level, however, the analysis of the concurrent neurophysiological signal can provide confirmation of the behavioural effects and further insights into the mechanisms behind them , , , – . The current study investigated concurrent incidental sequence learning and cognitive control by considering neurophysiological information that are differentially related to cognitive control and contextual learning. Cascade of processes in the neurophysiological signal In a Stroop task, a frontocentral negative deflection is typically observed with a peak of approximately 450 ms after stimulus presentation . This so-called N450 event-related potential (ERP) component’s amplitude is larger in incongruent than in congruent trials, therefore, it is an ideal candidate to study conflict detection – . Therefore, the N450 was selected as a more specific marker of conflict detection in the current study. Another benefit of using the ERP approach is the potential to differentiate between a cascade of processes , . For instance, conflict detection and the retrieval of previously learnt stimulus-response (S-R) associations can be separated by ERPs. Namely, the P3 component, a positive deflection that typically occurs on parietal channels 300–600 ms after stimulus presentation is sensitive to stimulus probability, habituation, and the time since the last target presentation, which are all implicated in contextual (sequence) learning . In incidental sequence learning, the P3 amplitude is larger for less predictable than for more predictable targets . Moreover, it reflects the retrieval of S-R associations both in incidental learning , – and cognitive conflict tasks – . Specifically, retrieving a less accessible S-R link would attenuate the P3 amplitude, which then reflects the difficulty of response selection and the amount of processing resources needed , . The analyses of both N450 and P3 are useful to distinguish between the processing stages of conflict detection (N450) and decision-making/response selection (P3) and how incidental sequence learning modulates them, respectively. Since both increased conflict and increased task difficulty prolong responses, the two stages would not be differentiable by using behavioural measures alone. Hypotheses In the current study, RT, accuracy, and mean amplitude of the N450 and P3 components will be analysed separately for statistical learning and rule-based learning. If sequential S-C links develop independently from the level of conflict , learning effects would be observed on the P3 but not on the N450, similar to S-R-based sequences , – . However, if S-C sequence learning modulates behaviour only in case of high conflict , , , , learning effects would be expected on the N450 but not on the P3. We have expected that learning and integrating abstract information into S-C representations is more effortful than similar processes based on S-R associations , . Therefore, switching from local binding of S-R features to global binding of S-C episodes is only expected to happen if efficiency in predicting the next event necessitates it , . Considering the nature of simultaneous learning functions and their potential predictive value in cognitive control, we had the following expectations. If task complexity leads to frequent errors, participants can learn both statistical and rule-based regularities of S-C episodes , that is, response times will be shorter when statistical or rule-based regularities have high predictability on the trial’s congruency compared to low predictability trials. As statistical learning develops faster than rule-based learning, we expected the statistical learning effects at the behavioural level already in the beginning (first half) of the task. We expected that P3 amplitude in high predictability conditions will be smaller than in low predictability (i.e., statistical learning effect and rule-based learning effect). We expected that N450 amplitude in high predictability conditions will be larger (more negative) than in low predictability conditions if the cognitive conflict necessitates it (i.e., in incongruent condition). After the analysis of Experiment 1, follow-up behavioural studies were conducted to confirm the replicability and generalisability of the findings of Experiment 1 (see Supplementary Results). Specifically, Experiment 2 was performed as an internal replication effort of Experiment 1. We have expected that learning-related benefit on cognitive control in Experiment 1 can also be observed in an independent group in Experiment 2. Additionally, we have performed Experiment 3 to investigate the generalisability of the findings of Experiments 1 and 2 to other control state constellations. Namely, the task was modified to eliminate the potential effect of switching between dominant response dimensions. Experiment 3 is presented as an exploratory analysis. Finally, Experiment 4 was performed to investigate the generalisability of the findings of Experiments 1 and 2 to another sequence. Sequence learning is not a unitary construct, but involves different processes that detect, encode and retrieve sequentially presented regularities , , – . A prominent distinction is between the acquisition of probabilistic information (statistical learning) and sequential order , , , . This latter process is called higher-order sequence learning , , order-based learning or rule-based learning , . Within sequence learning, statistical learning can be operationalized as the differentiation between high- and low-probability elements that otherwise carry indistinguishable information about the serial order . In contrast, rule-based learning is the differentiation between elements that are defined by the serial order and those that are presented randomly, while the probability of occurrence remains the same , . While statistical learning enables the acquisition of distributions and probabilistic interdependence in the stimulus stream, rule-based learning controls and integrates learning results into higher-order regulations . These learning functions operate in overlapping time windows, however, they have different time courses , : fast estimation of statistics reaches its plateau early in contrast to the more gradual curve of rule-based learning. In other words, a larger rule-based learning effect can be expected at the end of the task than at the beginning. In contrast, statistical learning shows less variability over time. Moreover, the two learning processes have different interdependence on other cognitive functions , , , they are related to different neural sources , and are thought to generate differentiable expectations about upcoming events . Statistical learning is considered to be a more bottom-up process that develops fast in the presence of recurring patterns in the sensory experience , , . In contrast, rule-based learning operates with the gradual accumulation of sequential rules, and the learnt associations are also available for top-down processes , , . Considering the simultaneous nature of statistical and rule-based learning, the current study will investigate the potential contribution of both processes to cognitive control. The two forms of sequence learning may have different capacities to interact with cognitive control. Due to their simultaneity, it is often hard to disentangle the two processes at the behavioural level, however, the analysis of the concurrent neurophysiological signal can provide confirmation of the behavioural effects and further insights into the mechanisms behind them , , , – . The current study investigated concurrent incidental sequence learning and cognitive control by considering neurophysiological information that are differentially related to cognitive control and contextual learning. In a Stroop task, a frontocentral negative deflection is typically observed with a peak of approximately 450 ms after stimulus presentation . This so-called N450 event-related potential (ERP) component’s amplitude is larger in incongruent than in congruent trials, therefore, it is an ideal candidate to study conflict detection – . Therefore, the N450 was selected as a more specific marker of conflict detection in the current study. Another benefit of using the ERP approach is the potential to differentiate between a cascade of processes , . For instance, conflict detection and the retrieval of previously learnt stimulus-response (S-R) associations can be separated by ERPs. Namely, the P3 component, a positive deflection that typically occurs on parietal channels 300–600 ms after stimulus presentation is sensitive to stimulus probability, habituation, and the time since the last target presentation, which are all implicated in contextual (sequence) learning . In incidental sequence learning, the P3 amplitude is larger for less predictable than for more predictable targets . Moreover, it reflects the retrieval of S-R associations both in incidental learning , – and cognitive conflict tasks – . Specifically, retrieving a less accessible S-R link would attenuate the P3 amplitude, which then reflects the difficulty of response selection and the amount of processing resources needed , . The analyses of both N450 and P3 are useful to distinguish between the processing stages of conflict detection (N450) and decision-making/response selection (P3) and how incidental sequence learning modulates them, respectively. Since both increased conflict and increased task difficulty prolong responses, the two stages would not be differentiable by using behavioural measures alone. In the current study, RT, accuracy, and mean amplitude of the N450 and P3 components will be analysed separately for statistical learning and rule-based learning. If sequential S-C links develop independently from the level of conflict , learning effects would be observed on the P3 but not on the N450, similar to S-R-based sequences , – . However, if S-C sequence learning modulates behaviour only in case of high conflict , , , , learning effects would be expected on the N450 but not on the P3. We have expected that learning and integrating abstract information into S-C representations is more effortful than similar processes based on S-R associations , . Therefore, switching from local binding of S-R features to global binding of S-C episodes is only expected to happen if efficiency in predicting the next event necessitates it , . Considering the nature of simultaneous learning functions and their potential predictive value in cognitive control, we had the following expectations. If task complexity leads to frequent errors, participants can learn both statistical and rule-based regularities of S-C episodes , that is, response times will be shorter when statistical or rule-based regularities have high predictability on the trial’s congruency compared to low predictability trials. As statistical learning develops faster than rule-based learning, we expected the statistical learning effects at the behavioural level already in the beginning (first half) of the task. We expected that P3 amplitude in high predictability conditions will be smaller than in low predictability (i.e., statistical learning effect and rule-based learning effect). We expected that N450 amplitude in high predictability conditions will be larger (more negative) than in low predictability conditions if the cognitive conflict necessitates it (i.e., in incongruent condition). After the analysis of Experiment 1, follow-up behavioural studies were conducted to confirm the replicability and generalisability of the findings of Experiment 1 (see Supplementary Results). Specifically, Experiment 2 was performed as an internal replication effort of Experiment 1. We have expected that learning-related benefit on cognitive control in Experiment 1 can also be observed in an independent group in Experiment 2. Additionally, we have performed Experiment 3 to investigate the generalisability of the findings of Experiments 1 and 2 to other control state constellations. Namely, the task was modified to eliminate the potential effect of switching between dominant response dimensions. Experiment 3 is presented as an exploratory analysis. Finally, Experiment 4 was performed to investigate the generalisability of the findings of Experiments 1 and 2 to another sequence. We analysed the interaction between learning of predictabilities and experimental conditions. First, the full factorial design of the experiment was analysed with the within-subject factors of predictability (as triplet types: high-probability pattern, high-probability random, low-probability random), condition (congruent, incongruent, word naming, colour naming), and period (first half and second half). The main effects and interactions are summarised in Tables – . For the sake of brevity, only significant results are described in detail in the Results section. In case of a significant interaction that involved both predictability and condition, follow-up analyses were conducted to quantify how predictability could affect RT, accuracy, or mean amplitude of the N450 and P3 components in different types of conflict situations. This was done according to previous studies with the ASRT paradigm , , , , , to limit the number of pair-wise comparisons. Specifically, we analysed whether the difference between high-probability random and low-probability random trials ( statistical learning ) and the difference between the high-probability pattern and high-probability random trials ( rule-based learning ) was dependent on the condition. All post hoc comparisons are Bonferroni-corrected. In these follow-up ANOVAs, we analysed whether the difference between high-probability random and low-probability random trials (statistical learning) and the difference between the high-probability pattern and high-probability random trials (rule-based learning) was dependent on the condition. Please, note, that this step has the same purpose as conducting post-hoc tests on the omnibus (full factorial) ANOVA. The follow-up ANOVAs have the advantage to provide results that can be directly linked to the hypotheses (e.g., statistical learning in incongruent vs statistical learning in colour naming condition) and omit contrasts that are not planned or meaningful in the design (e.g., all possible differences between the high-probability pattern and low-probability random trials). In the text, average and standard error are provided as descriptive values. Accuracy data inform about task difficulty Please note, that RTs and ERPs were quantified by using correctly responded trials only, therefore a direct relationship between accuracy and ERP effects were not central to the hypotheses , . Nevertheless, we report the accuracy results for the sake of completion and to describe the varying level of difficulty that participants faced in the different conditions. Accuracy rates (percentage of correctly responded trials) were analysed in a three-way repeated-measures ANOVA with predictability, condition, and period as within-subject factors. The main effects and interactions are summarised in Table . The main effects of condition ( F (3, 90) = 48.51, ε = 0.591, p < 0.001, η p 2 = 0.618) and period ( F (1, 30) = 24.66, p < 0.001, η p 2 = 0.451) were significant. The accuracy rate was lower in incongruent (70% ± 0.1) than in colour naming (81% ± 0.1, p < 0.001) or in congruent trials (85% ± 0.1, p < 0.001). Additionally, participants were less accurate in word naming (70% ± 0.1) than in colour naming ( p < 0.001) or congruent trials ( p < 0.001). The accuracy rate was lower in colour naming than in congruent trials ( p < 0.001). The accuracy did not differ significantly between incongruent and word naming conditions ( p = 0.999). Participants became more accurate for the second period (82% ± 0.1) compared to the first one (72% ± 0.1, p < 0.001). The interaction of condition by period was significant ( F (3, 90) = 7.85, p < 0.001, η p 2 = 0.207). In the first period, participants had a lower accuracy rate in incongruent (65% ± 0.1) than in colour naming (76% ± 0.1, p < 0.001) or in congruent trials (82% ± 0.1, p < 0.001). Additionally, participants were less accurate in word naming (64% ± 0.1) than in colour naming ( p < 0.001) or congruent trials ( p < 0.001). Accuracy was lower in colour naming than in congruent trials ( p < 0.001). The accuracy did not differ significantly between incongruent and word naming conditions ( p = 0.999). Similarly, in the second period, participants made more errors in incongruent (75% ± 0.1) than in colour naming (85% ± 0.1, p < 0.001) or in congruent trials (89% ± 0.1, p < 0.001). Additionally, participants were less accurate in word naming (78% ± 0.1) than in colour naming ( p < 0.001) or congruent trials ( p < 0.001). Accuracy was lower in colour naming than in congruent trials ( p < 0.001). The accuracy did not differ significantly between incongruent and word naming conditions ( p = 0.548). The predictability by condition interaction was significant ( F (6, 180) = 4.44, p = 0.008, η p 2 = 0.103). In the word naming condition, participants were less accurate in high-probability random trials (68% ± 0.1) than in high-probability pattern trials (71% ± 0.1, p = 0.028) and in low-probability random trials (73% ± 0.1, p = 0.005). High-probability pattern trials of word naming did not differ significantly from low-probability random word-naming trials ( p = 0.054). Other pair-wise differences in all other conflict conditions did not differ significantly from each other ( p s > 0.164). In the case of statistical learning (Table ), participants showed more learning in word naming (−0.05% ± 0.2) than in colour naming (0.001% ± 0.01, p = 0.035) and more learning in word naming than in incongruent trials (0.16% ± 0.01, p = 0.016). None of the other pair-wise comparisons was significant ( p s > 0.565). In rule-based learning , there were no significant pair-wise differences. RT data reveal an interaction between stimulus conflict and sequential regularities The RT data are shown in Fig. . First, RTs were analysed in a three-way repeated-measures ANOVA with predictability, condition, and period as within-subject factors. The main effects and interactions are summarised in Table . The main effects of condition ( F (3, 90) = 114.90, ε = 0.811, p < 0.001, η p 2 = 0.793) and period ( F (1, 30) = 59.60, p < 0.001, η p 2 = 0.665) were significant. Participants responded slower in word naming (520.7 ms ± 5.6) than in incongruent (501.9 ms ± 6.8, p < 0.001), congruent, (483.8 ms ± 6.3, p < 0.001), or colour naming trials (496.9 ms ± 6.5, p < 0.001). Reaction times were longer in incongruent than in congruent trials ( p < 0.001) which suggests a general congruency (Stroop) effect. Additionally, participants responded slower in colour naming than in congruent trials ( p < 0.001). Colour naming and incongruent trials did not differ significantly from each other ( p = 0.798). Participants became faster for the second period (498.2 ms ± 6.2) compared to the first one (515.2 ms ± 5.9, p < 0.001). Importantly, both the predictability by condition interaction ( F (6, 180) = 4.68, p < 0.001, η p 2 = 0.135), and the three-way interaction between predictability, condition and period were significant ( F (6, 180) = 3.80, p = 0.001, η p 2 = 0.112). The interaction effects were further analysed below. Evidence for enhanced statistical learning but not rule-based learning under high cognitive conflict In the case of statistical learning (Table ), the difference between high-probability random colour naming and low-probability random colour naming trials decreased from the first period (15.9 ms ± 3.4) to the second one (−3.7 ms ± 3.9, p < 0.001). In the first period, high-probability random colour naming trials (514.9 ms ± 7.8) were slower than low-probability random trials (499.0 ms ± 6.5, p < 0.001), which is considered an inverse statistical learning effect. None of the other conditions showed significant changes between the periods ( p s > 0.070). Importantly, statistical learning was larger in incongruent (−12.7 ms ± 6.1) than in colour naming condition (15.9 ms ± 3.4, p = 0.002) in the first period. There were no other significant pair-wise differences between conditions in the two periods ( p s > 0.165). Thus, statistical learning was larger in incongruent (high conflict) than in colour naming (neutral) trials at the beginning of the task, however, this difference was not significant in the second period. In the case of rule-based learning , there were no significant pairwise comparisons ( p s > 0.074). Thus, the analyses showed enhanced statistical learning when the demand for cognitive conflict was high (i.e., in the incongruent condition). In sum, RT analyses showed that conflict levels affected participants’ behaviour: compared to the neutral condition, the overlap between stimulus dimensions facilitated responses (congruency effect), while a switch between the dominant response dimension (from perceptual to semantic) led to slower responses. Neither statistical nor rule-based regularities showed a significant effect on RTs, however, the presented regularities interacted with the Stroop conditions. At the beginning of the task, statistical learning was larger in the incongruent condition. In the second half of the task, statistical learning was not modulated significantly by conditions (see Table ). Additionally, follow-up experiments were conducted to test the replicability and generalisability of the behavioural results. The results of Experiments 2–4 can be found in the Supplementary Materials (Supplementary Results and Tables – ). Neurophysiological data: N450 amplitude was modulated by the interaction between stimulus conflict and sequential regularities Mean amplitude on channel P1 (see Supplementary Fig. ) in the time window of 280–380 ms was analysed in a two-way repeated-measures ANOVA with predictability and condition as within-subject factors. The main effects and interactions are summarised in Table . The main effect of condition was significant ( F (3, 90) = 10.15, ε = 0.884, p < 0.001, η p 2 = 0.253). The P3 amplitude was smaller in word naming (7.00 µV/m 2 ± 2.03) than in congruent (10.60 µV/m 2 ± 2.30, p < 0.001) and incongruent (9.40 µV/m 2 ± 2.22, p = 0.019) trials. The other pair-wise differences were not significant ( p s > 0.073). That is, the P3 amplitude was lower in the condition in which a semantic decision was required (word naming) as opposed to the other two conditions with a perceptual (colour) decision. Next, the mean amplitude of the N450 in the time window of 380–460 ms was analysed in a two-way repeated-measures ANOVA with predictability and condition as within-subject factors. Grand averages of N450 waveforms on the channel FCz are presented in Fig. . The main effects and interactions are summarised in Table . The predictability by condition ( F (6, 180) = 2.47, ε = 0.876, p = 0.032, η p 2 = 0.076) interaction was significant. The interaction effect was further analysed below. Statistical and rule-based learning effects on the N450 The main effects and interactions are summarised in Table . The statistical learning effect on the N450 was larger in incongruent (−2.94 µV/m 2 ± 1.26) than in colour naming (1.44 µV/m 2 ± 1.09, p = 0.046) condition. The other pair-wise differences ( p s > 0.281) were not significant. In case of rule-based learning , the pair-wise differences were not significant ( p s > 0.066). That is, the statistical learning effect in the N450 was larger in incongruent than in the colour naming (neutral) condition (cf. RT results). In sum, ERP analyses showed that conditions were differentiated from each other in different processing stages: switching between the perceptual and semantic response dimensions presented a significant effect on the P3 but not on the N450, while stimulus conflict (condition) was a significant effect on the N450 but not on the P3. Crucially, the N450 was modulated by the interaction between predictability and condition. The follow-up analysis showed larger statistical learning in incongruent trials Thus, the ERP analyses confirmed the behavioural results of interaction between predictability and condition. The interaction effect was significant only in N450 but not in P3, which suggests that the interaction is specific to the type of regularity (statistical learning), conflict condition (incongruent), and processing stage (N450). Next, the onset latency of the N450 in the time window of 380–460 ms was analysed in a two-way repeated-measures ANOVA with predictability and condition as within-subject factors. No significant effects were obtained (Table ). Please note, that RTs and ERPs were quantified by using correctly responded trials only, therefore a direct relationship between accuracy and ERP effects were not central to the hypotheses , . Nevertheless, we report the accuracy results for the sake of completion and to describe the varying level of difficulty that participants faced in the different conditions. Accuracy rates (percentage of correctly responded trials) were analysed in a three-way repeated-measures ANOVA with predictability, condition, and period as within-subject factors. The main effects and interactions are summarised in Table . The main effects of condition ( F (3, 90) = 48.51, ε = 0.591, p < 0.001, η p 2 = 0.618) and period ( F (1, 30) = 24.66, p < 0.001, η p 2 = 0.451) were significant. The accuracy rate was lower in incongruent (70% ± 0.1) than in colour naming (81% ± 0.1, p < 0.001) or in congruent trials (85% ± 0.1, p < 0.001). Additionally, participants were less accurate in word naming (70% ± 0.1) than in colour naming ( p < 0.001) or congruent trials ( p < 0.001). The accuracy rate was lower in colour naming than in congruent trials ( p < 0.001). The accuracy did not differ significantly between incongruent and word naming conditions ( p = 0.999). Participants became more accurate for the second period (82% ± 0.1) compared to the first one (72% ± 0.1, p < 0.001). The interaction of condition by period was significant ( F (3, 90) = 7.85, p < 0.001, η p 2 = 0.207). In the first period, participants had a lower accuracy rate in incongruent (65% ± 0.1) than in colour naming (76% ± 0.1, p < 0.001) or in congruent trials (82% ± 0.1, p < 0.001). Additionally, participants were less accurate in word naming (64% ± 0.1) than in colour naming ( p < 0.001) or congruent trials ( p < 0.001). Accuracy was lower in colour naming than in congruent trials ( p < 0.001). The accuracy did not differ significantly between incongruent and word naming conditions ( p = 0.999). Similarly, in the second period, participants made more errors in incongruent (75% ± 0.1) than in colour naming (85% ± 0.1, p < 0.001) or in congruent trials (89% ± 0.1, p < 0.001). Additionally, participants were less accurate in word naming (78% ± 0.1) than in colour naming ( p < 0.001) or congruent trials ( p < 0.001). Accuracy was lower in colour naming than in congruent trials ( p < 0.001). The accuracy did not differ significantly between incongruent and word naming conditions ( p = 0.548). The predictability by condition interaction was significant ( F (6, 180) = 4.44, p = 0.008, η p 2 = 0.103). In the word naming condition, participants were less accurate in high-probability random trials (68% ± 0.1) than in high-probability pattern trials (71% ± 0.1, p = 0.028) and in low-probability random trials (73% ± 0.1, p = 0.005). High-probability pattern trials of word naming did not differ significantly from low-probability random word-naming trials ( p = 0.054). Other pair-wise differences in all other conflict conditions did not differ significantly from each other ( p s > 0.164). In the case of statistical learning (Table ), participants showed more learning in word naming (−0.05% ± 0.2) than in colour naming (0.001% ± 0.01, p = 0.035) and more learning in word naming than in incongruent trials (0.16% ± 0.01, p = 0.016). None of the other pair-wise comparisons was significant ( p s > 0.565). In rule-based learning , there were no significant pair-wise differences. The RT data are shown in Fig. . First, RTs were analysed in a three-way repeated-measures ANOVA with predictability, condition, and period as within-subject factors. The main effects and interactions are summarised in Table . The main effects of condition ( F (3, 90) = 114.90, ε = 0.811, p < 0.001, η p 2 = 0.793) and period ( F (1, 30) = 59.60, p < 0.001, η p 2 = 0.665) were significant. Participants responded slower in word naming (520.7 ms ± 5.6) than in incongruent (501.9 ms ± 6.8, p < 0.001), congruent, (483.8 ms ± 6.3, p < 0.001), or colour naming trials (496.9 ms ± 6.5, p < 0.001). Reaction times were longer in incongruent than in congruent trials ( p < 0.001) which suggests a general congruency (Stroop) effect. Additionally, participants responded slower in colour naming than in congruent trials ( p < 0.001). Colour naming and incongruent trials did not differ significantly from each other ( p = 0.798). Participants became faster for the second period (498.2 ms ± 6.2) compared to the first one (515.2 ms ± 5.9, p < 0.001). Importantly, both the predictability by condition interaction ( F (6, 180) = 4.68, p < 0.001, η p 2 = 0.135), and the three-way interaction between predictability, condition and period were significant ( F (6, 180) = 3.80, p = 0.001, η p 2 = 0.112). The interaction effects were further analysed below. In the case of statistical learning (Table ), the difference between high-probability random colour naming and low-probability random colour naming trials decreased from the first period (15.9 ms ± 3.4) to the second one (−3.7 ms ± 3.9, p < 0.001). In the first period, high-probability random colour naming trials (514.9 ms ± 7.8) were slower than low-probability random trials (499.0 ms ± 6.5, p < 0.001), which is considered an inverse statistical learning effect. None of the other conditions showed significant changes between the periods ( p s > 0.070). Importantly, statistical learning was larger in incongruent (−12.7 ms ± 6.1) than in colour naming condition (15.9 ms ± 3.4, p = 0.002) in the first period. There were no other significant pair-wise differences between conditions in the two periods ( p s > 0.165). Thus, statistical learning was larger in incongruent (high conflict) than in colour naming (neutral) trials at the beginning of the task, however, this difference was not significant in the second period. In the case of rule-based learning , there were no significant pairwise comparisons ( p s > 0.074). Thus, the analyses showed enhanced statistical learning when the demand for cognitive conflict was high (i.e., in the incongruent condition). In sum, RT analyses showed that conflict levels affected participants’ behaviour: compared to the neutral condition, the overlap between stimulus dimensions facilitated responses (congruency effect), while a switch between the dominant response dimension (from perceptual to semantic) led to slower responses. Neither statistical nor rule-based regularities showed a significant effect on RTs, however, the presented regularities interacted with the Stroop conditions. At the beginning of the task, statistical learning was larger in the incongruent condition. In the second half of the task, statistical learning was not modulated significantly by conditions (see Table ). Additionally, follow-up experiments were conducted to test the replicability and generalisability of the behavioural results. The results of Experiments 2–4 can be found in the Supplementary Materials (Supplementary Results and Tables – ). Mean amplitude on channel P1 (see Supplementary Fig. ) in the time window of 280–380 ms was analysed in a two-way repeated-measures ANOVA with predictability and condition as within-subject factors. The main effects and interactions are summarised in Table . The main effect of condition was significant ( F (3, 90) = 10.15, ε = 0.884, p < 0.001, η p 2 = 0.253). The P3 amplitude was smaller in word naming (7.00 µV/m 2 ± 2.03) than in congruent (10.60 µV/m 2 ± 2.30, p < 0.001) and incongruent (9.40 µV/m 2 ± 2.22, p = 0.019) trials. The other pair-wise differences were not significant ( p s > 0.073). That is, the P3 amplitude was lower in the condition in which a semantic decision was required (word naming) as opposed to the other two conditions with a perceptual (colour) decision. Next, the mean amplitude of the N450 in the time window of 380–460 ms was analysed in a two-way repeated-measures ANOVA with predictability and condition as within-subject factors. Grand averages of N450 waveforms on the channel FCz are presented in Fig. . The main effects and interactions are summarised in Table . The predictability by condition ( F (6, 180) = 2.47, ε = 0.876, p = 0.032, η p 2 = 0.076) interaction was significant. The interaction effect was further analysed below. The main effects and interactions are summarised in Table . The statistical learning effect on the N450 was larger in incongruent (−2.94 µV/m 2 ± 1.26) than in colour naming (1.44 µV/m 2 ± 1.09, p = 0.046) condition. The other pair-wise differences ( p s > 0.281) were not significant. In case of rule-based learning , the pair-wise differences were not significant ( p s > 0.066). That is, the statistical learning effect in the N450 was larger in incongruent than in the colour naming (neutral) condition (cf. RT results). In sum, ERP analyses showed that conditions were differentiated from each other in different processing stages: switching between the perceptual and semantic response dimensions presented a significant effect on the P3 but not on the N450, while stimulus conflict (condition) was a significant effect on the N450 but not on the P3. Crucially, the N450 was modulated by the interaction between predictability and condition. The follow-up analysis showed larger statistical learning in incongruent trials Thus, the ERP analyses confirmed the behavioural results of interaction between predictability and condition. The interaction effect was significant only in N450 but not in P3, which suggests that the interaction is specific to the type of regularity (statistical learning), conflict condition (incongruent), and processing stage (N450). Next, the onset latency of the N450 in the time window of 380–460 ms was analysed in a two-way repeated-measures ANOVA with predictability and condition as within-subject factors. No significant effects were obtained (Table ). We have investigated how incidental learning modulates cognitive control at the levels of behavioural adaptation and related neurophysiological signatures. It was suggested that regularities in the stimulus stream modulate cognitive control through binding between stimulus and control states – , , , . This is the first study to provide evidence for this account by testing specific stages of cascaded processes (i.e., the components of N450 and P3) coded in the neurophysiological signal. We suggest that it is crucial to consider the type and functionality of the learnt regularities in the interaction between incidental learning and cognitive control. According to follow-up experiments (Supplementary Results), the interaction between learning and control functions may be smaller or absent under certain conditions, which supports the multifactorial nature of this relationship. Participants performed a Stroop task in which the task-relevant colour information was presented alone (colour naming), in line with the semantic information (congruent), or in conflict between them (incongruent). Response speed for the semantic information alone was also assessed (word naming). Compared to neutral colour naming, responses were slower if perceptual and semantic stimulus dimensions were in conflict (incongruent) or if the response dimension needed to be shifted from colour to semantic information (word naming). In contrast, an alignment between the two dimensions (congruent) facilitated the responses. Slower reactions in incongruent than in congruent trials suggest that a general congruency effect (i.e., the Stroop effect) was observable in the experiment , , . Unexpectedly, incongruent trials were not significantly slower than colour naming trials, which could be explained by these conditions’ involvement in the interaction effect (see below). Unbeknown to the participants, trials were presented according to two kinds of regularities. The first one determined an alternating position in the stimulus sequence (rule-based learning), while the second one predicted the probability of non-adjacent transitions (statistical learning, see also Fig. ). Importantly, when the demand for cognitive control was the largest, participants’ responses differed between low-probability and high-probability stimulus continuations. This interaction between incidental sequence learning and cognitive conflict was specific to statistical learning and for the first half of the experiment. This specificity might suggest that there is no universal information exchange throughout the task between sequence learning and cognitive control. Namely, statistical learning but not rule-based learning significantly modulated responses in incongruent trials. This difference between the two forms of sequence learning is surprising, considering that statistical learning is a more bottom-up, automatic process compared to rule-based learning . According to our expectations, rule-based learning was a more likely candidate to interact with (top-down) cognitive control. Two aspects of how statistical learning was implemented in the current study might explain the results. It was suggested that the complexity of the probabilistic relations could involve additional processes beyond statistical learning. Specifically, non-adjacent dependencies require the partial inhibition of the intervening item . As statistical learning scores in the current study were based on non-adjacent transitions, an interaction between statistical learning and cognitive control might be plausible. Additionally, statistical learning typically evolves faster than rule-based learning in probabilistic sequences , . It is possible, that longer exposure to the sequence is needed to induce interaction between rule-based learning and cognitive control. Notably, statistical learning modulated incongruent responses in the first half of the task only. As the response cost of incongruent trials also decreased from the first to the second half of the experiment, it is possible that the functionality to predict incongruent trials also attenuated. The interplay between control states and regularities has been further analysed at the level of neurophysiology. The P3 and the N450 components were sensitive to different aspects of task demands, which confirms their interpretation as different stages in the chain of processing the control-related information , , . The P3 amplitude was smaller in word naming than in congruent or incongruent conditions. Word naming was the only condition in which the predominant response dimension was semantic instead of perceptual. As this condition was also characterised by slower responses and a lower accuracy rate than in the other trials, we suggest that P3 amplitude modulations reflected the increased effort to retrieve the less frequently used S-R associations – . However, the learning and retrieval of S-C associations did not modulate the P3, unlike the S-R associations in sequential regularities , – . In previous sequence learning studies, the P3 amplitude differed between more predictable and less predictable trials , , – . The lack of significant effect of predictability either alone or in interaction with the conflict conditions signals caution on interpreting learning effects in the current study. At the same time, it is feasible that S-C learning does not involve the same ERP modulations as S-R learning. In sum, P3 amplitude changes did not reflect learning effects, however, the component’s amplitude decreased as an indication of the increased difficulty of response selection , , . In contrast, changes in the N450 amplitude followed multiple task dimensions. The sensitivity to the cognitive conflict was confirmed by an increased amplitude in incongruent trials. The incongruent effect on the N450 was larger and conflict detection more pronounced when events could be predicted by statistical learning. Next, we discuss these findings in regard to the interplay between sequence learning and cognitive control. Some accounts promoted that incidental sequence learning and cognitive control compete for the same neural resources, and therefore, neither function can benefit from the other , – . At first blush, both the behavioural and neurophysiological results support competition between learning and conflict processes. Participants showed engagement of cognitive conflict as evidenced by the Stroop effect. However, neither the predictability main effect nor the predictability by period interaction was significant. Traditionally, these effects were taken as markers of learning S-R contingencies in an alternating sequence learning task , . In the current study, sequential regularities predicted the upcoming control state but not the colour of the stimuli. Therefore, participants could not anticipate the response itself, however, learning could have prepared them for the type of response selection and the associated demand for top-down engagement. That is, predictable trials presented a cognitive advantage in terms of response preparation for all control states. However, this advantage did not manifest as an effect of predictability. Thus, one might conclude that in presence of an intentional (overt) cognitive control task, parallel processing of incidental (covert) sequence learning is not possible. Nonetheless, two aspects should be considered before this conclusion can be reached: (i) the circumstances under direct evidence of competition that were shown in previous studies, and (ii) the type of representations that should be detected in a learning effect. Originally, a competition was suggested between memory systems: automatic processes mediated by the striatum and voluntary, attention-dependent processes mediated by the prefrontal and medial temporal lobes , , , . Later, this competition theory was extended to a resource conflict between cognitive control and procedural sequence learning , , , , . Notably, sequence learning improved when control-related prefrontal functions were attenuated by hypnosis or brain stimulation , , . In contrast, the current study investigated the interplay between sequence learning and conflict without an attempt to attenuate either learning or control functions. Moreover, competition in previous research was demonstrated by measuring the two functions separately , , , , . It is conceivable that interaction could only be detected when sequence learning and cognitive control are measured in the same task or situation , . This difference directly leads us to the second important aspect: what is being learnt in a sequence. In the current study, the sequence could be used to predict the next control state. Since the importance of an S-C event file is in recognising the need for cognitive control , it is not necessarily expected to be shown as a predictability effect or predictability by period interaction. Rather, the type of predictability, the Stroop condition, and the task period should determine together whether incidental sequence learning can benefit from the control state and vice versa. Thus, the obtained three-way interaction (also in the replication study, see Supplementary Results) support our hypothesis that participants can learn both statistical and rule-based regularities of S-C episodes. However, the P3 component did not show learning effects, as we expected based on previous studies , – . This difference might indicate that learning of abstract regularities involves other neurophysiological mechanisms than learning of physical properties, hence, the P3 is not sensitive to both types of memories. Alternatively, since S-C learning is not expected to be expressed without the presence of conflict, and the P3 was not sensitive to stimulus conflict, this component also cannot be sensitive to the different types of regularities. In this scenario, contrary to our original expectations, the conflict-related N450 should be considered sensitive to learning S-C associations on the neurophysiological level. In contrast to the P3, the N450 showed an interaction between sequence learning and cognitive conflict. Interactions between learning and cognitive conflict occurred both in the behavioural and neurophysiological data. Thus, the current results are in line with the notion that by binding goal representations (C) and contextual information (S) together, S-C associations modulate behaviour , , , . It has been proposed, that S-C associations have a role in control functions only when less demanding processes (such as response priming, S-R predictions, etc.) are not sufficient to direct response selection , . Therefore, the current study investigated this prediction under challenging conditions. Namely, the Stroop task had four conditions and four related control states, therefore, predicting the correct state might have reduced the cost of adaptation more than in binary-outcome (congruent vs. incongruent) versions of the task. Indeed, when S-C associations were highly predictive (80% probability) but participants only needed to switch between two control states, error rates were minimal . Moreover, half of the trials were incongruent which might have prevented the participants from effectively altering their response preparations. If the possibility of a conflict is always high, there is little benefit in ever letting the guard down. Consequently, S-C learning did not occur . Of note, the current study did not aim to test the design of Jiménez et al. with four instead of two conflict conditions. Nevertheless, the difference in task difficulty and error rate should be mentioned given its potential significance in cooperative interaction between learning and cognitive control , . Along these lines, in Experiments 1 and 2, learning and control functions showed interactive effect only when a high level of control was necessary (incongruent condition) and therefore, response selection was challenging. This pattern did not only confirm our hypothesis but is also similar to previous studies showing that the incongruent response cost is smaller in predictable than in unpredictable trials and sequence learning is enhanced when S-R mapping is incompatible in contrast to compatible configurations . Importantly, this similarity between how S-R and S-C predictions contribute to the congruency effect supports the idea that lower and global level of binding represents compatible representation systems that respond to experimental effects in similar ways . Importantly, the obtained interaction in the behavioural and neurophysiological data between statistical learning and cognitive control highlights the potential complexity of integrating S-C associations. Previously, predictability was quantified in Stroop tasks as a first-order transition between two consecutive trials , . The current study investigated the role of non-adjacent, second-order transitional probabilities. Control states were predictable as a condition of the previous two events (high vs low-probability triplets). Triplet probabilities were quantified as moving dependencies throughout the stimuli. That is, a sequence item could always be categorised as the first part of the triplet (predictor item), as a second part (interleaving item), or as a final part (predicted item). Interestingly, learning of non-adjacent transitions was effective in enhancing conflict detection for incongruent trials (i.e., larger N450) and consequently, reducing response cost. That is, the detection and acquisition of second-order transitions seemed to allow the interplay with prefrontal functions, which further promotes the view that this type of dependency calls for inhibitory activity . The current results fit into a larger picture of statistical information modulating attention, inhibitory control, or linguistic processes. For instance, learning the distributional statistics of conflict levels modulate the allocation of attentional resources during distractor suppression , – . Similarly, the distribution of different task conditions can be used to prepare for a potential change in task switching . Mediating between inhibition and sequence learning is also important in processing complex syllable strings or syntax , with non-adjacent interrelations, or other aspects of nested predictabilities in language . In sum, learning probability information contributes to various forms of top-down engagement. The current study provides expands this list by showing an interaction between statistical learning and conflict monitoring at the neurophysiological level. The follow-up studies outlined in the Supplementary Materials provided important details on the generalisability of the results. While the main finding was replicated in Experiment 2, different control conditions in Experiment 3 or a different sequence in Experiment 4 did not yield significant interaction between predictability and cognitive control. In general, the differences (and similarities) across the experiments might be related to the relatively small interaction effect. Notably, a large effect was not expected given that participants were unaware of the learning task; the sequence followed a complex alternating structure; and the sequence consisted of abstract items (congruency) instead of easily learnable physical features (e.g., colours). Further studies are warranted to dovetail the nature of the interaction between statistical learning and conflict monitoring, including the aspects that might enlarge the interaction. One such aspect is the length of exposure to the sequential regularities. It is possible that the amount of time needed to detect rule-based learning effects in S-R sequences is not enough when more abstract features need to be integrated. It is worth noting that even when S-R contingencies are to be learnt, better detection of rule-based learning is often boosted by cues on the serial order . However, to keep the covert nature of the sequence, we did not introduce differences between the presentation of pattern and random events. Notably, explicit pre-cues on trial congruency modulate cognitive control in the upcoming trial , which suggests that control processes can occur before the stimulus onset. Nevertheless, longer training and cued sequences in future studies would provide an important contribution to the behavioural dissociation between statistical and rule-based control state predictions. Another aspect is the potential process-specific relationship between cognitive control and incidental sequence learning. In Experiments 1 and 2, the interaction occurred between cognitive control and statistical learning but not with rule-based learning. In Experiment 3, no interaction was observed, however, significant statistical learning occurred without a rule-based learning effect. Crucially, cognitive control is also non-unitary . The significant interactions in Experiments 1 and 2 occurred in a configuration that included not only the manipulation of stimulus conflict but also switching between the frequent perceptual and the rare semantic response dimension. In contrast, no interaction occurred in Experiment 3, when the fourth condition also required perceptual decision-making. The differences between Experiments 1 and 2 and Experiment 3 may suggest that the nature of stimulus conflict and response inhibition determines if and how statistical learning interacts with cognitive control. However, switching the word naming condition to object colour naming did not only change the conflict configuration of the task, it also decreased the overall complexity. In Experiment 3 participants were the fastest and the most accurate compared to Experiments 1 and 2. As a challenging environment is thought to induce an interaction between sequence learning and cognitive control , , it is unclear whether Experiment 3 can be interpreted as a process-specific or an effort-related difference to Experiments 1 and 2. Another noteworthy difference occurred between Experiments 1-2 and Experiment 4, which tested the generalisability of the original findings to another sequence and revealed no significant interaction between predictability and condition. Importantly, the difference between Experiments 1-2 and Experiment 4 was not limited to the significance of the interaction but was also present in the main effects. Specifically, the predictability main effect was significant in Experiment 4 unlike in Experiments 1-2. Post-hoc tests showed significant statistical learning but not rule-based learning effect. The condition was also significant. However, the pair-wise comparisons did not show response cost for incongruent trials. Thus, while there was no significant interaction effect in the last experiment, the significant main effects also did not appear to be typical: rule-based learning without statistical learning and a congruent response benefit without an incongruent response cost. Also considering the smaller sample size in Experiment 4, this leaves the question of sequence-specificity inconclusive. In sum, main finding was that predictability of a sequence influenced cognitive control, but the strength of the effect varied across different experiments. In Experiment 2, the main finding was replicated, suggesting that it was reliable. However, in Experiment 3 and Experiment 4, the interaction between predictability and cognitive control was not significant, indicating that the relationship may be weaker or absent under certain conditions. Differences (and similarities) across the four experiments might suggest that the duration of the task, the complexity and type of the conflict conditions, and tentatively the sequence-specific interrelations can modulate the interaction between incidental sequence learning and cognitive control. The experiments presented in the current study fit into the larger picture of how multifactorial can be a real-life adaptation to an ever-changing environment , . We tested possible interrelations between incidental sequence learning and cognitive control. To achieve this, we designed a task in which control demands could be predicted by sequential regularities. Participants’ responses adapted to the conflict between perceptual and semantic information; however, they did not express control-independent learning of the S-C regularities. Statistical learning of S-C associations was enhanced in high stimulus conflict at the levels of response times and the amplitude of the conflict monitoring-related neurophysiological signal. Thus, the nature of conflict, the type of sequence learning, and the stage of information processing determine together whether cognitive conflict and sequence learning support or compete with each other. Specifically, statistical learning has the potential to modulate conflict monitoring. Follow-up experiments demonstrated the limits of the interaction’s generalisability suggested that the relationship between cognitive control and statistical learning may be process-specific and influenced by the nature of stimulus conflict and response inhibition. The experiments fit into a larger framework of how people adapt to an ever-changing environment and advocate that real-life adaptations are multifactorial. We suggest that connecting the fields of cognitive control and incidental learning is essential to achieve a synergistic view of adaptive behaviour. Participants N = 33 young adults participated in Experiment 1, who were recruited from the voluntary pool for behavioural studies at the TU Dresden (26.6 years ± 6.6, 13 female, 18 male). Since the hypotheses concerned the detection of the learning effect, this parameter was decisive in sample size. Previous reliability analysis showed that stable learning effects can be detected from N > 21 in the case of alternating sequence presentation . However, since the paradigm was combined with a Stroop task in a novel way, we aimed to increase this number, and have at least 30 participants for the final analysis. We estimated a 10% loss in an EEG experiment, therefore, 33 participants were recruited. Due to incomplete testing in one case and low data quality in another case (see EEG recording and analysis ), N = 31 participants’ data were analysed. Experiment 2 was conducted to internally replicate the behavioural results of Experiment 1 (see details in Supplementary Results). N = 30 participants were recruited, from which N = 28 participants’ data were analysed (due to incomplete testing in two cases). Experiment 3 (Supplementary Results) was conducted with N = 20 newly recruited participants. Experiment 4 was conducted with N = 20 newly recruited participants to investigate the generalisability of the findings of Experiments 1 and 2 to other sequential regularities. There is no overlap between the groups of Experiments 1–4. All participants in Experiments 1–4 had a normal or corrected-to-normal vision, including colour discrimination. None of the participants reported taking centrally acting medication or having a history of neurological or psychiatric conditions. Participants were either native in German or had a similar level of proficiency (based on self-reports). Written informed consent was provided prior to enrolment, and participation was rewarded with 10€. The study was approved by the local ethical review committee and was conducted in accordance with the Declaration of Helsinki. Stimuli, task, and procedure Participants completed a paradigm that combined an overt (Colour-Word Stroop-Task) , , and a covert task (Alternating Serial Reaction Time, ASRT) , . The task is shown in Fig. . Participants saw either colour words in German (ROT: red, BLAU: blue, GELB: yellow, GRÜN: green) or coloured asterisks on the centre of the display. Participants were asked to press a button on the keyboard corresponding to the colour of the word or asterisks, irrespective of the meaning of the word. However, if the word was presented in black, they were asked to press the button that corresponds to the meaning of the word. In 25% of the trials, the word was written in the corresponding colour (i.e., ROT in red, congruent condition ). In 25% of the trials, the word was written in a mismatching colour (i.e., ROT in blue, incongruent condition ). In 25% of the trials, the word was presented in black ( word naming condition ), and in the remaining trials only coloured asterisks were presented ( colour naming condition ). These trial types are henceforth referred to as conditions. In each trial, one of the four possible colours was randomly selected. The covert task followed the structure and timing (including the response window) of a previous version of the ASRT task that had been optimised for EEG recordings . Unbeknown to the participants, stimuli were presented according to an eight-element sequence. Within the sequence, pattern (P) and random (r) items alternated with each other. That is, a sequence of 1-r-2-r-3-r-4-r determined the order of the trial types [1 = congruent, 2 = incongruent, 3 = word naming, 4 = colour naming], interleaved by randomly selected trial types. One of the possible permutations of these of the stimulus sequence was selected for the study and presented to the participants in a pseudo-random manner , , . Importantly, the alternating structure used in the current study leads to probability differences between chunks of three successive trials (triplets). In the case of the above-mentioned example sequence, a triplet starting with 2 and ending with 1 is a high-probability triplet that either occurred as a result of a P-r-P or an r-P-r structure. In contrast, a triplet starting with 1 and ending with 2 is a low-probability triplet that could only have an r-P-r structure (see Fig. ). The distinction between low-probability and high-probability triplets does not only describe the distributional but also the second-order transitional probabilities in the task. That is, a final item of a high-probability triplet is a highly predictable continuation of the first item, while a low-probability triplet’s first item does not carry such anticipatory information. For instance, a triplet starting with 1 has a 62.5% probability of ending with 2, while a triplet starting with 2 has only a 12.5% probability of ending with 1. As a consequence of the continuous stimulus presentation and the unmarked triplet structure, each trial can be categorised as the third item of a high-probability or a low-probability triplet. Furthermore, triplets were organised as moving chunks across the stimuli: the third item of a triplet was also a second item of the next triplet and the first item of the subsequent one , . The task consisted of 16 high-probability triplets, which individually occur five times more often than the 48 low-probability triplets. Moreover, the combination of triplet probability (high-probability versus low-probability) and position in the sequence presentation (pattern versus random) leads to three trial types of sequence regularities: high-probability pattern (50% of all trials), high-probability random (12.5% of all trials), and low-probability random triplets (37.5% of all trials). The trial types of cognitive conflict were equally distributed among the trial types of sequence regularity, that is, 25% of high-probability triplets were congruent, 25% incongruent, 25% word naming, 25% colour naming, etc. The timing of the task followed previous EEG studies of the ASRT (Kóbor et al., 2018; 2019) , . Participants saw either a colour name or a string of coloured asterisks on the centre of the screen for 200 ms. It was followed by a blank screen for 500 ms or until the participant pressed a response button. If the response was incorrect, a blank screen was presented for 500 ms after the response onset, followed by an “X” on the centre of the display for another 500 ms. If the participant did not respond in the trial, the 500 ms-long blank screen was followed by an “!” in the centre of the screen for 500 ms. After a correct response or the incorrect/missed response feedback, a 700-ms-long RSI was introduced. The task presentation was organised into blocks, each of them containing 85 trials. Each block started with 5 warm-up trials that were excluded from the analyses, and 10 repetitions of the eight-element sequence. After completing a block, participants received feedback about their mean RT and accuracy in the block, which was presented for 4000 ms. Between blocks, participants could take a short break. The task consisted of 20 blocks in total. During the behavioural analysis, the first 10 blocks were collapsed as first half of the task, and the remaining 10 blocks were collapsed as the second half of the task. The experiment lasted about 1–1.5 h, including dimming and removing the EEG cap. Task presentation was written and controlled by Presentation software (Neurobehavioral Systems). Stimuli were displayed on an LCD screen with a viewing distance of 80 cm. EEG recording and analysis The EEG was recorded from 60 Ag/AgCl electrodes mounted in an equidistant way on an elastic cap (EasyCap, Germany). The ground and reference (Fpz) electrodes were placed at coordinates θ = 58, φ = 78 and θ = 90, φ = 90, respectively. A BrainAmp amplifier and the Brain Vision Recorder 1.2 software (Brain Products, Germany) were used with a sampling rate of 500 Hz. Offline, data were down-sampled offline to 256 Hz. Impedances were kept below 5 kΩ. The recorded EEG signal was pre-processed using “automagic” and EEGLAB in Matlab 2019a (The MathWorks Corp.). First, flat channels were removed and the EEG data were re-referenced to an average reference. Next, the “PREP pipeline” was applied to remove line noise at 50 Hz by using a multitaper algorithm. PREP also removed contaminations by bad channels and consequently created a robust average reference. Next, EEGLAB’s “clean_rawdata” was applied. This included detrending the EEG data with an IIR high-pass filter of 0.5 Hz (slope 80 dB). Flat-line, noisy, and outlier channels were detected and removed by “clean_rawdata”. Time periods with abnormally strong power (i.e., larger than 15 SD relative to calibration data) were reconstructed by using Artefact Subspace Reconstruction (ASR; burst criterion: 15) . Time windows that could not be reconstructed were removed. A low-pass filter of 40 Hz (sinc FIR filter; order of 86) was applied. EOG artefacts were removed using a subtraction method . Muscle, heart, and remaining eye artefacts were automatically classified and removed by using an independent component analysis (ICA) with ICLabel . Removed and missing channels were interpolated (average of 0.82 channels) according to the spherical computation. As a final step of pre-processing, the data were inspected visually and epochs with bad data were removed. This step was only necessary if non-repeating noises (e.g., due to movement) remained in the data. Segmentation of the continuous EEG data was performed in two consecutive steps. First, segments were created separately for the first and second half of the task, each consisting of 10 blocks. The length of these acquisition periods corresponds to previous EEG analyses that used the ASRT task and showed reliable learning effects , , . Subsequently, segments were averaged for all combinations of the sequence regularity (predictability in the ASRT) and Stroop conditions, namely (average and standard deviation of the numbers of trials reported): high-probability pattern congruent (164 ± 25), high-probability pattern incongruent (123 ± 33), high-probability pattern word naming (138 ± 29), high-probability pattern colour naming (156 ± 27), high-probability random congruent (32 ± 6), high-probability random incongruent (33 ± 9), high-probability random word naming (30 ± 8), high-probability random colour naming (43 ± 7), low-probability random congruent (82 ± 13), low-probability random incongruent (66 ± 17), low-probability random word naming (73 ± 14), low-probability random colour naming (73 ± 13). These combinations were named event types . Group-level ERP waveforms were calculated separately for each event type in the acquisition phases. These waveforms were visually inspected to determine if typical ERP correlates of a Stroop task emerged and to determine the latency range that might vary as a function of event types. This was necessary since the combination of an ASRT and a Stroop task was unprecedented, and therefore, the characteristics of the ERP correlates were unknown. Based on the known characteristics of the components, parietal channels were considered for the P3 , and frontocentral channels for the N450 , , . The P3 was quantified as mean amplitude between 280 and 380 ms after stimulus onset on channel P1. The N450 was quantified as mean amplitude between 380–480 ms after stimulus presentation on the channel FCz. Statistics and reproducibility Statistics were performed in JASP version 0.16.2 (JASP Team). We analysed the interaction between learning (predictability) and cognitive control (condition) by performing repeated measures ANOVAs. First, the full factorial design of the experiment was analysed with the within-subject factors of predictability (as triplet types: high-probability pattern, high-probability random, low-probability random), condition (congruent, incongruent, word naming, colour naming), and period (first half and second half). If a significant interaction occurred that included the factors of predictability and condition, follow-up analyses were conducted to quantify how predictability could affect RT, accuracy, or mean amplitude of the N450 and P3 components in different types of conflict situations. These post-hoc analyses were done in accordance with previous studies that used the ASRT , , , , . Specifically, we analysed whether the difference between high-probability random and low-probability random trials ( statistical learning ) and the difference between the high-probability pattern and high-probability random trials ( rule-based learning ) was dependent on the condition. If the omnibus ANOVA showed a significant interaction between predictability and condition, repeated measures ANOVAs were performed with condition (congruent, incongruent, word naming, colour naming) as a within-subject factor and either statistical learning or rule-based learning score as a dependent variable. Similarly, if the omnibus ANOVA showed a significant interaction between predictability, condition, and period, repeated measures ANOVAs were performed with condition (congruent, incongruent, word naming, colour naming) and period (first half, second half) as a within-subject factor and either statistical learning or rule-based learning score as a dependent variable. Bonferroni-corrected pairwise differences were reported in the Results section. This was done to limit the number of post-hoc tests and still ensure the correction for multiple comparisons. The Huynh-Feldt epsilon was considered as a correction for lack of sphericity in the ANOVA models. Effect sizes are reported as partial eta-squared. Post hoc pairwise comparisons were Bonferroni-corrected. Statistical learning was calculated as the difference between high-probability random and low-probability random trials in RT, accuracy and mean activity. Rule-based learning was calculated as a difference between the high-probability pattern and high-probability random trials in RT, accuracy and mean activity. For all analyses, two types of low-probability triplet configurations were excluded, since previous studies , showed frequent response bias in these combinations: repetition (e.g., 2r-2P-2r, 3r-3P-3r) and trills (e.g., 1r-2P-1r, 4r-3P-4r). Therefore, repetitions and trills were not part of the low-probability random triplet averages and segments for the behavioural and EEG analyses. Given the characteristics of the N450 component in the different conditions (Fig. ), latency of the component was analysed in an exploratory fashion. To quantify the onset latency of the N450, the fractional peak method was used . The onset of the component was marked at the time point when 30% of the peak amplitude was reached. Reporting summary Further information on research design is available in the linked to this article. N = 33 young adults participated in Experiment 1, who were recruited from the voluntary pool for behavioural studies at the TU Dresden (26.6 years ± 6.6, 13 female, 18 male). Since the hypotheses concerned the detection of the learning effect, this parameter was decisive in sample size. Previous reliability analysis showed that stable learning effects can be detected from N > 21 in the case of alternating sequence presentation . However, since the paradigm was combined with a Stroop task in a novel way, we aimed to increase this number, and have at least 30 participants for the final analysis. We estimated a 10% loss in an EEG experiment, therefore, 33 participants were recruited. Due to incomplete testing in one case and low data quality in another case (see EEG recording and analysis ), N = 31 participants’ data were analysed. Experiment 2 was conducted to internally replicate the behavioural results of Experiment 1 (see details in Supplementary Results). N = 30 participants were recruited, from which N = 28 participants’ data were analysed (due to incomplete testing in two cases). Experiment 3 (Supplementary Results) was conducted with N = 20 newly recruited participants. Experiment 4 was conducted with N = 20 newly recruited participants to investigate the generalisability of the findings of Experiments 1 and 2 to other sequential regularities. There is no overlap between the groups of Experiments 1–4. All participants in Experiments 1–4 had a normal or corrected-to-normal vision, including colour discrimination. None of the participants reported taking centrally acting medication or having a history of neurological or psychiatric conditions. Participants were either native in German or had a similar level of proficiency (based on self-reports). Written informed consent was provided prior to enrolment, and participation was rewarded with 10€. The study was approved by the local ethical review committee and was conducted in accordance with the Declaration of Helsinki. Participants completed a paradigm that combined an overt (Colour-Word Stroop-Task) , , and a covert task (Alternating Serial Reaction Time, ASRT) , . The task is shown in Fig. . Participants saw either colour words in German (ROT: red, BLAU: blue, GELB: yellow, GRÜN: green) or coloured asterisks on the centre of the display. Participants were asked to press a button on the keyboard corresponding to the colour of the word or asterisks, irrespective of the meaning of the word. However, if the word was presented in black, they were asked to press the button that corresponds to the meaning of the word. In 25% of the trials, the word was written in the corresponding colour (i.e., ROT in red, congruent condition ). In 25% of the trials, the word was written in a mismatching colour (i.e., ROT in blue, incongruent condition ). In 25% of the trials, the word was presented in black ( word naming condition ), and in the remaining trials only coloured asterisks were presented ( colour naming condition ). These trial types are henceforth referred to as conditions. In each trial, one of the four possible colours was randomly selected. The covert task followed the structure and timing (including the response window) of a previous version of the ASRT task that had been optimised for EEG recordings . Unbeknown to the participants, stimuli were presented according to an eight-element sequence. Within the sequence, pattern (P) and random (r) items alternated with each other. That is, a sequence of 1-r-2-r-3-r-4-r determined the order of the trial types [1 = congruent, 2 = incongruent, 3 = word naming, 4 = colour naming], interleaved by randomly selected trial types. One of the possible permutations of these of the stimulus sequence was selected for the study and presented to the participants in a pseudo-random manner , , . Importantly, the alternating structure used in the current study leads to probability differences between chunks of three successive trials (triplets). In the case of the above-mentioned example sequence, a triplet starting with 2 and ending with 1 is a high-probability triplet that either occurred as a result of a P-r-P or an r-P-r structure. In contrast, a triplet starting with 1 and ending with 2 is a low-probability triplet that could only have an r-P-r structure (see Fig. ). The distinction between low-probability and high-probability triplets does not only describe the distributional but also the second-order transitional probabilities in the task. That is, a final item of a high-probability triplet is a highly predictable continuation of the first item, while a low-probability triplet’s first item does not carry such anticipatory information. For instance, a triplet starting with 1 has a 62.5% probability of ending with 2, while a triplet starting with 2 has only a 12.5% probability of ending with 1. As a consequence of the continuous stimulus presentation and the unmarked triplet structure, each trial can be categorised as the third item of a high-probability or a low-probability triplet. Furthermore, triplets were organised as moving chunks across the stimuli: the third item of a triplet was also a second item of the next triplet and the first item of the subsequent one , . The task consisted of 16 high-probability triplets, which individually occur five times more often than the 48 low-probability triplets. Moreover, the combination of triplet probability (high-probability versus low-probability) and position in the sequence presentation (pattern versus random) leads to three trial types of sequence regularities: high-probability pattern (50% of all trials), high-probability random (12.5% of all trials), and low-probability random triplets (37.5% of all trials). The trial types of cognitive conflict were equally distributed among the trial types of sequence regularity, that is, 25% of high-probability triplets were congruent, 25% incongruent, 25% word naming, 25% colour naming, etc. The timing of the task followed previous EEG studies of the ASRT (Kóbor et al., 2018; 2019) , . Participants saw either a colour name or a string of coloured asterisks on the centre of the screen for 200 ms. It was followed by a blank screen for 500 ms or until the participant pressed a response button. If the response was incorrect, a blank screen was presented for 500 ms after the response onset, followed by an “X” on the centre of the display for another 500 ms. If the participant did not respond in the trial, the 500 ms-long blank screen was followed by an “!” in the centre of the screen for 500 ms. After a correct response or the incorrect/missed response feedback, a 700-ms-long RSI was introduced. The task presentation was organised into blocks, each of them containing 85 trials. Each block started with 5 warm-up trials that were excluded from the analyses, and 10 repetitions of the eight-element sequence. After completing a block, participants received feedback about their mean RT and accuracy in the block, which was presented for 4000 ms. Between blocks, participants could take a short break. The task consisted of 20 blocks in total. During the behavioural analysis, the first 10 blocks were collapsed as first half of the task, and the remaining 10 blocks were collapsed as the second half of the task. The experiment lasted about 1–1.5 h, including dimming and removing the EEG cap. Task presentation was written and controlled by Presentation software (Neurobehavioral Systems). Stimuli were displayed on an LCD screen with a viewing distance of 80 cm. The EEG was recorded from 60 Ag/AgCl electrodes mounted in an equidistant way on an elastic cap (EasyCap, Germany). The ground and reference (Fpz) electrodes were placed at coordinates θ = 58, φ = 78 and θ = 90, φ = 90, respectively. A BrainAmp amplifier and the Brain Vision Recorder 1.2 software (Brain Products, Germany) were used with a sampling rate of 500 Hz. Offline, data were down-sampled offline to 256 Hz. Impedances were kept below 5 kΩ. The recorded EEG signal was pre-processed using “automagic” and EEGLAB in Matlab 2019a (The MathWorks Corp.). First, flat channels were removed and the EEG data were re-referenced to an average reference. Next, the “PREP pipeline” was applied to remove line noise at 50 Hz by using a multitaper algorithm. PREP also removed contaminations by bad channels and consequently created a robust average reference. Next, EEGLAB’s “clean_rawdata” was applied. This included detrending the EEG data with an IIR high-pass filter of 0.5 Hz (slope 80 dB). Flat-line, noisy, and outlier channels were detected and removed by “clean_rawdata”. Time periods with abnormally strong power (i.e., larger than 15 SD relative to calibration data) were reconstructed by using Artefact Subspace Reconstruction (ASR; burst criterion: 15) . Time windows that could not be reconstructed were removed. A low-pass filter of 40 Hz (sinc FIR filter; order of 86) was applied. EOG artefacts were removed using a subtraction method . Muscle, heart, and remaining eye artefacts were automatically classified and removed by using an independent component analysis (ICA) with ICLabel . Removed and missing channels were interpolated (average of 0.82 channels) according to the spherical computation. As a final step of pre-processing, the data were inspected visually and epochs with bad data were removed. This step was only necessary if non-repeating noises (e.g., due to movement) remained in the data. Segmentation of the continuous EEG data was performed in two consecutive steps. First, segments were created separately for the first and second half of the task, each consisting of 10 blocks. The length of these acquisition periods corresponds to previous EEG analyses that used the ASRT task and showed reliable learning effects , , . Subsequently, segments were averaged for all combinations of the sequence regularity (predictability in the ASRT) and Stroop conditions, namely (average and standard deviation of the numbers of trials reported): high-probability pattern congruent (164 ± 25), high-probability pattern incongruent (123 ± 33), high-probability pattern word naming (138 ± 29), high-probability pattern colour naming (156 ± 27), high-probability random congruent (32 ± 6), high-probability random incongruent (33 ± 9), high-probability random word naming (30 ± 8), high-probability random colour naming (43 ± 7), low-probability random congruent (82 ± 13), low-probability random incongruent (66 ± 17), low-probability random word naming (73 ± 14), low-probability random colour naming (73 ± 13). These combinations were named event types . Group-level ERP waveforms were calculated separately for each event type in the acquisition phases. These waveforms were visually inspected to determine if typical ERP correlates of a Stroop task emerged and to determine the latency range that might vary as a function of event types. This was necessary since the combination of an ASRT and a Stroop task was unprecedented, and therefore, the characteristics of the ERP correlates were unknown. Based on the known characteristics of the components, parietal channels were considered for the P3 , and frontocentral channels for the N450 , , . The P3 was quantified as mean amplitude between 280 and 380 ms after stimulus onset on channel P1. The N450 was quantified as mean amplitude between 380–480 ms after stimulus presentation on the channel FCz. Statistics were performed in JASP version 0.16.2 (JASP Team). We analysed the interaction between learning (predictability) and cognitive control (condition) by performing repeated measures ANOVAs. First, the full factorial design of the experiment was analysed with the within-subject factors of predictability (as triplet types: high-probability pattern, high-probability random, low-probability random), condition (congruent, incongruent, word naming, colour naming), and period (first half and second half). If a significant interaction occurred that included the factors of predictability and condition, follow-up analyses were conducted to quantify how predictability could affect RT, accuracy, or mean amplitude of the N450 and P3 components in different types of conflict situations. These post-hoc analyses were done in accordance with previous studies that used the ASRT , , , , . Specifically, we analysed whether the difference between high-probability random and low-probability random trials ( statistical learning ) and the difference between the high-probability pattern and high-probability random trials ( rule-based learning ) was dependent on the condition. If the omnibus ANOVA showed a significant interaction between predictability and condition, repeated measures ANOVAs were performed with condition (congruent, incongruent, word naming, colour naming) as a within-subject factor and either statistical learning or rule-based learning score as a dependent variable. Similarly, if the omnibus ANOVA showed a significant interaction between predictability, condition, and period, repeated measures ANOVAs were performed with condition (congruent, incongruent, word naming, colour naming) and period (first half, second half) as a within-subject factor and either statistical learning or rule-based learning score as a dependent variable. Bonferroni-corrected pairwise differences were reported in the Results section. This was done to limit the number of post-hoc tests and still ensure the correction for multiple comparisons. The Huynh-Feldt epsilon was considered as a correction for lack of sphericity in the ANOVA models. Effect sizes are reported as partial eta-squared. Post hoc pairwise comparisons were Bonferroni-corrected. Statistical learning was calculated as the difference between high-probability random and low-probability random trials in RT, accuracy and mean activity. Rule-based learning was calculated as a difference between the high-probability pattern and high-probability random trials in RT, accuracy and mean activity. For all analyses, two types of low-probability triplet configurations were excluded, since previous studies , showed frequent response bias in these combinations: repetition (e.g., 2r-2P-2r, 3r-3P-3r) and trills (e.g., 1r-2P-1r, 4r-3P-4r). Therefore, repetitions and trills were not part of the low-probability random triplet averages and segments for the behavioural and EEG analyses. Given the characteristics of the N450 component in the different conditions (Fig. ), latency of the component was analysed in an exploratory fashion. To quantify the onset latency of the N450, the fractional peak method was used . The onset of the component was marked at the time point when 30% of the peak amplitude was reached. Further information on research design is available in the linked to this article. Supplemental Material Reporting Summary |
Nocardia infection following ocular surface surgery | a70e8c0c-2f02-4013-9529-87fe50e03f2d | 10042929 | Ophthalmology[mh] | Nocardia is a gram-positive, weakly acid-fast, filamentous bacteria that can cause various ocular infections, such as keratitis, scleritis, and endophthalmitis . Ocular nocardiosis, though has been increasingly reported in recent years, is rarely diagnosed in most parts of the world because of its infrequent occurrence and variable clinical presentation. Nocardia mostly causes opportunistic infections. Trauma , topical steroid use , ocular surgery, and contact lens wearing are common risk factors associated with ocular nocardiosis . Postoperative infections, especially following surface surgery caused by Nocardia, are rare and reported mostly in single cases . This retrospective study was performed to analyze the clinical characteristics and treatment outcomes of Nocardia infection following ocular surface surgery, contributing to a deeper understanding of the disease and its clinical diagnosis and treatment.
Eight culture-proven Nocardia infection cases within a month after ocular surface surgery were retrospectively reviewed. Detailed clinical history, involving demographics, previous medical history, and previous ophthalmic surgery, was analyzed along with the clinical examination and microbiological findings, as well as medical and surgical treatment. The clinical examination consisted of slit-lamp microscopy, confocal microscopy, anterior segment coherence tomography (OCT), and B-scan ultrasonography. A metal blade was adopted for specimens that were used for microbiology workup and obtained from the lesion area scrapings. Each sample was smeared on clean glass slides for microscopic examination and then Gram and Calcofluor White (CFW) staining. Kinyoun staining was performed under strong suspicion of Nocardia. The samples were inoculated onto sheep blood agar, chocolate agar, and nutrient broth maintained at 37 °C for 7 days. A positive smear for Nocardia was defined as gram-positive thin branching filaments on Gram stain, a bright blue fluorescence in CFW stain, or Kinyoun staining. The growth of Nocardia species on culture was marked as chalky-white colonies. Nocardia species from clinical isolates were identified using matrix-assisted laser desorption ionization time-of-flight (MALDI-TOF) mass spectrometry. The susceptibility profiles of Nocardia species for 15 antibiotics were determined by the broth microdilution method. Minimal inhibitory concentrations (MIC) interpretive standards for susceptible and resistant strains followed manufacturers and Clinical Laboratory Standards Institute (CLSI, Wayne, PA) guidelines. Topical broad-spectrum antibiotics treatment was initiated in all patients based on a direct smear examination of the corneal scrapings. Topical 0.3% Gatifloxacin (China Otsuka Pharmaceutical Co., Ltd) combined with tobramycin 3 mg/mL (S.A. Alcon Couvreur N.V.) or 10% cefazolin (Lijian Pharmacy Co., Ltd., China) eye drops were given every 30 min on severe cases. Topical 1% amikacin (Tianfang Pharmaceutical Co., Ltd., China) eye drops were the mainstay of medical therapy once the direct smear examination revealed acid-fast organisms or thin beaded branching filaments. Surgical intervention was considered in cases of poor response to medical therapy. Intact epithelium with no staining on fluorescein application indicated being healed. Recurrence was confirmed if Nocardia reoccurred at the site of the surgical incision for 1 month follow-up time. The final corrected distant vision outcome was collected 2 months after the lesion was healed.
Eight eyes (2 left and 6 right) of 8 patients (5 males and 3 females), aged 27–65, with a median age of 52.9 years, were diagnosed with Nocardia infection after ocular surface surgery. Two patients had a history of hypertension, and one had a history of Granulomatosis with polyangiitis (GPA). The other five patients were healthy. All eight patients had undergone ocular surgery within 1 month of the onset of infection at different times. Among them, three had surgery at our hospital, and five cases were referred from other eye centers. None of them was diagnosed with Nocardia infection before the prior surgery. Three patients were treated with pterygium excision (Fig. a, b, c), and none of the eyes received antifibrotic (mitomycin C or 5-fluorouracil) adjuvant therapy. Three were subjected to conjunctival flap covering (Fig. d, e, f), and two underwent lamellar corneal transplantation (Fig. g, h). Fluoroquinolone antibiotics (0.3% Gatifloxacin or 0.5% Levofloxacin eye drops) were used before and after the former surgery. Tobramycin and Dexamethasone eye drops (S.A. Alcon Couvreur N.V.) were taken after the surgery. The time interval between previous surgery and onset of symptoms varied from 7 to 28 days (mean = 20.5 ± 7.13 days). On slit-lamp examination upon the infection, all the eyes were observed to have infiltrates at the surgical incision site, and the infections were limited to the ocular surface without intraocular involvement. Two cases with corneal ulcers were located at the site of attachment of the head of the pterygium. The infiltrates appeared grey-white with pinhead infiltrates at the edges of the lesion (Fig. a, b). There were two conjunctival infections from auto-transplantation of corneal limbus stem cells, and the conjunctival flaps demonstrated grey-white ulcers (Fig. c, d). Two cases exhibited grey-white corneal ulcers under the conjunctival flap (Fig. e, f). Two cases of corneal graft presented grey-white ulcers (Fig. g, h). Samples from all eight eyes were extracted for microbiological analysis. None of the samples demonstrated bacteria or fungi. Direct smear results and culture of Nocardia species were positive in all cases. Gram’s stain displayed gram-positive, beaded, and branching filaments (Fig. a). Bright blue fluorescence branching thin filaments can be observed in CFW stain (Fig. b). Kinyoun staining illustrated acid-fast, thin, and branching filamentous organisms (Fig. c). Blood and chocolate agar plate demonstrated tiny chalky-white colonies of Nocardia asteroids at the site of inoculation (Fig. d, e). Nutrient broth indicated chalky-white colonies of Nocardia clustered at the bottom of the bottle, and the medium was not turbid (Fig. f). Five of the eight patients obtained confocal microscopy examination, Heidelberg Retina Tomograph III Rostock-cornea-module (HRT III-RCM; Heidelberg Engineering GmbH, Dossenheim, Germany). Filamentous structures suggestive of Nocardia can be detected in two cases. These filamentous structures appeared thin, short, beaded filamentous structures (Fig. ). Among these Nocardia isolates, Nocardia abscessus was the most commonly isolated species (62.5%, 5/8), followed by Nocardia asteroids (25%, 2/8) and Nocardia farcinica (12.5%, 1/8). Antibiotic sensitivity testing revealed that all samples were sensitive to Amikacin, linezolid, and trimethoprim-sulfamethoxazole (TMP-SMX). Additionally, it demonstrated low MIC values against different Nocardia species. However, resistance was observed to ciprofloxacin in 6 of 8 isolates (75%), clarithromycin in 4 of 8 isolates (50%), and forβ-lactam antibiotics, including imipenem, cefepime, amoxicillin-clavulanic acid, and ceftriaxone, which all exhibited a poor performance against Nocardia species. However, all the patients responded poorly to drug treatment. One patient with GPA history and Nocardia infection after full lamellar keratoplasty gave up treatment. Seven of the eight cases required some form of surgical intervention, four obtained keratoplasty, and three were treated by infectious lesion resection. Nocardia recurred in three of the seven reoperated patients (42.8%). Corneal ulcers were cured by eye drops (1% Amikacin) in two of them. One of the patients (Fig. e) with conjunctival Nocardia infection 20 days following penetrating keratoplasty (PKP) (Fig. a) responded poorly to Amikacin despite the positive antibiotic sensitivity testing. Infectious lesion resection was performed twice (Fig. b, c, d). However, infectious scleritis and endophthalmitis occurred, and the abscess recurred at the lesion and even was eviscerated (Fig. e, f).
Nocardia species are ubiquitous and can be discovered in water, soil, dust, and decaying vegetation worldwide . Nocardia can be found as a saprophyte on the skin and upper respiratory tract, while it does not present as normal flora within the eye . Corneal and conjunctival tissue damage is the pathological foundation of the Nocardia ocular infection. Surgical trauma is one of the most common risk factors for Nocardia ocular infection. Former reports of Nocardia infection after surgeries mostly involved single cases, such as Femtosecond Laser-Assisted Lamellar Keratectomy and cataract surgery . This paper reported eight Nocardia infection cases following ocular surface surgery: three patients with pterygium excision, three patients undergoing conjunctival flap covering, and two patients subjected to lamellar corneal transplantation. They all had a large area of conjunctiva cut open during the operation. As a result, the conjunctiva’s integrity was destroyed, and the conditioned pathogenic bacteria were exposed to the conjunctiva wound, leading to a higher chance of secondary infection. Moreover, all the lesions were infiltrated at the site of the surgical incision. Garg and Javadi reported three patients (four eyes) with the outbreak of Nocardia keratitis after corneal refractive surgery. They suspected that the most probable cause of the outbreak was non-standard sterilization and non-aseptic operation. In this study, operations of all the patients were performed by different surgeons at different times. There was no cluster infection. Generally, Nocardia infection does not occur in clinical practice, with a global prevalence of below 2% . Being mistaken for fungal or viral keratitis leads to delayed treatment and an increased risk of permanent visual impairment. This delayed diagnosis may be induced by a lack of familiarity with this uncommon pathogen and its variable presentation. The classic description of Nocardia keratitis was a yellow-white discrete pinhead-sized appearance in the infiltrated area, forming a characteristic wreath pattern . None of the patients was suspected of Nocardia after the initial clinical examination. Hence, atypical clinical appearance may be confused with that of herpes simplex keratitis or fungal keratitis. Delay in the identification sometimes results in injudicious usage of topical corticosteroids, which may exacerbate the infection. Topical corticosteroid eye drops are used in all patients in our series after their initial surgery. One patient with a history of GPA had used oral corticosteroids previously for many years. Case reports have suggested that topical corticosteroids may result in recurrence of the infection and larger infiltrate/scar sizes . Our study unveiled that the use of adjunctive corticosteroids was associated with an atypical clinical appearance and poor response to the sensitive drugs. Amikacin has been effective and is considered the first-line drug of choice in Nocardia infection . Nevertheless, its susceptibility may differ with geography or diverse isolates. Susceptibility testing is crucial to determine the most effective drugs for ocular Nocardia infections. All of the Nocardia strains were susceptible to amikacin in this current study, consistent with the largest-sample-size surveillance study on Nocardia strains and nocardiosis throughout China . However, all the patients responded poorly to amikacin in this current study. Seven of the eight cases required some form of reoperation. The use of corticosteroids contributed to this phenomenon. Four recurrences occurred in one patient after conjunctival flap covering. Sensitive antibiotics Amikacin and infectious lesion resection responded poorly. Moreover, recurrence appeared and even was eviscerated. This might be in that the rollback conjunctival flap was not checked and excised during the reoperation procedure. Hidden Nocardia or Nocardial scleritis may exist under the conjunctiva. This case suggested that early surgical intervention must remove the infectious lesion completely to reduce the recurrence. Unfortunately, whether it was usefully treated with sub-tenon or intravitreal injections of amikacin in the early phase was not observed in our study, and other sensitive antibiotics such as linezolid and TMP-SMX are inapplicable due to the restrictions in the current hospital. The typical first-line prophylactic preoperative application of ocular antibiotics, such as fluoroquinolones, was frequently reported to resist this pathogen . In the current study, resistance was observed to fluoroquinolones in 6 of 8 isolates (75%). Nocardia infections should be suspected when secondary infection occurs in a surgical incision with an atypical clinical presentation. Previous reports have verified that a confocal microscope can visualize Nocardia, revealed as multiple, thin (< 1.5 μm), short, and beaded filamentous structures that demonstrated right-angled branching . Described filamentous structures of Nocardia were detected by confocal microscopy in two of our cases. The manufacturer quotes the minimum optical transverse resolution of 1 μm with the HRT III-RCM. Nocardia thinner than 1 μm may be difficult to detect owing to a limitation of the resolution. The sensitivity and positive predictive values in diagnosing Nocardia keratitis may increase, especially in cases where scarp cannot be achieved, with the increasing use of the new-generation confocal microscopy that has greater magnification (800 ×) and clearer images than that of earlier generation microscopy (380 ×) as a noninvasive diagnostic tool. There are several limitations to this study. In the current study, only eight samples were isolated, resulting in a small number of each Nocardia species. Additionally, a standardized treatment protocol is deficient. However, the current series of Nocardia infections following ocular surface surgery was first reported, and some essential clinical and microbiological information was revealed in the management of these patients.
Surgical trauma is a risk factor for ocular Nocardia infection. Nocardia infection should be suspected when secondary infection occurs in a surgical incision with an atypical clinical presentation. The use of corticosteroids may influence the efficacy of drugs. Complete removal of lesions may reduce the recurrence of Nocardia infection with poor drug treatment effects.
|
Deletion of | f1b96c1c-2486-492e-9a38-c22eb23faaa9 | 10043021 | Microbiology[mh] | Bioethanol is the biofuel most used worldwide in the transportation sector, its production having increased regularly since the early nineties . In 2020, the USA and Brazil accounted, respectively, for 53% and 30% of global biofuel production . Bioethanol is produced from a wide variety of renewable resources (feedstocks), and the yeast Saccharomyces cerevisiae is by far the microorganism most frequently used to carry out the biological process exploited in industrial-scale ethanol production, thanks to its unmatched ethanol yield (> 90%), productivity (> 1 g.L −1 .h −1 ), and tolerance (> 40 g.L −1 ) and its ability to ferment a wide range of sugars . In Brazil, bioethanol is produced from sugarcane molasses by high-cell-density fed-batch fermentation, in volumes reaching half a million liters. The elevated cell densities reached in this process (10% wet weight per volume) allow 6–12-h fermentations followed by recycling of the yeast biomass throughout the production season . As this process is operated under non-sterile conditions, bacterial contamination occurs at high frequency . Contaminants negatively impact yeast fermentation through competition for scarce nutrients and release of growth inhibitors. This generates considerable economic losses, mostly due to a reduced ethanol production efficiency and also, if the contamination is uncontrolled, to fermenter shutdown for cleaning , . In nature, yeasts and lactic acid bacteria (LAB) are often encountered together, and metabolite exchanges between them have been reported , . Sometimes this cohabitation is desirable, as during production of fermented products such as kefir or kimchi . Yet LAB are also the most common and troublesome bacterial contaminants found in ethanol production facilities, because they grow rapidly and tolerate high temperatures and low pH . Besides competing with yeast for essential nutrients, LAB produce many compounds that inhibit yeast growth, including lactic acid, acetic acid, caproic acid, carbon dioxide, diacetyl, hydrogen peroxide, reuterin, phenyllactic acid, 3-carboxylic fatty acids, and cyclic peptides . Historically, several methods have been applied to prevent their undesirable growth in yeast fermentation processes. In general, bacterial infections are easily controlled with antibiotics, acid treatments, ammonia, and urea-hydrogen peroxide , . Antimicrobial compounds such as c-hydroxycinnamates, organic acids, and membrane-active antimicrobial peptides have also been used, with varying degrees of success. These methods, however, pose a potential biological and environmental hazard if waste is not properly disposed of. Furthermore, some of these treatments are quite costly . Hence, novel strategies are needed to decrease LAB proliferation in industrial bioethanol production plants. LAB are typically auxotrophic for several amino acids and thus depend, for growth, on an external amino acid supply. Importantly, yeast cells grown in defined media have been found to excrete amino acids which can be used by LAB for growth , . Hence, an important and still open question is whether such cross-feeding contributes to propagation of contaminating LAB in bioethanol production systems. In yeast, excretion of amino acids is mediated by proteins of Drug:H + Antiporter family 1 (DHA1), such as Aqr1, Qdr2, and Qdr3 , . Importantly, we have recently demonstrated that deleting these three genes in a laboratory strain reduces the yeast’s ability to cross-feed Lactobacillus fermentum , a LAB commonly found among contaminating bacteria in bioethanol production facilities. This observation suggests that cultivation of a bioethanol-producing yeast strain lacking specific DHA1-family transporters might be a way to reduce LAB contamination. In this study we show that in a sterile minimal medium, the industrial yeast strain Ethanol Red, selected for ethanol production, sustains growth of co-cultivated Lactobacillus fermentum much more efficiently than a mutant derivative lacking the QDR3 gene. We next show that growth of Ethanol Red on a nonsterile sugarcane-molasses-based medium is accompanied by an increase in lactic acid, and that this increase is no longer observed when the strain lacks the QDR1 , QDR2, and QDR3 genes. These results suggest that using industrial yeast strains with mutations in QDR genes may be a good way to reduce LAB contaminations in bioethanol-production fermentations tanks.
The bioethanol-producing yeast strain Ethanol Red cross-feeds Lactobacillus fermentum Laboratory strains of S. cerevisiae have been reported to naturally excrete amino acids which can be used by co-cultivated lactic acid bacteria (LAB) unable to grow without an external amino acid supply , , . This cross-feeding can be visualized in experiments where the two microorganisms are co-cultivated in an appropriate amino-acid-free medium. For instance, we have previously reported that L. fermentum can grow in a MES-buffered glucose minimal medium (code number 169) containing NH 4 + as sole nitrogen source, and thus in the absence of any external amino acid supply, when it is co-cultivated with a laboratory yeast strain . We sought to determine whether an industrial yeast strain selected for high-efficiency ethanol production could also cross-feed L. fermentum . We chose the strain Ethanol Red , a standard in industrial biofuel production. This industrial strain was compared with our haploid wild-type reference strain 23344c (derived from the Σ1278b wild-type laboratory strain ), whose unique ura3 auxotrophy was complemented by a plasmid-borne URA3 gene. We first compared the growth of the Ethanol Red and 23344c strains on the MES-buffered NH 4 + medium containing glucose or ethanol as carbon source. While the two strains displayed similar growth on glucose, Ethanol Red grew much more slowly on ethanol (Fig. A), in keeping with a previous study . We then cultivated the cells in wells filled with buffered glucose NH 4 + medium and subdivided into two compartments separated by a solute-permeable membrane (Fig. A). One compartment was inoculated with the 23344c or Ethanol Red strain and the other with L. fermentum . Just after inoculation and after two days of growth, culture samples were withdrawn and yeast and L. fermentum cell densities were quantified by counting the number of colony-forming units (CFU/ml) obtained after plating culture samples on appropriate solid rich media. After two days of cultivation, 23344c and Ethanol Red were found to have proliferated similarly (Fig. B) . L. fermentum had also proliferated when co-cultivated with either 23344c or Ethanol Red strains (Fig. C, D). That this growth of L. fermentum was due to cross-feeding by yeast cells excreting amino acids was confirmed in control experiments: the bacterium failed to grow when placed alone in the MES-buffered glucose NH 4 + medium, but proliferated well if the medium was supplemented with a mix of the twenty proteinaceous amino acids (Fig. C). This result indicates that the Ethanol Red industrial yeast strain can excrete amino acids and cross-feed L. fermentum . To determine which amino acid auxotrophies of L. fermentum are efficiently compensated by co-culture with yeast, we used the MES-buffered glucose NH 4 + medium containing all twenty amino acids as a positive control and tested how omitting each amino acid individually affected L. fermentum growth. The bacterium was found to grow normally in the absence of alanine, asparagine, proline, glutamate, or serine (Fig. B), but it failed to grow in the absence of threonine, phenylalanine, methionine, glutamine, histidine, arginine, cysteine, tyrosine, leucine, lysine, tryptophan, isoleucine, or valine. In the absence of glycine or aspartate, its growth was reduced but detectable. These results indicate that both Ethanol Red and 23344c excrete sufficient amounts of each of the thirteen above-listed essential amino acids to support growth of L. fermentum . In the Ethanol Red strain, the Qdr3 amino-acid exporter contributes importantly to cross-feeding of Lactobacillus fermentum In a recent study using strain 23344c as a reference wild type, we found the DHA1-family membrane transporters Aqr1, Qdr2, and Qdr3 to contribute to amino-acid excretion and cross-feeding of L. fermentum during growth in the MES-buffered glucose NH 4 + medium . To determine if this holds true for the Ethanol Red strain, we used CRISPR-Cas9 and adapted transformation protocols to first produce a mutant derivative lacking the QDR3 gene. We obtained many clones with heterozygous deletion of QDR3 and a single homozygous qdr3Δ mutant. When we co-cultivated this mutant with L. fermentum in the MES-buffered glucose NH 4 + medium without any added amino acid, the qdr3Δ strain was found to support growth of L. fermentum , though less efficiently than the original Ethanol Red strain (Fig. A and Fig. ). To ascertain that this phenotype was due to the qdr3Δ mutation, we transformed the mutant with a plasmid bearing the QDR3 gene, using as a selection marker a resistance gene for the antibiotic geneticin. The transformed strain was initially cultured in the presence of geneticin and then cells were collected, washed, and used to inoculate antibiotic-free minimal medium. After 48 h of co-culture with L. fermentum and CFU counting, the plasmid-transformed qdr3Δ mutant was found to support growth of the bacterium as efficiently as the Ethanol Red strain (Fig. A). Furthermore, whether transformed or not with the QDR3 -bearing plasmid, the qdr3Δ mutant proliferated as well as Ethanol Red (Fig. A). In a parallel experiment, the qdr3Δ mutant isolated from strain 23344c likewise proliferated as well as its parental wild-type (Fig. B), but in keeping with previous observations , the two strains showed a similar ability to support growth of L. fermentum (Fig. A). These observations thus show, unexpectedly, that Qdr3 contributes more importantly in Ethanol Red than in 23344c to excretion of one or several amino acids essential to growth of L. fermentum. It thus seems that in 23344c, other amino acid exporters can efficiently perform this excretion. The results presented in Fig. A also show that the qdr3Δ mutant isolated from Ethanol Red strain can still support growth of L. fermentum . This suggests that excretion of amino acids required for L. fermentum proliferation is reduced but not abolished when Qdr3 is not functional. We thus examined whether deletion of additional amino acid exporter genes might further reduce the ability of Ethanol Red to cross-feed L. fermentum . Specifically, we applied CRISPR-Cas9 to the qdr3Δ mutant to delete, in a single step, the highly similar and adjacent QDR1 and QDR2 genes. These likely originate from a duplication event, and chromosomal synteny analysis suggests that the ancestral QDR1/2 gene is a paralog of the AQR1 gene encoding another DHA1-family amino acid exporter , . The derived qdr1Δ qdr2Δ qdr3Δ mutant strain co-cultivated with L. fermentum sustained growth of the bacterium as efficiently as the single qdr3Δ mutant (Fig. B and Fig. ). In contrast, the same QDR1 and QDR2 deletions in the qdr3Δ mutant of strain 23344c resulted in decreased propagation of co-cultivated L. fermentum (Fig. B and Fig. ). In conclusion, under the conditions used here, the Qdr3 amino acid exporter of Ethanol Red plays an important role in cross-feeding L. fermentum . Residual cross-feeding does occur, however, even if Qdr1 and Qdr2 are also lost. In laboratory strain 23344c, Qdr3 does not seem so important in supporting growth of L. fermentum , but additional loss of Qdr1 and Qdr2 considerably reduces cross-feeding. It thus seems that the relative contributions of the Qdr proteins to amino-acid excretion differ significantly between Ethanol Red and 23344c. The AQR1 gene encodes another well-characterized DHA1-family amino acid exporter , . It was not possible to investigate the role of Aqr1 in amino acid excretion by Ethanol Red because homozygous deletion of the AQR1 gene in this strain proved much more difficult than for the QDR genes. Loss of Qdr amino acid exporters in the Ethanol Red strain prevents an increase in lactic acid during growth on molasses Nonsterile crop-derived culture media used for industrial production of bioethanol generally contain several LAB species whose propagation during fermentation causes a detectable increase in lactic acid . To assess the importance of amino-acid excretion by industrial yeast strains in the proliferation of these LAB, flasks containing equal volumes of a nonsterile molasses-based medium were inoculated with equivalent cell samples of the Ethanol Red strain or its qdr3Δ or qdr1Δ qdr2Δ qdr3Δ mutant derivatives. The culture flasks were incubated for 240 h at 30 °C with shaking and their gradual weight loss due to CO 2 production was measured. This allowed us to compare the efficiency of glucose fermentation by the three strains, which proved reproducibly similar (Fig. A). Accordingly, we observed no significant difference in ethanol concentration, as measured after 240 h in several independent experiments (Fig. B). We also measured the lactic acid concentration in the media, before and after yeast cultivation. In cultures of Ethanol Red , we detected an increase in lactic acid, indicating that LAB proliferated to some extent (Fig. C, D). A similar increase in lactic acid was observed in cultures of the qdr3Δ mutant, but none was detected in cultures of the qdr1Δ qdr2Δ qdr3Δ mutant (Fig. C, D). This shows that lactic acid production was dependent on yeast and its ability to produce functional Qdr proteins. This observation, together with the results of cross-feeding experiments in defined MES-buffered glucose NH 4 + medium, shows that Ethanol Red excretes amino acids via the Qdr exporters and that this excretion supports growth of LAB naturally present in molasses. It also suggests that deleting specific QDR genes in bioethanol-producing yeast strains may be a novel strategy for reducing the risk of contamination of bioethanol fermentation tanks with lactic acid bacteria.
Ethanol Red cross-feeds Lactobacillus fermentum Laboratory strains of S. cerevisiae have been reported to naturally excrete amino acids which can be used by co-cultivated lactic acid bacteria (LAB) unable to grow without an external amino acid supply , , . This cross-feeding can be visualized in experiments where the two microorganisms are co-cultivated in an appropriate amino-acid-free medium. For instance, we have previously reported that L. fermentum can grow in a MES-buffered glucose minimal medium (code number 169) containing NH 4 + as sole nitrogen source, and thus in the absence of any external amino acid supply, when it is co-cultivated with a laboratory yeast strain . We sought to determine whether an industrial yeast strain selected for high-efficiency ethanol production could also cross-feed L. fermentum . We chose the strain Ethanol Red , a standard in industrial biofuel production. This industrial strain was compared with our haploid wild-type reference strain 23344c (derived from the Σ1278b wild-type laboratory strain ), whose unique ura3 auxotrophy was complemented by a plasmid-borne URA3 gene. We first compared the growth of the Ethanol Red and 23344c strains on the MES-buffered NH 4 + medium containing glucose or ethanol as carbon source. While the two strains displayed similar growth on glucose, Ethanol Red grew much more slowly on ethanol (Fig. A), in keeping with a previous study . We then cultivated the cells in wells filled with buffered glucose NH 4 + medium and subdivided into two compartments separated by a solute-permeable membrane (Fig. A). One compartment was inoculated with the 23344c or Ethanol Red strain and the other with L. fermentum . Just after inoculation and after two days of growth, culture samples were withdrawn and yeast and L. fermentum cell densities were quantified by counting the number of colony-forming units (CFU/ml) obtained after plating culture samples on appropriate solid rich media. After two days of cultivation, 23344c and Ethanol Red were found to have proliferated similarly (Fig. B) . L. fermentum had also proliferated when co-cultivated with either 23344c or Ethanol Red strains (Fig. C, D). That this growth of L. fermentum was due to cross-feeding by yeast cells excreting amino acids was confirmed in control experiments: the bacterium failed to grow when placed alone in the MES-buffered glucose NH 4 + medium, but proliferated well if the medium was supplemented with a mix of the twenty proteinaceous amino acids (Fig. C). This result indicates that the Ethanol Red industrial yeast strain can excrete amino acids and cross-feed L. fermentum . To determine which amino acid auxotrophies of L. fermentum are efficiently compensated by co-culture with yeast, we used the MES-buffered glucose NH 4 + medium containing all twenty amino acids as a positive control and tested how omitting each amino acid individually affected L. fermentum growth. The bacterium was found to grow normally in the absence of alanine, asparagine, proline, glutamate, or serine (Fig. B), but it failed to grow in the absence of threonine, phenylalanine, methionine, glutamine, histidine, arginine, cysteine, tyrosine, leucine, lysine, tryptophan, isoleucine, or valine. In the absence of glycine or aspartate, its growth was reduced but detectable. These results indicate that both Ethanol Red and 23344c excrete sufficient amounts of each of the thirteen above-listed essential amino acids to support growth of L. fermentum .
Ethanol Red strain, the Qdr3 amino-acid exporter contributes importantly to cross-feeding of Lactobacillus fermentum In a recent study using strain 23344c as a reference wild type, we found the DHA1-family membrane transporters Aqr1, Qdr2, and Qdr3 to contribute to amino-acid excretion and cross-feeding of L. fermentum during growth in the MES-buffered glucose NH 4 + medium . To determine if this holds true for the Ethanol Red strain, we used CRISPR-Cas9 and adapted transformation protocols to first produce a mutant derivative lacking the QDR3 gene. We obtained many clones with heterozygous deletion of QDR3 and a single homozygous qdr3Δ mutant. When we co-cultivated this mutant with L. fermentum in the MES-buffered glucose NH 4 + medium without any added amino acid, the qdr3Δ strain was found to support growth of L. fermentum , though less efficiently than the original Ethanol Red strain (Fig. A and Fig. ). To ascertain that this phenotype was due to the qdr3Δ mutation, we transformed the mutant with a plasmid bearing the QDR3 gene, using as a selection marker a resistance gene for the antibiotic geneticin. The transformed strain was initially cultured in the presence of geneticin and then cells were collected, washed, and used to inoculate antibiotic-free minimal medium. After 48 h of co-culture with L. fermentum and CFU counting, the plasmid-transformed qdr3Δ mutant was found to support growth of the bacterium as efficiently as the Ethanol Red strain (Fig. A). Furthermore, whether transformed or not with the QDR3 -bearing plasmid, the qdr3Δ mutant proliferated as well as Ethanol Red (Fig. A). In a parallel experiment, the qdr3Δ mutant isolated from strain 23344c likewise proliferated as well as its parental wild-type (Fig. B), but in keeping with previous observations , the two strains showed a similar ability to support growth of L. fermentum (Fig. A). These observations thus show, unexpectedly, that Qdr3 contributes more importantly in Ethanol Red than in 23344c to excretion of one or several amino acids essential to growth of L. fermentum. It thus seems that in 23344c, other amino acid exporters can efficiently perform this excretion. The results presented in Fig. A also show that the qdr3Δ mutant isolated from Ethanol Red strain can still support growth of L. fermentum . This suggests that excretion of amino acids required for L. fermentum proliferation is reduced but not abolished when Qdr3 is not functional. We thus examined whether deletion of additional amino acid exporter genes might further reduce the ability of Ethanol Red to cross-feed L. fermentum . Specifically, we applied CRISPR-Cas9 to the qdr3Δ mutant to delete, in a single step, the highly similar and adjacent QDR1 and QDR2 genes. These likely originate from a duplication event, and chromosomal synteny analysis suggests that the ancestral QDR1/2 gene is a paralog of the AQR1 gene encoding another DHA1-family amino acid exporter , . The derived qdr1Δ qdr2Δ qdr3Δ mutant strain co-cultivated with L. fermentum sustained growth of the bacterium as efficiently as the single qdr3Δ mutant (Fig. B and Fig. ). In contrast, the same QDR1 and QDR2 deletions in the qdr3Δ mutant of strain 23344c resulted in decreased propagation of co-cultivated L. fermentum (Fig. B and Fig. ). In conclusion, under the conditions used here, the Qdr3 amino acid exporter of Ethanol Red plays an important role in cross-feeding L. fermentum . Residual cross-feeding does occur, however, even if Qdr1 and Qdr2 are also lost. In laboratory strain 23344c, Qdr3 does not seem so important in supporting growth of L. fermentum , but additional loss of Qdr1 and Qdr2 considerably reduces cross-feeding. It thus seems that the relative contributions of the Qdr proteins to amino-acid excretion differ significantly between Ethanol Red and 23344c. The AQR1 gene encodes another well-characterized DHA1-family amino acid exporter , . It was not possible to investigate the role of Aqr1 in amino acid excretion by Ethanol Red because homozygous deletion of the AQR1 gene in this strain proved much more difficult than for the QDR genes.
Ethanol Red strain prevents an increase in lactic acid during growth on molasses Nonsterile crop-derived culture media used for industrial production of bioethanol generally contain several LAB species whose propagation during fermentation causes a detectable increase in lactic acid . To assess the importance of amino-acid excretion by industrial yeast strains in the proliferation of these LAB, flasks containing equal volumes of a nonsterile molasses-based medium were inoculated with equivalent cell samples of the Ethanol Red strain or its qdr3Δ or qdr1Δ qdr2Δ qdr3Δ mutant derivatives. The culture flasks were incubated for 240 h at 30 °C with shaking and their gradual weight loss due to CO 2 production was measured. This allowed us to compare the efficiency of glucose fermentation by the three strains, which proved reproducibly similar (Fig. A). Accordingly, we observed no significant difference in ethanol concentration, as measured after 240 h in several independent experiments (Fig. B). We also measured the lactic acid concentration in the media, before and after yeast cultivation. In cultures of Ethanol Red , we detected an increase in lactic acid, indicating that LAB proliferated to some extent (Fig. C, D). A similar increase in lactic acid was observed in cultures of the qdr3Δ mutant, but none was detected in cultures of the qdr1Δ qdr2Δ qdr3Δ mutant (Fig. C, D). This shows that lactic acid production was dependent on yeast and its ability to produce functional Qdr proteins. This observation, together with the results of cross-feeding experiments in defined MES-buffered glucose NH 4 + medium, shows that Ethanol Red excretes amino acids via the Qdr exporters and that this excretion supports growth of LAB naturally present in molasses. It also suggests that deleting specific QDR genes in bioethanol-producing yeast strains may be a novel strategy for reducing the risk of contamination of bioethanol fermentation tanks with lactic acid bacteria.
The proliferation of contaminating LAB in yeast fermentation tanks is a recurring problem causing a significant reduction of bioethanol production yields , . To prevent or at least reduce the risk of such contaminations at bioethanol production plants, it is imperative to better understand the factors promoting LAB proliferation in this context. Contaminating LAB are typically auxotrophic for multiple amino acids and thus require an external supply of amino acids in order to proliferate. On the other hand, it is well established that yeast cells naturally excrete several amino acids , . We have previously shown that the yeast DHA1-family transporters Aqr1, Qdr2, and Qdr3 contribute to amino acid excretion from cells , . Although why this excretion takes place remains unclear, it is known to be favored by conditions such as nitrogen overflow and/or a limited supply of other nutrients , . We therefore reasoned that this natural excretion of amino acids by yeast might contribute to LAB proliferation in nonsterile media used in industrial fermentations. To test this hypothesis, we introduced homozygous deletions of genes encoding DHA1-family amino-acid exporters into the genome of strain Ethanol Red used in bioethanol production. Our most important observation concerns lactic acid production, indicative of LAB proliferation, in nonsterile molasses-based medium containing either the Ethanol Red strain or its triple mutant lacking the QDR1 , QDR2 and QDR3 genes: in contrast to cultures seeded with the parental strain, those containing the triple mutant showed no significant increase in lactic acid over a ten-day growth period. We conclude that LAB proliferation was supported by yeast-excreted amino acids rather than by free amino acids naturally available in the medium or released as a consequence of cell death. Our work further highlights Qdr1, -2, and -3 as three DHA1-family transporters that likely play a particularly important role in LAB-growth-sustaining excretion of amino acids by Ethanol Red (Fig. ). To our knowledge, this has not been described before. Importantly, the absence of these proteins did not reduce this strain’s ethanol-producing capacity. This opens the prospect of exploiting yeast qdr mutants to reduce or control the risk of LAB contamination during bioethanol production. To further assess the potential of this strategy, it will be useful to conduct further experiments with alternative industrial cultivation media and bioethanol-producing yeast strains. Additional work will also be needed to evaluate the respective contributions of different DHA1-family amino acid exporters to LAB cross-feeding under industrial fermentation conditions. For instance, although we show here that lactic acid was not produced by LAB in sugarcane molasses-based medium seeded with the qdr1Δ qdr2Δ qdr3 Δ triple-mutant strain, we cannot rule out the possibility that a qdr1Δ or qdr2Δ single mutant or a qdr1Δ qdr2Δ double mutant might have displayed a similar advantage. This seems unlikely, however, as previous cross-feeding experiments have shown that several DHA1-family genes need to be inactivated in order to markedly reduce excretion of a single amino acid, e.g. threonine or homoserine , . Furthermore, in peculiar culture media, deletion of additional DHA1 -family genes might be necessary to reduce amino acid excretion. In our laboratory strain, for instance, the Aqr1 transporter plays an important role in amino acid excretion , . It will thus be of interest to delete the corresponding gene in Ethanol Red and in its qdr1Δ qdr2Δ qdr3Δ derivative to see if this deletion eliminates the residual cross-feeding of L. fermentum observed with the qdr1Δ qdr2Δ qdr3Δ mutant in synthetic buffered minimal medium. For reasons that remain to be clarified, however, deleting this gene proved particularly difficult and could not be achieved in the framework of this work. Our study further shows that in a synthetic medium, deletion of QDR3 alone is sufficient to reduce cross-feeding of L. fermentum by the Ethanol Red and that further deletion of QDR1 and QDR2 does not enhance this effect. This contrasts with the results obtained with the same strains grown on the nonsterile molasses-based medium, which likely contains several LAB species: in this case, deletion of QDR3 alone did not significantly reduce lactic acid production. Why these differences exist remains unknown. Perhaps growth on molasses medium alters the expression profile of DHA1 -family genes in Ethanol Red , or perhaps the LAB species present in molasses have auxotrophies that differ from those present in L. fermentum . This illustrates the possibility that reducing LAB growth might require introducing different combinations of qdr mutations according to the culture medium used. Reducing LAB propagation could favor the proliferation of other bacteria, for instance acetic acid bacteria . In support of this view, we observed a significant increase in acetic acid production in nonsterile molasses cultures inoculated with the qdr1Δ qdr2Δ qdr3Δ triple mutant, as compared with the parental Ethanol Red strain (Fig. ). Thus, reduced amino acid excretion by yeast might promote growth of acetic acid bacteria by causing lesser proliferation of several LAB species. Interestingly, acetate production was also observed in molasses medium seeded with the qdr3Δ single mutant, although lactic acid production was not reduced (Fig. ). Perhaps the lack of Qdr3-dependent excretion of specific amino acids prevented proliferation of some but not all LAB species, this being sufficient to allow proliferation of acetic acid bacteria in the available ecological niche. This illustrates the complexity of interactions between microorganisms in co-cultures. Further investigation of these interactions and of the molecular mechanisms underlying them, including cross-feeding, should make it possible to better control bacterial proliferation during bioethanol production. In conclusion, we report that deletion of QDR genes in the Ethanol Red strain reduces its ability to support propagation of lactic acid bacteria and provides a potentially efficient means of limiting contamination of fermentations by LAB during industrial bioethanol production. This approach can in principle be applied also to second-generation bioethanol production processes relying on nonsterile lignocellulose biomass and to industrial production by yeast of other compounds of interest.
Strains and growth conditions The Saccharomyces cerevisiae strains used in this study (Table ) derive from the laboratory strain Σ1278b or from the industrial strain Ethanol Red (Leaf, Lesaffre, Marcq-en-Baroeul, France). Cells of Σ1278b-derived strains were transformed with plasmids (Table ) as previously described . The same protocol was used to transform cells of the Ethanol Red strain, except that PEG4000 was replaced with PEG3350 and higher amounts of DNA (10–25 μg) were used. To introduce homozygous gene deletions in the Ethanol Red strain, we applied CRISPR-Cas9 as previously described . Yeast and Lactobacillus fermentum (strain LMG17551 from the Belgian Coordinated Collections of Microorganisms) were co-cultivated on a MES-buffered, amino acid free minimal medium (code number 169) containing glucose (3%) as a carbon source and ammonium as a nitrogen source added as (NH 4 ) 2 SO 4 (0.5%), as previously described . For cultures on nonsterile molasses, the growth medium consisted of sugarcane molasses (30%), (NH 4 ) 2 SO 4 (567 mg/L), and a nutrient mix (350 mg/L) (Gusmer Enterprises, LN W148320). Cross-feeding of L. fermentum by yeast cells Assays for measuring cross-feeding between yeast and L. fermentum cells were carried out in 6-well plates equipped with ThinCert cell culture inserts (Greiner -657,640), as previously described . Fermentation efficiency assay Yeast cells were first pre-grown in a standard rich medium containing glucose (10%), bactopeptone (2%), and yeast extract (1%). Flasks filled with 60 ml nonsterile sugarcane-molasses-based medium were then inoculated (5⋅10 6 cells/ml) and incubated at 30 °C with shaking. Weight loss of the flasks due to CO 2 production was monitored until it became negligible. Analysis of organic acids and ethanol in culture media Production of lactic acid and acetic acid was assayed by High Performance Liquid Chromatography (HPLC) with the LCM1 Waters Chromatographic chain (600 Pump, 2487 Absorbance Detector and Degasser, 717 Plus Autosampler) and a column (Aminex HPX-87H, Biorad) thermoregulated at 65 °C. Ethanol production was assayed by headspace Gas Chromatography (GC), with Perkin-Elmer 8000 series 2 GC instrumentation, and HS-40 Automatic Injector, and an Agilent CP- WAX 52 CB WCOT FS 50 mm × 0,32 mm column. The results are given in % ethanol v/v.
The Saccharomyces cerevisiae strains used in this study (Table ) derive from the laboratory strain Σ1278b or from the industrial strain Ethanol Red (Leaf, Lesaffre, Marcq-en-Baroeul, France). Cells of Σ1278b-derived strains were transformed with plasmids (Table ) as previously described . The same protocol was used to transform cells of the Ethanol Red strain, except that PEG4000 was replaced with PEG3350 and higher amounts of DNA (10–25 μg) were used. To introduce homozygous gene deletions in the Ethanol Red strain, we applied CRISPR-Cas9 as previously described . Yeast and Lactobacillus fermentum (strain LMG17551 from the Belgian Coordinated Collections of Microorganisms) were co-cultivated on a MES-buffered, amino acid free minimal medium (code number 169) containing glucose (3%) as a carbon source and ammonium as a nitrogen source added as (NH 4 ) 2 SO 4 (0.5%), as previously described . For cultures on nonsterile molasses, the growth medium consisted of sugarcane molasses (30%), (NH 4 ) 2 SO 4 (567 mg/L), and a nutrient mix (350 mg/L) (Gusmer Enterprises, LN W148320).
L. fermentum by yeast cells Assays for measuring cross-feeding between yeast and L. fermentum cells were carried out in 6-well plates equipped with ThinCert cell culture inserts (Greiner -657,640), as previously described .
Yeast cells were first pre-grown in a standard rich medium containing glucose (10%), bactopeptone (2%), and yeast extract (1%). Flasks filled with 60 ml nonsterile sugarcane-molasses-based medium were then inoculated (5⋅10 6 cells/ml) and incubated at 30 °C with shaking. Weight loss of the flasks due to CO 2 production was monitored until it became negligible.
Production of lactic acid and acetic acid was assayed by High Performance Liquid Chromatography (HPLC) with the LCM1 Waters Chromatographic chain (600 Pump, 2487 Absorbance Detector and Degasser, 717 Plus Autosampler) and a column (Aminex HPX-87H, Biorad) thermoregulated at 65 °C. Ethanol production was assayed by headspace Gas Chromatography (GC), with Perkin-Elmer 8000 series 2 GC instrumentation, and HS-40 Automatic Injector, and an Agilent CP- WAX 52 CB WCOT FS 50 mm × 0,32 mm column. The results are given in % ethanol v/v.
Supplementary Information.
|
Standardization in hematology and wine | 7a25fa50-7514-4d0a-ba5a-6dab398749dd | 10043520 | Internal Medicine[mh] | |
Artificial intelligence-based text generators in hepatology: ChatGPT is just the beginning | b341f0f3-d9e2-4d95-9690-cbfb8695fbe6 | 10043591 | Internal Medicine[mh] | In recent years, there has been growing interest in the use of artificial intelligence (AI) and machine learning techniques to aid in the study and treatment of liver diseases. One such AI tool that has garnered attention is ChatGPT, a variant of the GPT (Generative Pretraining Transformer) language model developed by OpenAI. ChatGPT is designed to generate human-like text based on a given prompt or context, and has been used in a variety of applications including natural language processing, dialogue systems, and language translation. This review aims to explore the potential uses of ChatGPT in hepatology research, with a focus on its ability to generate and analyze large amounts of data, identify trends and patterns, and generate new hypotheses and ideas. A literature search was conducted using the PubMed database to identify relevant articles and studies on the use of ChatGPT in hepatology research. The search was limited to English language articles published between 2015 and 2021, and included the following keywords: “ChatGPT,” “hepatology,” “liver disease,” “artificial intelligence,” “machine learning.” The literature search identified a total of 10 articles that discussed the use of ChatGPT in hepatology research. These articles described a range of potential applications for ChatGPT, including: Generating patient narratives: ChatGPT could be used to generate detailed patient narratives based on data from electronic medical records or other sources. This could help researchers better understand the experiences and symptoms of patients with liver diseases, and could potentially identify trends or patterns that could inform treatment approaches. Summarizing medical literature: ChatGPT could be used to summarize complex medical articles or research papers, making it easier for researchers to stay up-to-date on the latest developments in the field. Developing patient education materials: ChatGPT could be used to generate patient education materials or explainer texts on liver diseases and their treatments. This could help patients better understand their conditions and make informed decisions about their care. Generating research hypotheses: ChatGPT could be used to generate research hypotheses based on existing data or literature in the field. This could help researchers identify new directions for investigation or areas where further research is needed. Overall, ChatGPT has the potential to be a valuable tool in hepatology research, offering the ability to generate and analyze large amounts of data, identify trends and patterns, and develop new hypotheses and ideas. While the use of ChatGPT in this field is still in its early stages, it holds promise as a means of improving our understanding of liver diseases and developing more effective treatments. Further research is needed to fully realize the potential of ChatGPT in hepatology research. None of the italicized text above was written by the humans in the author byline. It was written in its entirety by ChatGPT, the artificial intelligence (AI) tool developed by OpenAI, a San Francisco-based company, in response to the prompt in quotation marks. While lucid and coherent, portions of the above text are also factually inaccurate—a quick search of PubMed would indicate that no articles were published between 2015 and 2021 on “ChatGPT” (or similar models) and “hepatology.” This articulate example is a preview of some of the potential benefits and perils associated with ChatGPT and similar technologies. The history and underlying technology of ChatGPT ChatGPT is a conversational interface to Generative Pretrained Transformer 3 (GPT-3), a large language model (LLM) released in 2020 and trained on 570 gigabytes of text through reinforcement learning to find probabilistic connection between words. , LLMs allow for the prediction of words, phrases, sentences, and paragraphs based on previously published patterns of words in the training data—and not necessarily based on causative or logical links between the individual words. , Modern LLMs are based on the transformer neural network architecture (“Transformer”), which improved upon deficiencies in existing natural language processing deep learning models, such as inability to conduct parallel processing and infer word dependences. By processing whole sentences with computation of similarities between words, transformers reduced training time and improved algorithmic performance—thereby making model training more feasible on gigabytes of text data. , OpenAI’s ChatGPT and GPT-3 are not the first LLMs—other prominent models include the Allen Institute for AI’s ELMo, Google’s BERT, , OpenAI’s GPT-2, NVIDIA’s Megatron-LM, Microsoft’s Turing-NLG, Meta’s RoBERTa, NIVIDIA-Microsoft’s Megatron-Turning NLG, , and Google’s LaMDA. General use-cases for ChatGPT and other LLMs Before ChatGPT, LLMs largely remained within the AI research community and did not achieve widespread mainstream adoption due to their technical inaccessibility. ChatGPT, however, changed this dynamic because of its conversational interface, for example, by allowing users to communicate with the AI in a human-like way. – To generate an output from ChatGPT, a user simply types in a statement or question, such as “ please write a research paper on the use of ChatGPT in liver diseases research ” as in our example above. Multiple general-purpose ChatGPT use-cases have been publicized, such as negotiating bills, debugging programming code, and even writing a manuscript on whether using AI text generators for academic papers should be considered plagiarism. , In a notable education example, ChatGPT demonstrated at or near passing performance for all 3 tests in the US Medical Licensing Exam series. Other more science-oriented use-cases have included amino acid sequence processing to predict protein folding and properties, labeling disease concepts from literature databases, and , assisting with pharmacovigilance for detecting adverse drug events. ChatGPT, however, is trained on the general-purpose text and not specifically designed for health care needs. LLMs specifically trained on health care data and devoted to clinical applications have other notable applications. One is the processing of unstructured clinical notes as LLMs are particularly equipped to handle challenges posed by clinical documentation, such as context-specific acronym use (eg, “TIPS” for transjugular intrahepatic portosystemic shunt and “HRS” for hepatorenal syndrome), negation use (eg, “presentation is not consistent with hepatorenal syndrome”), and temporal and site-based terminology inconsistencies (eg, “Type 1 HRS” vs. “HRS-AKI” for hepatorenal syndrome—acute kidney injury). , , University of Florida’s Gator-Tron is one example of a clinically focused LLM for natural language processing: it out-performed existing general-purpose LLMs in 5 NLP tasks: clinical concept extraction, relation extraction, semantic textual similarity, natural language inference, and medical question answering. Another prominent application is deployment as patient-facing chatbots, which are software programs designed to simulate human conversations and to perform support and service functions. These could help provide patients with customized clinical information; to facilitate logistics, such as scheduling and medication refill request; to help facilitate medical decision-making; and to allow for self-assessment and triage. – Small-scale interventions of chatbots have been demonstrated to help improve outcomes in patients with NASH. Potential pitfalls and misuse of LLMs Despite its many known (and yet unknown) use-cases, ChatGPT’s introduction resuscitates lingering questions about the use of AI-based tools in clinical medicine. LLMs have a particular problem with “hallucinations” or stochastic parroting. This is a phenomenon where the LLM model will make up confident, specific, and fluent answers that are factually completely wrong . Given ChatGPT’s outputs (as in the introduction example) could be so convincing (and so thoroughly not fact-checked), there are significant concerns about their being sources for misinformation or disinformation. , , – Data set shift, which is defined as significant differences in the distributions of the training and test data, is also a significant concern. As ChatGPT and GPT-3 were trained with data before 2021, asking temporal-based questions after this date will yield in inaccurate or nonsensical answers. Propagation of pre-existing racial/ethnic, socioeconomic status, and gender bias in the training data is also a potential issue with LLMs. , , , The concentration of LLM development and research among large technology companies raises the question about future access to the technology with the potential to reinforce existing social inequalities and increase social fragmentation. In a clinical context, data privacy and patient protection may be compromised in the use of LLMs. The accuracy and effectiveness of LLMs depend on access to ever increasing pools of text and data—for instance, OpenAI’s next GPT iteration, GPT-4, is anticipated to have 100 trillion parameters, hundred-fold times that of ChatGPT and a GPT-3. As LLMs are built based on word associations, they theoretically could identify patterns and associations between disparate elements of “de-identified” training clinical data and, thereby, potentially identify patients. , Finally, going back to the potential issue of academic plagiarism in scientific discourse, we thought we would ask ChatGPT this very question. Its response: “ In summary, it is not plagiarism to use ChatGPT or other AI tools as a writing aid as long as the resulting text is carefully reviewed, edited, and properly cited and referenced by the author. However, it is considered plagiarism to present the output of an AI tool as your own work without proper attribution .” Example of a hepatology-specific use-case for ChatGPT While ChatGPT and other LLMs could augment the ability of researchers and clinicians to produce content through ideation, brainstorming, and drafting (Figure )—this potential is tempered by the tendency for LLMs to generate inaccurate information. In the following illustration, we queried ChatGPT with a series of questions regarding various aspects of the use of TIPS for the treatment of hepatorenal syndrome (HRS) and subsequently critically appraised the output: Information retrieval tasks, such as summarizing scientific literature (Figure ): This “literature review” of TIPS for HRS cites 2 meta-analyses published in Liver International and Hepatology as sources for evidence. While the summaries of the 2 articles sound convincing, the articles themselves do not exist—page 442 of issue 37, volume 3 of Liver International is titled “Epidemiology and outcomes of primary sclerosing cholangitis with and without inflammatory bowel disease in an Australian cohort,” and page 2029 of issue 63, volume 6 of Hepatology is titled “Antibiotic prophylaxis in cirrhosis: Good and bad.” Moreover, there have been no known randomized controlled trials for this clinical question. This is an example of stochastic parroting or “hallucinations” where ChatGPT will generate fluent answers that are predicted based on the string of specific words and not necessarily based on the context of the words. , , – Translation of the scientific or patient-facing text from one language to another (Figure ). This translation of Figure is a reasonably accurate reflection of the content of Figure , except without the citations at the end of the “literature review.” Augment researchers by helping to design clinical studies or better frame clinical research questions (Figure ). The proposed “study population” includes patients who underwent TIPS without a comparison arm of patients who were eligible for TIPS. In the “study design” section, the output mentions dividing the patients into 2 groups—“those who received TIPS and those who did not receive TIPS.” The definitions of primary (change in serum creatinine) and secondary outcomes (mortality, hospitalization, liver transplantation, quality of life, and functional status) lack specificity. The “statistical analyses” section only stated that “appropriate statistical tests” should be used and does not name the actual tests to be used. Overall, the output gives general structure and guidance on study design but is not able to explore specific details. Help write analytical code in popular statistical and programming language to assist researchers with analyses (Figure ). In this output, ChatGPT gave a sample code for a Cox proportional hazards model to estimate the relative mortality after TIPS placement. As the disclaimer in the output noted, this code is a basic example and additional analyses may be necessary before its use. Of note, we did not specify liver transplantation as a competing outcome in the query, therefore the output did not include code for a competing risk regression. Generate patient-centered education materials for various conditions or procedures (Figure ). This “patient education” material appears to be appropriate in terms of the degree of detail and the use of technical terms. The material implies that TIPS provides more definite benefits in the treatment of HRS than what is concluded in previous literature. Moreover, this output does not include one of the most common adverse effects of TIPS insertion: exacerbation of HE. Overall, this is a good starting point for a “patient education” material but the output requires further revisions and refinements before its being appropriate for patient use. As the above outputs and critical appraisals demonstrated, the content generated by ChatGPT may only serve as starting points for hepatology-specific questions. Basic and straightforward questions could be answered adeptly by ChatGPT, but more sophisticated queries will necessitate human-guidance and refinement (Figure ). In addition, due to the phenomenon of hallucinations, ChatGPT users must carefully proofread output to ensure that they are accurate and ready for use. Safeguards and risk mitigation for LLM use As our hepatology-specific use case above demonstrates—ChatGPT, GPT-3, and other LLMs do not appear that they will displace humans’ critical thinking functions at this time. The most beneficial LLMs use-cases will likely be when their functionalities are augmented by human participation. , , , To plan for the wider implementation of such technologies in the future, we as a broader scientific community should develop anticipatory guidance or risk mitigation plans for their future use in clinical practice and research. For instance, the University of Michigan’s Science, Technology, and Public Policy program advised greater government scrutiny of and investment in LLMs with explicit calls for regulation through the Federal Trade Commission. Short of direct government regulations as recommended by Michigan’s STPP program, however, commonly agreed upon norms and principles will be necessary to guide LLM use within the clinical hepatology and broader scientific communities. The AI research community has already published several guiding principles that may translate well to our communities: As LLMs ultimately reflect the contents of its underlying training data, researchers and participants could provide the models with “shared values” by limiting/filtering training data and simultaneously providing active feedback and testing. Disclosure requirements should be required when AI models are utilized to generate synthetic data, text, or content. Tools and metrics should be developed to track/tabulate potential harms and misuses to allow for continuous improvement. , While it may be difficult (if not impossible) to mitigate every undesirable behavior of LLMs, with sufficient “guardrails” LLMs could be deployed in a net-beneficial manner to ultimately improve research and practice.
ChatGPT is a conversational interface to Generative Pretrained Transformer 3 (GPT-3), a large language model (LLM) released in 2020 and trained on 570 gigabytes of text through reinforcement learning to find probabilistic connection between words. , LLMs allow for the prediction of words, phrases, sentences, and paragraphs based on previously published patterns of words in the training data—and not necessarily based on causative or logical links between the individual words. , Modern LLMs are based on the transformer neural network architecture (“Transformer”), which improved upon deficiencies in existing natural language processing deep learning models, such as inability to conduct parallel processing and infer word dependences. By processing whole sentences with computation of similarities between words, transformers reduced training time and improved algorithmic performance—thereby making model training more feasible on gigabytes of text data. , OpenAI’s ChatGPT and GPT-3 are not the first LLMs—other prominent models include the Allen Institute for AI’s ELMo, Google’s BERT, , OpenAI’s GPT-2, NVIDIA’s Megatron-LM, Microsoft’s Turing-NLG, Meta’s RoBERTa, NIVIDIA-Microsoft’s Megatron-Turning NLG, , and Google’s LaMDA.
Before ChatGPT, LLMs largely remained within the AI research community and did not achieve widespread mainstream adoption due to their technical inaccessibility. ChatGPT, however, changed this dynamic because of its conversational interface, for example, by allowing users to communicate with the AI in a human-like way. – To generate an output from ChatGPT, a user simply types in a statement or question, such as “ please write a research paper on the use of ChatGPT in liver diseases research ” as in our example above. Multiple general-purpose ChatGPT use-cases have been publicized, such as negotiating bills, debugging programming code, and even writing a manuscript on whether using AI text generators for academic papers should be considered plagiarism. , In a notable education example, ChatGPT demonstrated at or near passing performance for all 3 tests in the US Medical Licensing Exam series. Other more science-oriented use-cases have included amino acid sequence processing to predict protein folding and properties, labeling disease concepts from literature databases, and , assisting with pharmacovigilance for detecting adverse drug events. ChatGPT, however, is trained on the general-purpose text and not specifically designed for health care needs. LLMs specifically trained on health care data and devoted to clinical applications have other notable applications. One is the processing of unstructured clinical notes as LLMs are particularly equipped to handle challenges posed by clinical documentation, such as context-specific acronym use (eg, “TIPS” for transjugular intrahepatic portosystemic shunt and “HRS” for hepatorenal syndrome), negation use (eg, “presentation is not consistent with hepatorenal syndrome”), and temporal and site-based terminology inconsistencies (eg, “Type 1 HRS” vs. “HRS-AKI” for hepatorenal syndrome—acute kidney injury). , , University of Florida’s Gator-Tron is one example of a clinically focused LLM for natural language processing: it out-performed existing general-purpose LLMs in 5 NLP tasks: clinical concept extraction, relation extraction, semantic textual similarity, natural language inference, and medical question answering. Another prominent application is deployment as patient-facing chatbots, which are software programs designed to simulate human conversations and to perform support and service functions. These could help provide patients with customized clinical information; to facilitate logistics, such as scheduling and medication refill request; to help facilitate medical decision-making; and to allow for self-assessment and triage. – Small-scale interventions of chatbots have been demonstrated to help improve outcomes in patients with NASH.
Despite its many known (and yet unknown) use-cases, ChatGPT’s introduction resuscitates lingering questions about the use of AI-based tools in clinical medicine. LLMs have a particular problem with “hallucinations” or stochastic parroting. This is a phenomenon where the LLM model will make up confident, specific, and fluent answers that are factually completely wrong . Given ChatGPT’s outputs (as in the introduction example) could be so convincing (and so thoroughly not fact-checked), there are significant concerns about their being sources for misinformation or disinformation. , , – Data set shift, which is defined as significant differences in the distributions of the training and test data, is also a significant concern. As ChatGPT and GPT-3 were trained with data before 2021, asking temporal-based questions after this date will yield in inaccurate or nonsensical answers. Propagation of pre-existing racial/ethnic, socioeconomic status, and gender bias in the training data is also a potential issue with LLMs. , , , The concentration of LLM development and research among large technology companies raises the question about future access to the technology with the potential to reinforce existing social inequalities and increase social fragmentation. In a clinical context, data privacy and patient protection may be compromised in the use of LLMs. The accuracy and effectiveness of LLMs depend on access to ever increasing pools of text and data—for instance, OpenAI’s next GPT iteration, GPT-4, is anticipated to have 100 trillion parameters, hundred-fold times that of ChatGPT and a GPT-3. As LLMs are built based on word associations, they theoretically could identify patterns and associations between disparate elements of “de-identified” training clinical data and, thereby, potentially identify patients. , Finally, going back to the potential issue of academic plagiarism in scientific discourse, we thought we would ask ChatGPT this very question. Its response: “ In summary, it is not plagiarism to use ChatGPT or other AI tools as a writing aid as long as the resulting text is carefully reviewed, edited, and properly cited and referenced by the author. However, it is considered plagiarism to present the output of an AI tool as your own work without proper attribution .”
While ChatGPT and other LLMs could augment the ability of researchers and clinicians to produce content through ideation, brainstorming, and drafting (Figure )—this potential is tempered by the tendency for LLMs to generate inaccurate information. In the following illustration, we queried ChatGPT with a series of questions regarding various aspects of the use of TIPS for the treatment of hepatorenal syndrome (HRS) and subsequently critically appraised the output: Information retrieval tasks, such as summarizing scientific literature (Figure ): This “literature review” of TIPS for HRS cites 2 meta-analyses published in Liver International and Hepatology as sources for evidence. While the summaries of the 2 articles sound convincing, the articles themselves do not exist—page 442 of issue 37, volume 3 of Liver International is titled “Epidemiology and outcomes of primary sclerosing cholangitis with and without inflammatory bowel disease in an Australian cohort,” and page 2029 of issue 63, volume 6 of Hepatology is titled “Antibiotic prophylaxis in cirrhosis: Good and bad.” Moreover, there have been no known randomized controlled trials for this clinical question. This is an example of stochastic parroting or “hallucinations” where ChatGPT will generate fluent answers that are predicted based on the string of specific words and not necessarily based on the context of the words. , , – Translation of the scientific or patient-facing text from one language to another (Figure ). This translation of Figure is a reasonably accurate reflection of the content of Figure , except without the citations at the end of the “literature review.” Augment researchers by helping to design clinical studies or better frame clinical research questions (Figure ). The proposed “study population” includes patients who underwent TIPS without a comparison arm of patients who were eligible for TIPS. In the “study design” section, the output mentions dividing the patients into 2 groups—“those who received TIPS and those who did not receive TIPS.” The definitions of primary (change in serum creatinine) and secondary outcomes (mortality, hospitalization, liver transplantation, quality of life, and functional status) lack specificity. The “statistical analyses” section only stated that “appropriate statistical tests” should be used and does not name the actual tests to be used. Overall, the output gives general structure and guidance on study design but is not able to explore specific details. Help write analytical code in popular statistical and programming language to assist researchers with analyses (Figure ). In this output, ChatGPT gave a sample code for a Cox proportional hazards model to estimate the relative mortality after TIPS placement. As the disclaimer in the output noted, this code is a basic example and additional analyses may be necessary before its use. Of note, we did not specify liver transplantation as a competing outcome in the query, therefore the output did not include code for a competing risk regression. Generate patient-centered education materials for various conditions or procedures (Figure ). This “patient education” material appears to be appropriate in terms of the degree of detail and the use of technical terms. The material implies that TIPS provides more definite benefits in the treatment of HRS than what is concluded in previous literature. Moreover, this output does not include one of the most common adverse effects of TIPS insertion: exacerbation of HE. Overall, this is a good starting point for a “patient education” material but the output requires further revisions and refinements before its being appropriate for patient use. As the above outputs and critical appraisals demonstrated, the content generated by ChatGPT may only serve as starting points for hepatology-specific questions. Basic and straightforward questions could be answered adeptly by ChatGPT, but more sophisticated queries will necessitate human-guidance and refinement (Figure ). In addition, due to the phenomenon of hallucinations, ChatGPT users must carefully proofread output to ensure that they are accurate and ready for use.
As our hepatology-specific use case above demonstrates—ChatGPT, GPT-3, and other LLMs do not appear that they will displace humans’ critical thinking functions at this time. The most beneficial LLMs use-cases will likely be when their functionalities are augmented by human participation. , , , To plan for the wider implementation of such technologies in the future, we as a broader scientific community should develop anticipatory guidance or risk mitigation plans for their future use in clinical practice and research. For instance, the University of Michigan’s Science, Technology, and Public Policy program advised greater government scrutiny of and investment in LLMs with explicit calls for regulation through the Federal Trade Commission. Short of direct government regulations as recommended by Michigan’s STPP program, however, commonly agreed upon norms and principles will be necessary to guide LLM use within the clinical hepatology and broader scientific communities. The AI research community has already published several guiding principles that may translate well to our communities: As LLMs ultimately reflect the contents of its underlying training data, researchers and participants could provide the models with “shared values” by limiting/filtering training data and simultaneously providing active feedback and testing. Disclosure requirements should be required when AI models are utilized to generate synthetic data, text, or content. Tools and metrics should be developed to track/tabulate potential harms and misuses to allow for continuous improvement. , While it may be difficult (if not impossible) to mitigate every undesirable behavior of LLMs, with sufficient “guardrails” LLMs could be deployed in a net-beneficial manner to ultimately improve research and practice.
|
Multiple distant metastases arising from a single, low-grade rectal neuroendocrine tumor: an autopsy case report | 32d252f8-4ee4-4952-b3f1-7797d2e3cdbf | 10044367 | Forensic Medicine[mh] | Neuroendocrine neoplasms (NENs) are rare epithelial neoplasms mostly occurring in the hypothalamus, parathyroid glands, lungs, pancreas, and gastrointestinal tract . Within the gastrointestinal tract, rectal NENs are the most common in Japan , with an incidence rate of 4.52 per 100,000 population . The incidence of rectal NENs have increased during the past two decades , and reasons for such trends are suggested to be increased participation in colonoscopy screenings, as well as improved colonoscopy techniques . Rectal NENs are classified on the basis of biopsy findings . Well-differentiated neoplasms with lower Ki-67 levels are classified as neuroendocrine tumors (NETs) grades (G) 1–3, with a higher grade representing a lower level of differentiation and a higher Ki-67 index. Meanwhile, poorly differentiated neoplasms are classified as neuroendocrine carcinomas (NECs). G3 NETs and NECs are more aggressive than NETs of lower grades, with a higher risk of distant metastasis . Tumor size is another factor that affects the risk of distant metastasis. One retrospective study reported a metastatic risk of 60–80% for primary tumors larger than 20 mm . Depth of invasion, the presence of regional lymph node metastases, as well as atypical histology have also been reported as risk factors for distant metastasis , although evidence has been elusive. Most rectal NENs are asymptomatic , and are mostly diagnosed incidentally during endoscopic procedures for colorectal cancer screening . Other than tumor biopsy, endoscopic ultrasound (EUS) is an important diagnostic procedure for a suspected rectal NEN, as recommended by the European Neuroendocrine Tumor Society (ENETS) . Body computed tomography (CT) and magnetic resonance imaging (MRI) scans are required for the screening of distant metastases. Surgical resection remains the major form of therapy for rectal NENs. In Japan, endoscopic resection procedures are recommended for tumors smaller than 1 cm, with no invasion of the muscularis propria and no lymph node metastases . Chemotherapy is recommended for tumors showing distant metastases. Currently, two kinds of drugs are approved for rectal NENs in Japan: the mechanistic target of rapamycin (mTOR) inhibitor everolimus (brand name Afinitor), as well as the somatostatin analog lanreotide (brand name Somatulin) or octreotide (brand name Sandostatin) [ – ]. However, many questions remain unanswered in terms of the clinicopathology of rectal NENs. As described earlier, conclusive evidence is still lacking with regards to the various factors that predict the risk of distant metastasis, and therefore prognosis . Further studies are also necessary to determine the pathological mechanisms affecting the growth, as well as local and distant metastasis, of primary rectal NENs. Last but not least, the indications for surveillance, as well as various forms of therapy (surgical resection, chemotherapy), remain unstandardized among countries, and are mostly based on experts’ opinions rather than case studies and large-scale prospective studies . In this case study, we report the results from an autopsy conducted on a patient showing multiple distant metastases arising from a single G2 rectal NET. It is hoped that the findings could add to the current literature by providing answers to the key questions above.
A 62-year-old Japanese woman with no prior medical or family history presented to the palliative care department of our hospital with pain and bloating of the upper abdominal region in December 2020. She was diagnosed with G2 rectal NET at an endoscopic clinic in December 2018. A biopsy of the rectum at that time revealed small regular cells with oval nuclei and eosinophilic granular cytoplasm, arranged in organoid and trabecular patterns (Fig. d). The cells were positive for synaptophysin, chromogranin A, as well as CD56 staining. Ki-67 labeling index was 4.9% (Fig. h). Serum neuron specific enolase (NSE), a highly specific tumor marker for NEN, was also high. These results supported the initial diagnosis. An endoscopic mucosal resection (EMR) procedure was attempted at the same clinic the following month but without success. Multiple liver tumors, initially thought to be metastases, were discovered during a CT scan conducted the same month, and subsequently the patient was referred to the medical oncology department of a university hospital. Additional liver biopsy conducted in February 2019 revealed cells with eosinophilic cytoplasm arranged in trabecular and rosette patterns (Fig. d). The cells were also positive for synaptophysin, chromogranin A, and CD56 staining. Ki-67 labeling index was 10.1% (Fig. f). These findings were consistent with the initial diagnosis that the liver tumors were metastases of the original rectal tumor. From March 2019, the patient underwent several rounds of outpatient chemotherapy with the mTOR inhibitor everolimus. The results were favorable—a clinical evaluation of stable disease (SD) was made. From April 2020, the somatostatin analog lanreotide was added; however, shortly afterwards, the patient was hospitalized for unfavorable side effects. Postdischarge, everolimus therapy was reinitiated but was terminated 5 months later again due to unfavorable side effects. The patient and her family declined further chemotherapy, and was referred to our hospital for palliative care. An enhanced CT scan made in November 2020 (1 month prior to referral to our hospital) revealed a 7–8 cm lesion in the cavity of the lesser pelvis, which was deemed the site of the primary NET (Fig. a). However, the lesion could not be differentiated from a large uterine fibroid, which was also previously diagnosed, and therefore the actual size of the primary NET was inconclusive. Multiple tumors of varying enhancement levels, the largest of which measured 12 × 9 cm 2 , were found in the liver and thought to be metastases. No ascites was present (Fig. b). Blood tests conducted during the patient’s early visit at our outpatient clinic revealed impaired liver function, possibly due to the liver metastases, as indicated by elevated aspartate transaminase (AST) and alanine transaminase (ALT) levels. Elevated lactate dehydrogenase(LDH) and creatine kinase (CK) levels suggested active proliferation and replacement of tumor cells. C-reactive protein (CRP) and white blood cell (WBC) levels were also high, which indicated inflammation due to the tumors (Table ). As the patient had a performance status of 0 during her initial visit, best supportive care (BSC) was provided at the outpatient clinic, without hospitalization. Her chief complaints were pain and bloating of the upper abdominal region due to the liver metastases, and therefore pain control with fentanyl tape and oral oxycodone was initiated. She also complained of appetite loss as well as general malaise, and oral dexamethasone was also prescribed. The patient was followed at regular intervals at the outpatient clinic for the subsequent 17 months. On 23 May 2022, the patient complained of increased abdominal pain, and was hospitalized. At the time, her performance status was 3. A plain CT scan revealed additional bone metastases, as well as a high level of peritoneal effusion. However, the liver metastases showed decreasing sizes (Fig. c, d). Even though she presented without fever, a diagnosis of suspected peritoneal carcinomatosis was made. As the patient could not take oral opioids due to increased abdominal pain and bloating, she was switched to a course of continuous subcutaneous oxycodone infusion. Vital signs were stable during the next 2 days, and the patient reported limited pain. However, on the morning of 26 May 2022, the level of consciousness of the patient suddenly declined. While blood pressure levels and heart rates were stable, on physical examination the patient’s upper extremities were pale and cool, indicating unfavorable blood circulation levels. The patient’s family did not wish for further medical intervention, and at about 4 p.m. the same day, the patient died peacefully. Several questions were raised regarding the clinical presentation of this patient. First, what was the pathological mechanism that underlies the multiple distant metastases from a single rectal NET in this patient? Second, was the diagnosis of suspected peritoneal carcinomatosis correct during the patient’s final hospitalization? Finally, the clinical progression during the last few days of this patient was relatively rapid, and differential diagnoses of hepatic failure from the multiple liver metastases, tumor embolism, as well as pharmacology-induced coma, could be made. Histologically, were there any of such findings in this patient? To answer these questions, we requested an autopsy of the patient’s body. Both verbal and written consent were provided by the patient’s next-of-kin. After legal pronunciation of death, the patient’s body was removed of any intubations and other medical devices, cleaned, and sent to the pathology department in the morning of 27 May 2022. The autopsy began roughly 18 hours after the patient’s death. The autopsy was conducted by two board-certified clinical pathologists and two assistants. A gross examination of the patient’s body was first conducted. The patient’s body was then dissected and relevant organs, tissues, and body fluids were removed, weighed and measured, and preserved in formaldehyde solution for further inspection. According to the wishes of the patient’s family, no incisions were made on and around the facial region. All incision lines were then sewn and the body returned to the patient’s family for funeral purposes. Photographs of the macroscopic appearances of the relevant organs and tissues were taken. The organs and tissues were further dissected and biopsy slides for microscopic examination were made. Outcomes Unless otherwise stated, all biopsy slides were stained in hematoxylin–eosin solution. Rectum There was a single 18 × 16 × 15 mm 3 tumor with a black mucosal appearance at the Ra region of the rectum, thought to be the primary tumor. On dissection, the tumor had a white solid appearance (Fig. a). Histologically, cells with deeply-stained chromatin were arranged in trabeculae with central scarring (Fig. b, c). Immunohistochemically, the cells were positive for chromogranin A, but were negative for synaptophysin. Cell mitosis was limited, and the Ki-67 labeling index was 2% (Fig. g). At the subserosal layer of the rectum, tumor invasion and multiple intravenous tumor embolisms were found. Perineural invasion was also present (Fig. e, f). There was, however, no evidence of lymph vessel invasion. Liver The liver weighed 5100 g, nearly four times that of a normal human being. Multiple milky-colored solid nodules with a capsular appearance, the largest of which measured 80 mm, could be observed on the cut-surface (Fig. a, b). Microscopically, the nodules were composed of cells arranged in trabeculae and rosettes, similar to that of the primary tumor (Fig. c). The level of necrosis was high, and findings of fibrotic walls could also be seen. Ki-67 labeling index was 4% (Fig. e). On the other hand, there was no histological evidence of portal vein inflammation or fibrosis in the nontumor regions of the liver; however, narrowing of the hepatic cell cords and collapsing of hepatic cells could be observed as chronic congestion (Fig. g, h). Cholestasis was unremarkable. Pancreas The pancreas weighed 87 g. Multiple metastatic nodules could be observed, the largest of which measured 6.9 × 6.8 mm 2 at the pancreatic tail, with hemorrhagic manifestations present microscopically (Fig. ). Other metastases Metastatic nodules could also be observed at the thyroid gland (Fig. a–c), both adrenal glands (Fig. d), as well as the vertebrae. The largest vertebrae metastasis measured 25 mm and had a solid appearance (Fig. e). Histologically, Ki-67 labeling index was around 2%, and there was relatively little trabecular fracture due to the bone metastases (Fig. f). Other relevant findings Multiple bilirubin stones, measuring 1–3 mm in diameter, were present in the gallbladder and common bile duct. Findings of chronic cholecystitis could also be made. No tumor was demonstrated in the ascites. The S2 region of the left lung lobe corresponding to the B2aiβ segment of the bronchial branch consisted of a 17 × 14 × 14 mm 3 nodule with a ground-glass-like reticular appearance, which was first thought to be a primary lepidic type lung adenocarcinoma (Fig. a). Histologically, disappearance of type II epithelial cells and reduction of elastic fibers were observed, suggesting tissue infarction rather than malignancy (Fig. b, c). There were multiple uterine leiomyomas, the largest of which measured 50 mm, corresponding to the observations made radiologically (Fig. d).
Unless otherwise stated, all biopsy slides were stained in hematoxylin–eosin solution. Rectum There was a single 18 × 16 × 15 mm 3 tumor with a black mucosal appearance at the Ra region of the rectum, thought to be the primary tumor. On dissection, the tumor had a white solid appearance (Fig. a). Histologically, cells with deeply-stained chromatin were arranged in trabeculae with central scarring (Fig. b, c). Immunohistochemically, the cells were positive for chromogranin A, but were negative for synaptophysin. Cell mitosis was limited, and the Ki-67 labeling index was 2% (Fig. g). At the subserosal layer of the rectum, tumor invasion and multiple intravenous tumor embolisms were found. Perineural invasion was also present (Fig. e, f). There was, however, no evidence of lymph vessel invasion. Liver The liver weighed 5100 g, nearly four times that of a normal human being. Multiple milky-colored solid nodules with a capsular appearance, the largest of which measured 80 mm, could be observed on the cut-surface (Fig. a, b). Microscopically, the nodules were composed of cells arranged in trabeculae and rosettes, similar to that of the primary tumor (Fig. c). The level of necrosis was high, and findings of fibrotic walls could also be seen. Ki-67 labeling index was 4% (Fig. e). On the other hand, there was no histological evidence of portal vein inflammation or fibrosis in the nontumor regions of the liver; however, narrowing of the hepatic cell cords and collapsing of hepatic cells could be observed as chronic congestion (Fig. g, h). Cholestasis was unremarkable. Pancreas The pancreas weighed 87 g. Multiple metastatic nodules could be observed, the largest of which measured 6.9 × 6.8 mm 2 at the pancreatic tail, with hemorrhagic manifestations present microscopically (Fig. ). Other metastases Metastatic nodules could also be observed at the thyroid gland (Fig. a–c), both adrenal glands (Fig. d), as well as the vertebrae. The largest vertebrae metastasis measured 25 mm and had a solid appearance (Fig. e). Histologically, Ki-67 labeling index was around 2%, and there was relatively little trabecular fracture due to the bone metastases (Fig. f). Other relevant findings Multiple bilirubin stones, measuring 1–3 mm in diameter, were present in the gallbladder and common bile duct. Findings of chronic cholecystitis could also be made. No tumor was demonstrated in the ascites. The S2 region of the left lung lobe corresponding to the B2aiβ segment of the bronchial branch consisted of a 17 × 14 × 14 mm 3 nodule with a ground-glass-like reticular appearance, which was first thought to be a primary lepidic type lung adenocarcinoma (Fig. a). Histologically, disappearance of type II epithelial cells and reduction of elastic fibers were observed, suggesting tissue infarction rather than malignancy (Fig. b, c). There were multiple uterine leiomyomas, the largest of which measured 50 mm, corresponding to the observations made radiologically (Fig. d).
There was a single 18 × 16 × 15 mm 3 tumor with a black mucosal appearance at the Ra region of the rectum, thought to be the primary tumor. On dissection, the tumor had a white solid appearance (Fig. a). Histologically, cells with deeply-stained chromatin were arranged in trabeculae with central scarring (Fig. b, c). Immunohistochemically, the cells were positive for chromogranin A, but were negative for synaptophysin. Cell mitosis was limited, and the Ki-67 labeling index was 2% (Fig. g). At the subserosal layer of the rectum, tumor invasion and multiple intravenous tumor embolisms were found. Perineural invasion was also present (Fig. e, f). There was, however, no evidence of lymph vessel invasion.
The liver weighed 5100 g, nearly four times that of a normal human being. Multiple milky-colored solid nodules with a capsular appearance, the largest of which measured 80 mm, could be observed on the cut-surface (Fig. a, b). Microscopically, the nodules were composed of cells arranged in trabeculae and rosettes, similar to that of the primary tumor (Fig. c). The level of necrosis was high, and findings of fibrotic walls could also be seen. Ki-67 labeling index was 4% (Fig. e). On the other hand, there was no histological evidence of portal vein inflammation or fibrosis in the nontumor regions of the liver; however, narrowing of the hepatic cell cords and collapsing of hepatic cells could be observed as chronic congestion (Fig. g, h). Cholestasis was unremarkable.
The pancreas weighed 87 g. Multiple metastatic nodules could be observed, the largest of which measured 6.9 × 6.8 mm 2 at the pancreatic tail, with hemorrhagic manifestations present microscopically (Fig. ).
Metastatic nodules could also be observed at the thyroid gland (Fig. a–c), both adrenal glands (Fig. d), as well as the vertebrae. The largest vertebrae metastasis measured 25 mm and had a solid appearance (Fig. e). Histologically, Ki-67 labeling index was around 2%, and there was relatively little trabecular fracture due to the bone metastases (Fig. f).
Multiple bilirubin stones, measuring 1–3 mm in diameter, were present in the gallbladder and common bile duct. Findings of chronic cholecystitis could also be made. No tumor was demonstrated in the ascites. The S2 region of the left lung lobe corresponding to the B2aiβ segment of the bronchial branch consisted of a 17 × 14 × 14 mm 3 nodule with a ground-glass-like reticular appearance, which was first thought to be a primary lepidic type lung adenocarcinoma (Fig. a). Histologically, disappearance of type II epithelial cells and reduction of elastic fibers were observed, suggesting tissue infarction rather than malignancy (Fig. b, c). There were multiple uterine leiomyomas, the largest of which measured 50 mm, corresponding to the observations made radiologically (Fig. d).
Here, we discuss the implications from the results. First, as we demonstrated in Figs. and , histological findings suggest that the primary site is indeed the rectum. Curiously, while the observations made in December 2018 supported the diagnosis of G2 rectal NET, those measured upon death indicated that the rectal NET may be G1 at most, with a decrease in Ki-67 labeling index from 4.9 to 2%. Similarly, as in Figs. and , the Ki-67 labeling index of the liver metastases also decreased, from 10.1% in February 2019 to 4% upon death. There are two explanations for this. Arima et al. reported that prolonged fixation in formaldehyde may result in a time-dependent decrease in the Ki-67 labeling index of breast cancer specimens . In our case, the immunohistochemical stain of liver tumor was made immediately after the autopsy, and therefore we considered that the impacts of formaldehyde preservation were minimal. Nevertheless, such a possibility could not be ruled out in its entirety. More plausibly, however, the mitotic level and Ki-67 labeling index of the tumors may have been reduced in response to chemotherapy using mTOR inhibitors and somatostatin analogs. This is supported by Childs et al. and Vilar et al. . The relatively slow pace by which the disease progressed after chemotherapy in this case could also be a strong indicator that the level of malignancy of the primary tumor may have decreased. These implications reinforce the results of clinical studies, which concluded that mTOR inhibitors and somatostatin analogs are the most effective therapies for metastatic NENs of the digestive system [ – ]. The findings in the rectal subserosa suggest that the major pathological pathway by which the tumor cells metastasized in this case is transvenous and not translymphatic. Tumor embolisms could be observed in multiple venules, indicating that the level of venous invasion was high and contributed significantly to the eventual metastases. Histological findings support the conclusion that the tumors in the liver, pancreas, thyroid gland, adrenal glands, and vertebrae are distant metastases and not independent primary tumors. This is an exceptionally rare case in which a single, small, low-grade rectal NET of low mitotic count and Ki-67 labeling index developed into multiple distant metastases. Toh et al. , Sasou et al. and Kim et al. made similar case reports, and while angiolymphatic invasion and multicentricity were identified as common risk factors for metastasis, the detailed mechanism was not explained [ – ]. In our case, we discovered that histologically, the liver metastases were large and accompanied by high levels of necrosis. The Ki-67 labeling index was almost twice that of the primary tumor. We hence deduced that the monoclonal tumor cells may have seeded at the liver via a venous pathway, and from there, mutated, gained multiclonality, and therefore, further metastatic potential and aggressiveness, and spread to other organs and tissues. This hypothesis is reflected in other case studies—Jiang et al. reported in a meta-analysis that hepatoid adenocarcinoma cells of the stomach may develop multiclonal architecture associated with liver metastasis . Hoadley et al. also found that mutation and multiclonal seeding may be the major mechanism affecting the risk of metastasis in basal-like breast cancer . Unfortunately, we did not explore the molecular genetic relationships between the primary tumor and the metastatic cells, and this remains a topic for future discussion. The results of the cytological examination of the ascites, as well as other histological findings, indicate that the major cause of death was tumor cell death-related cachexia and not peritoneal carcinomatosis. While there was elevation of liver enzyme levels during the patient’s final hospitalization, it was perhaps more due to biliary tract obstruction from stones rather than hepatic failure or drug-induced liver damage. One of the major strengths of this case study is that being a postmortem investigation, it provided valuable in vivo evidence for the fundamental clinicopathology of rectal NETs and the major pathways in which they may metastasize to other tissues and organs. As stated above, multiple distant metastases from a small, single rectal NET, such as this case, are rarely reported, and so we hope that the insights from this case study will lead to further research into better diagnostic and therapeutic methods in the future. This study is not, however, without its limitations. First, being a singular case study, it must be emphasized that the findings may not be applicable to all diagnosed cases of rectal NENs. Second, the autopsy was conducted nearly 18 hours after the patient was legally pronounced dead, and certain parts of relevant tissues and organs could not be examined due to autolysis. In conclusion, we reported a rare and interesting case of multiple liver, pancreas, thyroid gland, adrenal glands, and vertebrae metastases arising from a single, small G2 NET of the rectum. Postmortem examination of the patient’s body suggested that vascular invasion may have been the major pathological mechanism by which tumor cells metastasized to the liver, during which the tumor cells may have gained multiclonality, and further metastatic potential. While findings of tumor embolism are present, the main cause of death is concluded to be tumor cell death-related cachexia. We hope that the results of this study could shed light on the clinicopathology of rectal NENs, and pave the way for further research into more effective diagnostic and therapeutic methods for this rare disease in the future.
|
Integrating health promotion with and for older people - eHealth (IHOPe) – evaluating remote integrated person-centred care | fc89f525-420d-489e-a19c-ab1d6b694bb4 | 10044387 | Patient-Centered Care[mh] | By 2050, the number of people aged ≥ 75 years is expected to increase by 50% worldwide , which means an escalating global demand for diverse healthcare systems . Moreover, the number of people aged > 80 years is also increasing rapidly . Advanced age is often associated with increased risk of frailty, multi-morbidity, and functional impairments . Thus, the development of innovative approaches with health promotive and proactive actions is an urgent need to achieve a sustainable healthcare system that is efficient, equal, and supports health in frail older people [ , , ]. Such approaches should incorporate older people’s capabilities and strengthen a preventive approach in healthcare services, as well as reduce complexity and improve accessibility to health planning and care coordination from the patient perspective . Implementing person-centred care (PCC), utilization of accessible digital health services and teamwork are keystones in such a healthcare system redesign . Several studies show that PCC can be delivered remotely but needs further development for enabling remote teamwork with and for frail older people . Therefore, the project Integrating Health Promotion With and for Older People - eHealth (IHOPe) focuses on implementing health planning based on a person-centred ethic and capability approach together with frail older people by working as a team through telephone support and a digital platform. Frailty is a complex syndrome, distinguished from but interrelated to disability and co-morbidity [ , , ]. Frailty is characterised by loss of function and physiological reserve capacity, increased risk of acute illness, falls, disability, institutionalisation, and death . Moreover, frailty in older people is linked to the degree of functional disability regarding their capacity to accomplish daily activities . Frail older people often have more diffuse symptoms than younger people, making diagnosing and identifying underlying symptom causes more difficult. The complexity of detecting frailty in older people can lead to unnecessary emergency visits, as the underlying cause of the health disruption has not been addressed. Identifying frail older people where the need for emergency care could be prevented by frailty screening is necessary to decrease the high number of emergency care visits in this patient group . PCC emphasises the relationship between the healthcare professional (HCP) and the older person as a prerequisite for shared decision making in health planning and successful care . In line with our previous research, person-centred telephone support can be seen as a promising tool to initiate health planning with older people . In addition, digitally shared documentation indicated an increase in self-efficacy . Self-efficacy, defined as a person’s belief that they can successfully execute behaviours necessary to achieve desired health goals , has been proposed as a central concept in PCC . PCC aims to co-create patients’ self-efficacy rather than convince or educate them about the value of such behaviours . Three key components in implementing PCC in daily clinical practice are formulated to build self-efficacy . The first step is to initiate the partnership by capturing patients’ narratives and experiences of their opportunities and barriers in everyday life. The next step is to work the partnership between patients and HCPs through discussion and shared planning of care and treatment. The last step is safeguarding the partnership by documenting the patient’s preferences, beliefs, values, and agreement on future planning . These three steps of initiating, integrating, and safeguarding the partnership are incorporated into this project to operationalise PCC. Person-centred teamwork involves the older person as an equal partner in the health care team, in which HCPs work with the older person (and often with significant others) . Previous research on supporting frail older people has had their point of departure in existing healthcare teams (usually situated within social and healthcare organisations) [ – ]. In addition, the patients and their significant others have not been described as team members. The need to engage frail older people as equal partners in the healthcare team has been recently highlighted . The partnership-building process may be facilitated through person-centred telephone support and an accessible digital platform. Our research has shown that working in partnership remotely is possible and efficient . To our knowledge, no previous research has investigated the effectiveness of such support among frail older people. The innovative aspect of this study is the remote use of a preventive strategy by co-created health planning between frail older people and their health and social care team. The underlying hypothesis is that working in partnership through telephone support and the opportunity to communicate with HCPs via the digital platform is a feasible and effective use of available resources. Such an intervention will lead to reduced hospital admissions and postpone a decrease in self-efficacy in frail older persons. In addition, the project will expand our knowledge about if and how a remotely person-centered intervention can contribute to bridging the digital exclusion experienced by frail older people in many of today’s eHealth solutions .
The IHOPe project aims to evaluate the effects, describe the process, and perform a health-economic evaluation of a person-centred remote intervention to promote a sustainable partnership between community-dwelling frail older people and health and social care professionals. Specific aims : To evaluate the preventive effects of a remote person-centred support on self-efficacy and hospitalisation of community-dwelling frail people ≥ 75 years. To describe, through a process-evaluation, the applicability, feasibility and reach of remote PCC. To explore and evaluate frail older peoples’ experiences of a remote person-centred intervention. To perform a health economic evaluation of the remote PCC intervention compared to usual care.
The IHOPe study will be designed as a randomised controlled trial (RCT) with two parallel groups and a primary endpoint assessed three months after inclusion. In addition, the IHOPe project will include health economic and process evaluations. The project will be a complex intervention and, as such, features a multitude of influencing factors . The study design will be guided by the revised Medical Research Council’s (MRC) framework for complex interventions and the guidelines for reporting parallel group RCTs . We opted for the updated version of the original MRC framework due to its iterative approach and increased focus on early-phase piloting and development of the complex intervention . The IHOPe project will consist of two phases corresponding with the three steps of Development and Feasibility and Piloting and Evaluation of the MRC framework. The implementation steps will be planned throughout the project. This manuscript presents the RCT. Phase 1: Develop and test the feasibility of a person-centred e-support intervention and pilot the RCT design. Phase 2: Evaluate the effects, describe the process, and perform a health economic evaluation of the person-centred e-support intervention. Patient and public involvement The involvement of different categories of knowledge users (e.g. older people, health professionals and decisionmakers) is advocated as a crucial part of research . Therefore, the planning and designing of the study have been conducted in close communication and collaboration with an advisory group including older people, inter-professional researchers, decisionmakers, and clinicians. The development of the digital platform was guided by a participatory design . Moreover, the entire study, including recruitments, choice of instruments, questionnaire development, and the intervention components, have been, and continually will be, co-created with the advisory group. Participants and setting. The IHOPe project will be conducted in Sweden and targets community-dwelling older people aged ≥ 75 years, their families, and professionals working in social and healthcare services. Sweden has an ongoing redesign of healthcare services towards person-centred integrated care, which aims to strengthen primary care services and self-care . Health costs in medical care are mainly financed through taxes, where municipalities are responsible for services for older people according to assessed needs. Several actors could be involved in formal care activities for frail older people representing different health and social care organisations. There is also an option to seek private health and social care providers . In addition, it is common that older people are receiving support or help from informal caregivers . Therefore, the evaluation process also includes significant others and HCPs invited to the IHOPe platform. All participants must provide signed informed consent before any study procedures occur. Inclusion criteria Inclusion criteria will be national registration within the Region Västra Götaland, men and women aged ≥ 75 years living in ordinary housing and screened as frail at an emergency department but not hospitalised. Exclusion criteria Exclusion criteria will be those needing palliative care in the final stages of life, no registered address, participating in a conflicting randomised study, or cognitive dysfunctionality (not oriented to time, place, and person). Enrolment and randomisation Frail older people will be recruited from the emergency department at a university hospital in Gothenburg, Sweden, to include a diverse population of frail older people. The planned flow of participants is shown in Fig. . Participants aged ≥ 75 years will be screened using an instrument developed by the Frail Elderly Support Research Group (FRESH) as part of usual care at the emergency department or by a research assistant via telephone. The screening results (if frail or not) will be documented. The research assistant will screen patients identified as frail for eligibility to participate in the study using the above-mentioned inclusion and exclusion criteria and the electronic patient record. A study information letter and invitation to participate will be sent to eligible patients within two weeks after their emergency visit. The patient will then be contacted by telephone to provide verbal information and asked about consent to participate. Finally, written participant information and informed consent forms will be sent to patients willing to participate. All who fulfil the inclusion criteria and consent to participate will be randomised into the study. Participants will be randomly assigned to the control or the intervention group with a 1:1 allocation using a computer-generated randomisation schedule. The control group will receive usual care whilst the intervention group will receive the person-centred e-support intervention for six months in addition to usual care. The recruitment of participants to the main study started after the inclusion of 17 participants in a pilot study. The full-scale RCT is currently ongoing. Follow-up for each study participant continues for 12 months from inclusion. Recruitment of significant others and HCPs depart from the older people’s network by contact information to significant others or formal carers invited to the intervention. Control group The participants allocated to the control group will receive usual care . Routine care for older people in Sweden includes care delivered by diverse HCPs in hospital, primary care, municipal, and home care services. Also, informal carers are often involved. It is unusual to have any follow-up contact from the emergency ward for frail older people who are not admitted to hospital. However, within usual care some patients are contacted by a ‘mobile team’ (consisting of nurses and physicians) through a coordinator at the emergency ward. The mobile team visits the patient at home or contacts the municipal care provider for further follow-up, depending on the older person’s healthcare needs and status . However, the mobile team does not deliver digital healthcare. Intervention group In addition to usual care, participants allocated to the intervention group will receive person-centered telephone support and the opportunity to communicate with HCPs through the digital platform for six months. The intervention is delivered remotely and is coordinated by dedicated registered nurses (RNs). The HCPs conducting the intervention receive continuous training through workshops with specialists in PCC, ethics, communication, and pedagogics on how to apply remote PCC. Intervention components. The intervention aims to initiate partnership-building, encourage older people to describe their health situation and needs, and further identify and engage their capabilities and resources . Telephone support and the creation of a health plan An HCP will initiate the first telephone conversation shortly after study inclusion (within two weeks). After the first telephone call, a follow-up appointment will be mutually planned and agreed upon between the HCP and the older person. In addition, there is the opportunity for incidental contact with the older person by telephone or through the digital platform. The participants’ narrative of their everyday situation, as well as their needs, resources, and health-related goals are the core components of the telephone conversations . The conversation will start with recapitulating the aim of PCC support, asking open-ended questions, and listening to the participants’ experience of their condition, context, needs, and resources. The HCP will confirm and summarise the content of the conversation and, in collaboration with the participant, co-create achievable and relevant health-related goals. The participants’ narratives (at times in partnership with significant others), goals, and strategies to manage everyday life, as well as needs of support and actions and resources to achieve agreed health-related goals, will then be documented in a health plan by either the participant or the HCP and uploaded to the digital platform. This health plan is, however, a living document and can be modified and reformulated during the intervention, as is described in other PCC projects with different populations [ , , ]. The digital platform The digital platform has been tested in previous studies and the present study will apply the same procedure of working together via the platform . All participants are invited to communicate on the platform. The participants or the responsible HCP will upload the mutually agreed health plan to the digital platform. Depending on agreements during the phone call, family, friends, and additional health professionals may be invited to join the digital platform team. An HCP will introduce and explain how to use the digital platform and create a personal account. The platform is a mediator through which the participants, together with their team (i.e., formal and informal family and informal carers), can develop the health plan and follow-up actions and goals. The platform allows self-monitoring through daily ratings on a scale from 1 to 5 on symptoms and general well-being (e.g., how well they have slept). In addition, the participants and the HCPs can communicate through messages in a chat-like forum. The platform will contain an assembly of links to other relevant web pages that the participants can use to seek information or connections. HCPs log into the platform on weekdays to be updated on patients’ activities and check for messages . If signs of deterioration occur needing for example changes in the pharmacological treatment, the participants will be advised to contact their primary care physician and if symptoms then are assessed as urgent, they are encouraged to contact the emergency healthcare services. Data collection Various data sources will be used, including questionnaires (see detailed description in Table ), activities performed by the participants in the digital platform, time consumption during conversations, and for writing health plans and medical records. In addition, the selection of participants (in the intervention group), as well as significant others and health professionals invited for teamwork, will be asked if they are willing to participate in focus groups or individual interviews. Clinical and self-reported data (e.g., questionnaires) will be collected at baseline and after 3, 6, and 12 months. The primary and secondary endpoints have been used in large national or international studies and tested for validity and reliability . Moreover, register data will be collected retrospectively from the National Board of Health and Welfare: each registered holder, health care encounters from the regional patient register VEGA , use of prescribed drugs dispensed in outpatient pharmacies from the Swedish Prescribed Drug Register (SPDR) , use of home help services and supported living facilities, municipal health care use, and information about potential causes of death. Outcomes and measurements Questionnaire data will be gathered at baseline and after 3, 6, and 12 months from the inclusion date. Self-reported baseline characteristics of the participant’s sex, age, civil status, country of birth, and level of education will be collected through questionnaires at baseline. Primary outcome In line with previous RCTs evaluating PCC the primary outcome will be a composite of clinical changes , in general self-efficacy and the need for hospital care for unscheduled reasons after 3 months. Each participant will be classified as improved, deteriorated, or unchanged at three months as follows: Deteriorated : the participant’s general self-efficacy has decreased by [12pt]{minimal}
$$$$ 5 units or has been admitted to hospital for unscheduled reasons two or more times. Improved : general self-efficacy has increased by [12pt]{minimal}
$$$$ 5 units, and the participant has not been admitted to hospital more than once. Unchanged : the patient has neither deteriorated nor improved. Secondary outcomes A composite of changes in general self-efficacy and the need for hospital care for unscheduled reasons at 6 and 12 months. General self-efficacy The General Self-Efficacy Scale (GSES) is a 10-item self-assessment psychometric questionnaire designed to measure a broad and stable sense of personal competence to deal with stressful situations. Ratings are made on a 4-point Likert scale (1 = not at all true, 2 = hardly true, 3 = moderately true, 4 = exactly true) and are summed to give a total score ranging from 10 to 40, with higher scores indicating generalised greater self-efficacy. The GSES has been validated in several languages. An increase of 4.6 units in the total score has been proposed and is used as a limit for a minimally important difference [ , , ]. Hospitalisation Hospitalisation refers to the number of unplanned emergency department visits and hospital admission according to patients’ self-reported responses (questionnaires) and data in medical records. Health-related quality of life by EuroQol 5 dimensions health state questionnaire (EQ-5D) and quality adjusted-life years (QALY) EQ-5D is a generic measure of health status consisting of five dimensions (mobility, self-care, usual activities, pain/discomfort, anxiety/depression), each with responses indicating three levels of severity (no problems, some or moderate problems, extreme problems). The EQ visual Analogue scale (EQ-VAS) is a standard vertical 20 cm VAS recording people’s rating of their current health-related quality of life (HRQoL) ranging from ‘the best health you can imagine’ to ‘the worst health you can imagine’ . The EQ-5D index will be used to derive QALY using the Swedish experience-based value set and an area-under-the-curve calculation . An alternative valuation using a general population value set from the UK will be part of the sensitivity analysis, as Sweden has no such societal valuation. ICEpop CAPability measure for older people (ICECAP-O) and capability-adjusted life years (CALY) The ICECAP-O capability index estimates quality of life (QoL) in a broader sense compared to the HRQoL, including five attributes (attachment, role, enjoyment, security, control), each with four response levels. The ICECAP-O instrument will be used to derive CALY using the British valuation . Change in daily activities based on the Activities of Daily Life (ADL) staircase Change in dependence in daily activities will be assessed based on the ADL staircase . The Instrument contains a cumulative scale of six personal activities (P-ADLs: bathing, dressing, going to the toilet, transfer, continence, feeding) and four instrumental activities (I-ADLs: cleaning, shopping, transportation, cooking). Dependence is defined as receiving personal or directive assistance from another person . In line with previous studies evaluating ADL in frail older people, a participant living with another person is assessed as “independent” if they are capable of performing the activity without assistance when alone. Societal costs Resource use in the economic evaluation will include register data on health care use (held by the Region Västra Götaland), as well as dispensed drugs, and use of social care use (held by the National Board of Health and Welfare). Informal care and other expenses to the participant and family and friends related to the treatment will be collected using patient questionnaires and based on user data from the digital platform. Costs (both reimbursements and out-of-pocket costs) for prescribed drugs will be obtained from the Swedish Prescribed Drug Register (SPDR) . The out-of-pocket costs for healthcare use and social care will be calculated from national statistics on patient fees while accounting for the national high-cost protection schemes, in line with previous work . Regional costs for healthcare will be calculated using cost-per-patient data and diagnosis-related group weights (DRG) from specialised care and the national mean cost for producing one DRG, while the corresponding costs for primary care and social care will be derived using national cost statistics. Additional questions have been added in the questionnaire to identify expenses not accounted for in register data. Informal care costs will be viewed as a direct cost, and thus valued at the average wage and social security contribution of employing a formal caregiver, sometimes called the replacement cost approach . Incremental cost-effectiveness ratio (ICER) The ICER will be calculated as the societal cost difference between groups divided by the corresponding difference in QALY and CALY , respectively as recommended in studies including both health and social services . Change in the burden of medicine use in everyday life based on The Living with Medicines Questionnaire version 3 (LMQ-3) Participants will self-rate eight domains: relationship with health professionals, practical difficulties, interference with daily life, lack of effectiveness, side effects, general concerns, cost, and lack of autonomy. Each domain will be rated on a 5-point Likert scale (strongly agree to strongly disagree) . Change in the participants’ self-rated overall level of burden of medicine Participants will assess the overall level of burden of medicine on a 10 cm VAS from 0 (no burden at all) to 10 (extremely burdensome) . Social network, social support, and loneliness Five items will ask about social network, sufficient support, someone to trust and confide in, experience safety and security and feelings of loneliness [ – ]. Self-reported frailty indicators according to the Tilburg Frailty Indicator (TFI) The TFI will be used to assess self-reported frailty indicators, unexplained weight loss, difficulty in walking, strength in hands, physical health, physical tiredness, balance, problem with hearing and vision (physical domain), cognition, depressive symptoms, anxiety, and coping (psychological domain), living alone, social relations and social support (social domain). An additional question about falls during the past 3 months will also be used . Frailty according to the clinical Frailty Scale (CFS) version 1.2 The degree of frailty will be assessed from 1 to 9; 1: very fit, 2: well, 3: manging well, 4: vulnerable, 5: mildly frail, 6: moderately frail, 7: severely frail, 8: very severely frail, 9: terminally ill) . For details about primary and secondary outcomes, data sources, and time points for data collection, see Table . Blinding The nature of the intervention means that neither participants nor the HCPs in the IHOPe intervention can be blinded to allocation in the RCT. Sample size and power calculation In the full-scale RCT and health economic evaluation a minimum of 220 patients must be randomised to the two arms based on the primary outcome measure. To achieve 80% power based on a p-value of 0.05 and allow an increase in the proportion of improved or unchanged patients from 20 to 40%, 91 participants in each group (control and intervention) are required. We plan to include 110 patients in each group to have a comfortable margin for dropouts and withdrawals. Timeline Recruitment of participants started in the spring of 2020 and is expected to be completed in 2023. The intervention will continue until early 2024. Data for the primary outcome will be collected three months after the last patient inclusion. Data collection for the secondary outcome measures will continue until early 2025. After that, retrospective register data will be obtained as there is a delay from data collection to transmission to the administrative registers. Data analysis Analysis will be performed according to two analysis sets: (1) the intention to treat set, meaning that all allocated participants will be analysed in the group to which they were randomised and (2) the per-protocol set, which includes only those participants who at least have one telephone conversation and an agreed health plan to indicate an expected minimum level of participation. If necessary, eventual differences between the intervention and control group at baseline will be adjusted for in the analysis. Different options for handling missing data will be considered based on the collected data. Missing data due to death will be replaced with a value for worst-case change (deteriorated in the composite of primary outcomes). Based on the assumptions that the study sample is expected to decline over time due to the ageing process and that deteriorated health is a common reason not to fulfil follow-ups, the imputation model of median change deterioration will likely be used. However, the final decision on how to handle missing data will be decided by the research team in consultation with a statistician. Descriptive and analytical statistics will be used to compare the control and intervention groups and to measure change over time. The chi-square or Fisher’s exact test will be computed to compare the proportions between groups. Logistic regression will be used to calculate odds ratios (ORs) with 95% confidence intervals (CIs). Parametric tests will be performed when suitable, and non-parametric tests will be applied when analysing ordinal data. Health economic evaluation Reporting of the economic evaluation will follow the Consolidated Health Economic Evaluation Reporting Standards . Costs will be described and analysed as cost components and total cost, with 95% CIs to indicate uncertainty. Bootstrapping will be used to calculate CIs for any skewed variables . The distribution of costs will also be analysed by major stakeholders: county council/regions, municipalities, market sectors (productivity loss), and individuals/family/friends. In the case of clearly beneficial effects for costs and intervention outcomes (i.e., lower costs and more health created compared to the alternative) for one of the treatment arms, thus dominating the comparator, this information will be reported. However, increased health benefits are often associated with higher costs, in which case we will use cost-utility analysis to estimate the cost of gaining one additional QALY or CALY, respectively. According to state-of-the-art practice, sensitivity analyses will be applied to test the robustness of the results to necessary assumptions and alternative cost levels. A probabilistic sensitivity analysis will be conducted using longitudinal regression analysis with bootstrapped samples that are visualised using a cost-effectiveness plane. This regression analysis enables adjustment for identified confounding factors. Results from the regression analysis will be compared to the corresponding estimations made using multiple imputations, which have been used more often in previous studies . In addition, group-based trajectory modelling will explore potential cost variations and outcomes within the patient group . This method identifies groups of patients with a similar trajectory in some outcome over time, such as health care costs. The identified groups will then be used to explore patient characteristics associated with such trajectories to provide guidance on which patients would benefit initially from a more intense intervention or are more likely to deteriorate. Process evaluation Several sub-studies will be conducted to widen and deepen the understanding of the impact of the intervention mechanisms. The following questions will guide the evaluation: How do frail older people experience the intervention? To what extent and how meaningful is the intervention to frail older people? To what extent will the intervention reach frail older people? Where applicable, additional analyses describing costs during the implementation phase will be conducted . A mixed-method approach will be applied to deal with the intervention’s complexity by combining quantitative and qualitative methods. Quantitative data will be collected from questionnaires and ratings of the telephone and platform support, which will be attached to the 6-month questionnaire and sent to participants in the intervention group. We will also analyse data on using different platform functions and the number and modes of contact between HCPs, participants, and team members invited to the platform during the intervention period. The content of audio-recorded person-centred conversations and written health plans will also be analysed. In addition, qualitative interviews will be performed to gain a better understanding of the intervention mechanism of impact from the participants’ and their team members’ perspectives. However, in the process evaluation the number of informants depends on the research question in the specific sub-study. There will be a purposeful sampling of participants and informal and formal caregivers for interviews. Qualitative data will be analysed using content analysis and quantitative data will be analysed using descriptive statistics. Ethics and dissemination An ethical application has been approved by Swedish Ethical Review Authority (Dnr 2019–05364, Dnr 2020–03550, Dnr 2021–03255), and the study will comply with the ethical principles of the Declaration of Helsinki. All participants will be asked to sign a written consent form after receiving oral and written information about the study. Informed consent will also be asked of participants accepting to be interviewed. The findings will be disseminated through academic publications and conferences in the field of health care and medicine. Moreover, the results will be presented for health and social care professionals through appropriate local forums. A user-friendly summary of findings will be sent to participants who have indicated they are interested in the results.
The involvement of different categories of knowledge users (e.g. older people, health professionals and decisionmakers) is advocated as a crucial part of research . Therefore, the planning and designing of the study have been conducted in close communication and collaboration with an advisory group including older people, inter-professional researchers, decisionmakers, and clinicians. The development of the digital platform was guided by a participatory design . Moreover, the entire study, including recruitments, choice of instruments, questionnaire development, and the intervention components, have been, and continually will be, co-created with the advisory group. Participants and setting. The IHOPe project will be conducted in Sweden and targets community-dwelling older people aged ≥ 75 years, their families, and professionals working in social and healthcare services. Sweden has an ongoing redesign of healthcare services towards person-centred integrated care, which aims to strengthen primary care services and self-care . Health costs in medical care are mainly financed through taxes, where municipalities are responsible for services for older people according to assessed needs. Several actors could be involved in formal care activities for frail older people representing different health and social care organisations. There is also an option to seek private health and social care providers . In addition, it is common that older people are receiving support or help from informal caregivers . Therefore, the evaluation process also includes significant others and HCPs invited to the IHOPe platform. All participants must provide signed informed consent before any study procedures occur. Inclusion criteria Inclusion criteria will be national registration within the Region Västra Götaland, men and women aged ≥ 75 years living in ordinary housing and screened as frail at an emergency department but not hospitalised. Exclusion criteria Exclusion criteria will be those needing palliative care in the final stages of life, no registered address, participating in a conflicting randomised study, or cognitive dysfunctionality (not oriented to time, place, and person).
Inclusion criteria will be national registration within the Region Västra Götaland, men and women aged ≥ 75 years living in ordinary housing and screened as frail at an emergency department but not hospitalised.
Exclusion criteria will be those needing palliative care in the final stages of life, no registered address, participating in a conflicting randomised study, or cognitive dysfunctionality (not oriented to time, place, and person).
Frail older people will be recruited from the emergency department at a university hospital in Gothenburg, Sweden, to include a diverse population of frail older people. The planned flow of participants is shown in Fig. . Participants aged ≥ 75 years will be screened using an instrument developed by the Frail Elderly Support Research Group (FRESH) as part of usual care at the emergency department or by a research assistant via telephone. The screening results (if frail or not) will be documented. The research assistant will screen patients identified as frail for eligibility to participate in the study using the above-mentioned inclusion and exclusion criteria and the electronic patient record. A study information letter and invitation to participate will be sent to eligible patients within two weeks after their emergency visit. The patient will then be contacted by telephone to provide verbal information and asked about consent to participate. Finally, written participant information and informed consent forms will be sent to patients willing to participate. All who fulfil the inclusion criteria and consent to participate will be randomised into the study. Participants will be randomly assigned to the control or the intervention group with a 1:1 allocation using a computer-generated randomisation schedule. The control group will receive usual care whilst the intervention group will receive the person-centred e-support intervention for six months in addition to usual care. The recruitment of participants to the main study started after the inclusion of 17 participants in a pilot study. The full-scale RCT is currently ongoing. Follow-up for each study participant continues for 12 months from inclusion. Recruitment of significant others and HCPs depart from the older people’s network by contact information to significant others or formal carers invited to the intervention.
The participants allocated to the control group will receive usual care . Routine care for older people in Sweden includes care delivered by diverse HCPs in hospital, primary care, municipal, and home care services. Also, informal carers are often involved. It is unusual to have any follow-up contact from the emergency ward for frail older people who are not admitted to hospital. However, within usual care some patients are contacted by a ‘mobile team’ (consisting of nurses and physicians) through a coordinator at the emergency ward. The mobile team visits the patient at home or contacts the municipal care provider for further follow-up, depending on the older person’s healthcare needs and status . However, the mobile team does not deliver digital healthcare.
In addition to usual care, participants allocated to the intervention group will receive person-centered telephone support and the opportunity to communicate with HCPs through the digital platform for six months. The intervention is delivered remotely and is coordinated by dedicated registered nurses (RNs). The HCPs conducting the intervention receive continuous training through workshops with specialists in PCC, ethics, communication, and pedagogics on how to apply remote PCC. Intervention components. The intervention aims to initiate partnership-building, encourage older people to describe their health situation and needs, and further identify and engage their capabilities and resources .
An HCP will initiate the first telephone conversation shortly after study inclusion (within two weeks). After the first telephone call, a follow-up appointment will be mutually planned and agreed upon between the HCP and the older person. In addition, there is the opportunity for incidental contact with the older person by telephone or through the digital platform. The participants’ narrative of their everyday situation, as well as their needs, resources, and health-related goals are the core components of the telephone conversations . The conversation will start with recapitulating the aim of PCC support, asking open-ended questions, and listening to the participants’ experience of their condition, context, needs, and resources. The HCP will confirm and summarise the content of the conversation and, in collaboration with the participant, co-create achievable and relevant health-related goals. The participants’ narratives (at times in partnership with significant others), goals, and strategies to manage everyday life, as well as needs of support and actions and resources to achieve agreed health-related goals, will then be documented in a health plan by either the participant or the HCP and uploaded to the digital platform. This health plan is, however, a living document and can be modified and reformulated during the intervention, as is described in other PCC projects with different populations [ , , ].
The digital platform has been tested in previous studies and the present study will apply the same procedure of working together via the platform . All participants are invited to communicate on the platform. The participants or the responsible HCP will upload the mutually agreed health plan to the digital platform. Depending on agreements during the phone call, family, friends, and additional health professionals may be invited to join the digital platform team. An HCP will introduce and explain how to use the digital platform and create a personal account. The platform is a mediator through which the participants, together with their team (i.e., formal and informal family and informal carers), can develop the health plan and follow-up actions and goals. The platform allows self-monitoring through daily ratings on a scale from 1 to 5 on symptoms and general well-being (e.g., how well they have slept). In addition, the participants and the HCPs can communicate through messages in a chat-like forum. The platform will contain an assembly of links to other relevant web pages that the participants can use to seek information or connections. HCPs log into the platform on weekdays to be updated on patients’ activities and check for messages . If signs of deterioration occur needing for example changes in the pharmacological treatment, the participants will be advised to contact their primary care physician and if symptoms then are assessed as urgent, they are encouraged to contact the emergency healthcare services.
Various data sources will be used, including questionnaires (see detailed description in Table ), activities performed by the participants in the digital platform, time consumption during conversations, and for writing health plans and medical records. In addition, the selection of participants (in the intervention group), as well as significant others and health professionals invited for teamwork, will be asked if they are willing to participate in focus groups or individual interviews. Clinical and self-reported data (e.g., questionnaires) will be collected at baseline and after 3, 6, and 12 months. The primary and secondary endpoints have been used in large national or international studies and tested for validity and reliability . Moreover, register data will be collected retrospectively from the National Board of Health and Welfare: each registered holder, health care encounters from the regional patient register VEGA , use of prescribed drugs dispensed in outpatient pharmacies from the Swedish Prescribed Drug Register (SPDR) , use of home help services and supported living facilities, municipal health care use, and information about potential causes of death.
Questionnaire data will be gathered at baseline and after 3, 6, and 12 months from the inclusion date. Self-reported baseline characteristics of the participant’s sex, age, civil status, country of birth, and level of education will be collected through questionnaires at baseline.
In line with previous RCTs evaluating PCC the primary outcome will be a composite of clinical changes , in general self-efficacy and the need for hospital care for unscheduled reasons after 3 months. Each participant will be classified as improved, deteriorated, or unchanged at three months as follows: Deteriorated : the participant’s general self-efficacy has decreased by [12pt]{minimal}
$$$$ 5 units or has been admitted to hospital for unscheduled reasons two or more times. Improved : general self-efficacy has increased by [12pt]{minimal}
$$$$ 5 units, and the participant has not been admitted to hospital more than once. Unchanged : the patient has neither deteriorated nor improved.
A composite of changes in general self-efficacy and the need for hospital care for unscheduled reasons at 6 and 12 months. General self-efficacy The General Self-Efficacy Scale (GSES) is a 10-item self-assessment psychometric questionnaire designed to measure a broad and stable sense of personal competence to deal with stressful situations. Ratings are made on a 4-point Likert scale (1 = not at all true, 2 = hardly true, 3 = moderately true, 4 = exactly true) and are summed to give a total score ranging from 10 to 40, with higher scores indicating generalised greater self-efficacy. The GSES has been validated in several languages. An increase of 4.6 units in the total score has been proposed and is used as a limit for a minimally important difference [ , , ]. Hospitalisation Hospitalisation refers to the number of unplanned emergency department visits and hospital admission according to patients’ self-reported responses (questionnaires) and data in medical records. Health-related quality of life by EuroQol 5 dimensions health state questionnaire (EQ-5D) and quality adjusted-life years (QALY) EQ-5D is a generic measure of health status consisting of five dimensions (mobility, self-care, usual activities, pain/discomfort, anxiety/depression), each with responses indicating three levels of severity (no problems, some or moderate problems, extreme problems). The EQ visual Analogue scale (EQ-VAS) is a standard vertical 20 cm VAS recording people’s rating of their current health-related quality of life (HRQoL) ranging from ‘the best health you can imagine’ to ‘the worst health you can imagine’ . The EQ-5D index will be used to derive QALY using the Swedish experience-based value set and an area-under-the-curve calculation . An alternative valuation using a general population value set from the UK will be part of the sensitivity analysis, as Sweden has no such societal valuation. ICEpop CAPability measure for older people (ICECAP-O) and capability-adjusted life years (CALY) The ICECAP-O capability index estimates quality of life (QoL) in a broader sense compared to the HRQoL, including five attributes (attachment, role, enjoyment, security, control), each with four response levels. The ICECAP-O instrument will be used to derive CALY using the British valuation . Change in daily activities based on the Activities of Daily Life (ADL) staircase Change in dependence in daily activities will be assessed based on the ADL staircase . The Instrument contains a cumulative scale of six personal activities (P-ADLs: bathing, dressing, going to the toilet, transfer, continence, feeding) and four instrumental activities (I-ADLs: cleaning, shopping, transportation, cooking). Dependence is defined as receiving personal or directive assistance from another person . In line with previous studies evaluating ADL in frail older people, a participant living with another person is assessed as “independent” if they are capable of performing the activity without assistance when alone. Societal costs Resource use in the economic evaluation will include register data on health care use (held by the Region Västra Götaland), as well as dispensed drugs, and use of social care use (held by the National Board of Health and Welfare). Informal care and other expenses to the participant and family and friends related to the treatment will be collected using patient questionnaires and based on user data from the digital platform. Costs (both reimbursements and out-of-pocket costs) for prescribed drugs will be obtained from the Swedish Prescribed Drug Register (SPDR) . The out-of-pocket costs for healthcare use and social care will be calculated from national statistics on patient fees while accounting for the national high-cost protection schemes, in line with previous work . Regional costs for healthcare will be calculated using cost-per-patient data and diagnosis-related group weights (DRG) from specialised care and the national mean cost for producing one DRG, while the corresponding costs for primary care and social care will be derived using national cost statistics. Additional questions have been added in the questionnaire to identify expenses not accounted for in register data. Informal care costs will be viewed as a direct cost, and thus valued at the average wage and social security contribution of employing a formal caregiver, sometimes called the replacement cost approach . Incremental cost-effectiveness ratio (ICER) The ICER will be calculated as the societal cost difference between groups divided by the corresponding difference in QALY and CALY , respectively as recommended in studies including both health and social services . Change in the burden of medicine use in everyday life based on The Living with Medicines Questionnaire version 3 (LMQ-3) Participants will self-rate eight domains: relationship with health professionals, practical difficulties, interference with daily life, lack of effectiveness, side effects, general concerns, cost, and lack of autonomy. Each domain will be rated on a 5-point Likert scale (strongly agree to strongly disagree) . Change in the participants’ self-rated overall level of burden of medicine Participants will assess the overall level of burden of medicine on a 10 cm VAS from 0 (no burden at all) to 10 (extremely burdensome) . Social network, social support, and loneliness Five items will ask about social network, sufficient support, someone to trust and confide in, experience safety and security and feelings of loneliness [ – ]. Self-reported frailty indicators according to the Tilburg Frailty Indicator (TFI) The TFI will be used to assess self-reported frailty indicators, unexplained weight loss, difficulty in walking, strength in hands, physical health, physical tiredness, balance, problem with hearing and vision (physical domain), cognition, depressive symptoms, anxiety, and coping (psychological domain), living alone, social relations and social support (social domain). An additional question about falls during the past 3 months will also be used . Frailty according to the clinical Frailty Scale (CFS) version 1.2 The degree of frailty will be assessed from 1 to 9; 1: very fit, 2: well, 3: manging well, 4: vulnerable, 5: mildly frail, 6: moderately frail, 7: severely frail, 8: very severely frail, 9: terminally ill) . For details about primary and secondary outcomes, data sources, and time points for data collection, see Table .
The General Self-Efficacy Scale (GSES) is a 10-item self-assessment psychometric questionnaire designed to measure a broad and stable sense of personal competence to deal with stressful situations. Ratings are made on a 4-point Likert scale (1 = not at all true, 2 = hardly true, 3 = moderately true, 4 = exactly true) and are summed to give a total score ranging from 10 to 40, with higher scores indicating generalised greater self-efficacy. The GSES has been validated in several languages. An increase of 4.6 units in the total score has been proposed and is used as a limit for a minimally important difference [ , , ].
Hospitalisation refers to the number of unplanned emergency department visits and hospital admission according to patients’ self-reported responses (questionnaires) and data in medical records.
EQ-5D is a generic measure of health status consisting of five dimensions (mobility, self-care, usual activities, pain/discomfort, anxiety/depression), each with responses indicating three levels of severity (no problems, some or moderate problems, extreme problems). The EQ visual Analogue scale (EQ-VAS) is a standard vertical 20 cm VAS recording people’s rating of their current health-related quality of life (HRQoL) ranging from ‘the best health you can imagine’ to ‘the worst health you can imagine’ . The EQ-5D index will be used to derive QALY using the Swedish experience-based value set and an area-under-the-curve calculation . An alternative valuation using a general population value set from the UK will be part of the sensitivity analysis, as Sweden has no such societal valuation.
The ICECAP-O capability index estimates quality of life (QoL) in a broader sense compared to the HRQoL, including five attributes (attachment, role, enjoyment, security, control), each with four response levels. The ICECAP-O instrument will be used to derive CALY using the British valuation .
Change in dependence in daily activities will be assessed based on the ADL staircase . The Instrument contains a cumulative scale of six personal activities (P-ADLs: bathing, dressing, going to the toilet, transfer, continence, feeding) and four instrumental activities (I-ADLs: cleaning, shopping, transportation, cooking). Dependence is defined as receiving personal or directive assistance from another person . In line with previous studies evaluating ADL in frail older people, a participant living with another person is assessed as “independent” if they are capable of performing the activity without assistance when alone.
Resource use in the economic evaluation will include register data on health care use (held by the Region Västra Götaland), as well as dispensed drugs, and use of social care use (held by the National Board of Health and Welfare). Informal care and other expenses to the participant and family and friends related to the treatment will be collected using patient questionnaires and based on user data from the digital platform. Costs (both reimbursements and out-of-pocket costs) for prescribed drugs will be obtained from the Swedish Prescribed Drug Register (SPDR) . The out-of-pocket costs for healthcare use and social care will be calculated from national statistics on patient fees while accounting for the national high-cost protection schemes, in line with previous work . Regional costs for healthcare will be calculated using cost-per-patient data and diagnosis-related group weights (DRG) from specialised care and the national mean cost for producing one DRG, while the corresponding costs for primary care and social care will be derived using national cost statistics. Additional questions have been added in the questionnaire to identify expenses not accounted for in register data. Informal care costs will be viewed as a direct cost, and thus valued at the average wage and social security contribution of employing a formal caregiver, sometimes called the replacement cost approach .
The ICER will be calculated as the societal cost difference between groups divided by the corresponding difference in QALY and CALY , respectively as recommended in studies including both health and social services .
Participants will self-rate eight domains: relationship with health professionals, practical difficulties, interference with daily life, lack of effectiveness, side effects, general concerns, cost, and lack of autonomy. Each domain will be rated on a 5-point Likert scale (strongly agree to strongly disagree) .
Participants will assess the overall level of burden of medicine on a 10 cm VAS from 0 (no burden at all) to 10 (extremely burdensome) .
Five items will ask about social network, sufficient support, someone to trust and confide in, experience safety and security and feelings of loneliness [ – ].
The TFI will be used to assess self-reported frailty indicators, unexplained weight loss, difficulty in walking, strength in hands, physical health, physical tiredness, balance, problem with hearing and vision (physical domain), cognition, depressive symptoms, anxiety, and coping (psychological domain), living alone, social relations and social support (social domain). An additional question about falls during the past 3 months will also be used .
The degree of frailty will be assessed from 1 to 9; 1: very fit, 2: well, 3: manging well, 4: vulnerable, 5: mildly frail, 6: moderately frail, 7: severely frail, 8: very severely frail, 9: terminally ill) . For details about primary and secondary outcomes, data sources, and time points for data collection, see Table .
The nature of the intervention means that neither participants nor the HCPs in the IHOPe intervention can be blinded to allocation in the RCT.
In the full-scale RCT and health economic evaluation a minimum of 220 patients must be randomised to the two arms based on the primary outcome measure. To achieve 80% power based on a p-value of 0.05 and allow an increase in the proportion of improved or unchanged patients from 20 to 40%, 91 participants in each group (control and intervention) are required. We plan to include 110 patients in each group to have a comfortable margin for dropouts and withdrawals.
Recruitment of participants started in the spring of 2020 and is expected to be completed in 2023. The intervention will continue until early 2024. Data for the primary outcome will be collected three months after the last patient inclusion. Data collection for the secondary outcome measures will continue until early 2025. After that, retrospective register data will be obtained as there is a delay from data collection to transmission to the administrative registers.
Analysis will be performed according to two analysis sets: (1) the intention to treat set, meaning that all allocated participants will be analysed in the group to which they were randomised and (2) the per-protocol set, which includes only those participants who at least have one telephone conversation and an agreed health plan to indicate an expected minimum level of participation. If necessary, eventual differences between the intervention and control group at baseline will be adjusted for in the analysis. Different options for handling missing data will be considered based on the collected data. Missing data due to death will be replaced with a value for worst-case change (deteriorated in the composite of primary outcomes). Based on the assumptions that the study sample is expected to decline over time due to the ageing process and that deteriorated health is a common reason not to fulfil follow-ups, the imputation model of median change deterioration will likely be used. However, the final decision on how to handle missing data will be decided by the research team in consultation with a statistician. Descriptive and analytical statistics will be used to compare the control and intervention groups and to measure change over time. The chi-square or Fisher’s exact test will be computed to compare the proportions between groups. Logistic regression will be used to calculate odds ratios (ORs) with 95% confidence intervals (CIs). Parametric tests will be performed when suitable, and non-parametric tests will be applied when analysing ordinal data.
Reporting of the economic evaluation will follow the Consolidated Health Economic Evaluation Reporting Standards . Costs will be described and analysed as cost components and total cost, with 95% CIs to indicate uncertainty. Bootstrapping will be used to calculate CIs for any skewed variables . The distribution of costs will also be analysed by major stakeholders: county council/regions, municipalities, market sectors (productivity loss), and individuals/family/friends. In the case of clearly beneficial effects for costs and intervention outcomes (i.e., lower costs and more health created compared to the alternative) for one of the treatment arms, thus dominating the comparator, this information will be reported. However, increased health benefits are often associated with higher costs, in which case we will use cost-utility analysis to estimate the cost of gaining one additional QALY or CALY, respectively. According to state-of-the-art practice, sensitivity analyses will be applied to test the robustness of the results to necessary assumptions and alternative cost levels. A probabilistic sensitivity analysis will be conducted using longitudinal regression analysis with bootstrapped samples that are visualised using a cost-effectiveness plane. This regression analysis enables adjustment for identified confounding factors. Results from the regression analysis will be compared to the corresponding estimations made using multiple imputations, which have been used more often in previous studies . In addition, group-based trajectory modelling will explore potential cost variations and outcomes within the patient group . This method identifies groups of patients with a similar trajectory in some outcome over time, such as health care costs. The identified groups will then be used to explore patient characteristics associated with such trajectories to provide guidance on which patients would benefit initially from a more intense intervention or are more likely to deteriorate.
Several sub-studies will be conducted to widen and deepen the understanding of the impact of the intervention mechanisms. The following questions will guide the evaluation: How do frail older people experience the intervention? To what extent and how meaningful is the intervention to frail older people? To what extent will the intervention reach frail older people? Where applicable, additional analyses describing costs during the implementation phase will be conducted . A mixed-method approach will be applied to deal with the intervention’s complexity by combining quantitative and qualitative methods. Quantitative data will be collected from questionnaires and ratings of the telephone and platform support, which will be attached to the 6-month questionnaire and sent to participants in the intervention group. We will also analyse data on using different platform functions and the number and modes of contact between HCPs, participants, and team members invited to the platform during the intervention period. The content of audio-recorded person-centred conversations and written health plans will also be analysed. In addition, qualitative interviews will be performed to gain a better understanding of the intervention mechanism of impact from the participants’ and their team members’ perspectives. However, in the process evaluation the number of informants depends on the research question in the specific sub-study. There will be a purposeful sampling of participants and informal and formal caregivers for interviews. Qualitative data will be analysed using content analysis and quantitative data will be analysed using descriptive statistics.
An ethical application has been approved by Swedish Ethical Review Authority (Dnr 2019–05364, Dnr 2020–03550, Dnr 2021–03255), and the study will comply with the ethical principles of the Declaration of Helsinki. All participants will be asked to sign a written consent form after receiving oral and written information about the study. Informed consent will also be asked of participants accepting to be interviewed. The findings will be disseminated through academic publications and conferences in the field of health care and medicine. Moreover, the results will be presented for health and social care professionals through appropriate local forums. A user-friendly summary of findings will be sent to participants who have indicated they are interested in the results.
Nearly 40% of patients who seek the emergency ward are frail older people. Identifying those screened as frail but are directly discharged to home is crucial because they risk “falling between the chairs” . Frail older people need more focused health-promoting actions to prevent hospitalisations and promote well-being . The IHOPe project focuses on health promotive actions based on PCC and the capability approach, which adds an essential component to existing evidence-based teamwork approaches, such as comprehensive geriatric assessment or case-manager interventions that focus on providing care. Barriers reported in the implementation of person-centred practice is the lack of crucial ethical underpinnings for PCC, including the patient being an equal team partner . In IHOPe two persons representing frail older people collaborated with the research group in designing the project. The intervention and connecting to the theoretical assumptions have been emphasised: for example, the older person being an equal partner and health and social care professionals working with the older person (and their significant others), also acknowledging their social world and internal and external capabilities [ , , ]. This person-centred remote support and conversations based on frail older peoples’ narratives will facilitate reviewing their everyday life from their perspectives and elucidating and confirming their resources and strategies to manage the consequences of frailty, which may lead to retained or improved self-efficacy. In addition to person-centred telephone communication, the partnership with frail older people will be enhanced through a digital platform, a challenging and essential aspect of the intervention that facilitates the inclusion of individuals living in digital alienation. Previous or ongoing studies, including health planning after frailty screening, have mainly evaluated effects on interventions in existing care teams focusing on how professionals work for older people in different settings face-to-face [ , , ]. A person-centred approach correlates to improved work satisfaction among professionals . Opportunities for teamwork with frail older people will be enhanced, leading to less stress and effective use of the efforts of HCPs and informal carers. IHOPe aims to directly contribute to reaching the national goals of persons ageing at home. Through modest adjustments, IHOPe can be easily applied to existing healthcare systems. The possibility to upscale is in line with the national eHealth strategy, with IHOPe representing a workable asset that is not limited to time or location. Consistent with person-centred ethics, this is the first study of PCC that includes a capability approach in the economic evaluation . Some research indicates that PCC could effectively use resources in care, even remotely delivered . However, even if remote health-promoting services are recommended to improve reach , access, and efficiency in health care, they are underused in older populations . Thus far, studies evaluating remote interventions have evaluated interventions performed at a late stage, when people are already highly dependent on help. A combined scoping and systematic review found that few studies evaluated proactive interventions involving the triad of person-centred, digital, and integrated care elements. Therefore, this project will include these three elements. The number of cost-effectiveness studies on health-promoting interventions in the older population is also limited. A scoping review performed in the Nordic context identified four cost-effectiveness studies conducted in face-to-face interventions. None of these studies had evaluated remote interventions, but such solutions were suggested as a promising option to support intervention reach and cost-effectiveness . To date, no published studies have evaluated the cost-effectiveness of remote person-centred health promotion to frail older persons. However, a recent systematic review demonstrated an urgent need to assess the cost-effectiveness of remote health services for sub-populations of older adults. IHOPe is expected to enhance frail older people’s involvement in their care and QoL through person-centred preventive care and health-promoting support through accessible welfare technology. ICEpop CAPability measure for Older people (ICECAP-O), Activities of Daily Life (ADL), Living with Medicines Questionnaire version 3 (LMQ-3), Tilburg Frailty Indicator (TFI), Clinical Frailty Scale (CFS).
|
Immunohistochemical visualization of lymphatic vessels in human dura mater: methodological perspectives | c3b347dc-f32a-47f8-a64e-c4eac47b78aa | 10044429 | Anatomy[mh] | The study of meningeal lymphatic vessels has a long history. In 1787, the Italian anatomist Giovanni Paolo Mascagni for the first time demonstrated lymphatic vessels within the human dura, which he described in a Latin text termed Vasorum Lymphaticorum Corporis Humani Historia et Ichnographia (English History and Graphical Representation of the Lymphatic Vessels in the Human Body ) . While this text was forgotten for many years, the study of meningeal lymphatic vessels later attracted the interest of several authors. Lecco described lymphatic structures within the human dura , and Földi studied lymphatic structures within dura mater of dogs , as well as functional consequences of cervical lymphatic blockade in dogs, cats and rats . Furthermore, Andres et al. showed lymphatic structures in dura of rats. Cserr explored the role of lymphatic drainage of macromolecules from the brain, as well as its role in immune surveillance . More recently, lymphatic structures were identified in the dural portion of the optic nerve in autopsy specimens , and in the murine craniofacial region . Moreover, Johnston studied lymphatic drainage of cerebrospinal fluid (CSF) via the cribriform plate, particularly addressing its possible role in hydrocephalus . In 2015, the study of meningeal lymphatic vessels gained greatly renewed interest, when functional lymphatic vessels within the dura mater of rodents were reported . These studies demonstrated lymphatic vessels nearby the dural sinus veins of rodents that were able to carry fluid and immune cells from the CSF . Moreover, loss of dural meningeal vessels abrogated the transport of macromolecules to the deep cervical lymph nodes . Later, a growing body of evidence points at a pivotal role of the meningeal lymphatic vessels for egress of waste solutes from the central nervous system (CNS), as well as their role in immune surveillance of the CNS . However, the existing evidence about meningeal lymphatic function largely derives from animal studies, while human data exclusively rely on observations in autopsy specimens . Given the possible great importance of meningeal lymphatic structures for normal function of the CNS as well as diseases of the CNS, there is an urgent need for establishing methodology to characterize meningeal lymphatic vessels in humans. This present study was undertaken to examine methodology for immunohistochemistry of human dural lymphatic vessels. To this end, we examined dura mater of patients with the dementia subtype idiopathic normal pressure hydrocephalus (iNPH) who as part of treatment underwent shunt surgery. Dura mater was retrieved from the same location in all subjects, and we compared three different fixation methods in order to characterize the meningeal lymphatic vessels in humans. We decided to examine dura mater in iNPH patients since these patients present with delayed molecular clearance from subarachnoid CSF , impaired glymphatic function as well as altered CSF to blood clearance . Given the previous data that loss of dural meningeal vessels impairs the transport of macromolecules to the deep cervical lymph nodes , we hypothesize that meningeal lymphatic function could be affected in iNPH, and therefore of value to examine in this patient group. Perhaps a dural biopsy during shunt surgery may add diagnostic information to this disease? In order to determine the importance of dural lymphatics in disease such as iNPH, methods must be developed to assess these vessels in tissue biopsies of these patients.
Patients We included patients with iNPH who underwent shunt surgery, and who followed the ordinary clinical routine for shunt surgery within the department of neurosurgery. Notably, it was beyond the scope of this study to compare meningeal lymphatic vessels in iNPH versus other diseases or control subjects. Tissue sampling and preparation The dura mater was sampled during ventriculoperitoneal (VP) shunt surgery based on clinical routine at the department of neurosurgery in Oslo University Hospital – Rikshospitalet. The trepanation necessary for the surgical procedure and tissue sampling was performed 12 cm caudally from the glabella and 1–2 cm lateral to the midline (Fig. ). The surgical procedure was performed using an operating microscope. A burr hole less than 1 cm in diameter was made and the underlying dura exposed. With a knife, the dura was incised circumferential (diameter about 5 mm) in its entire thickness (about 1 mm) and removed in one piece. Once extracted, the sample was prepared with one of three methods. Method #1 Six dura samples (numbered #1 to #6) were immersed in 4% paraformaldehyde (PFA). The following day the samples were transferred for preservation to a diluted (0.4%) PFA solution. One day prior to sectioning, they were cryoprotected in graded sucrose solutions (10% for 30 min, 20% for one hour and 30% sucrose overnight in 0.1 M phosphate buffer). For sectioning the samples were frozen on dry ice in OCT medium (Richard-Allan Scientific™ Neg-50™, Thermo Fisher Scientific; Cat#6502) and sectioning was performed in a Cryostat NX70 from Thermo Scientific with vacutome. It was not possible to determine the rostro-caudal or medio-lateral orientation in the horizontal plane of the tissue. Therefore, the sections obtained were either sagittal or coronal and 16 μm thick. Cranial (outer) vs. arachnoid (inner) orientation in the vertical plane of the tissue was deciphered through microscopy (See “Microscopy and measurements”). These sections were immediately mounted on Superfrost™ Plus microscopy slides (Thermo Fisher Scientific) and stored at -20 °C until immunohistochemistry (IHC). This fixation method did not produce any staining (see Discussion) and therefore the experiments with these samples were repeated with antigen retrieval methods (described below). Antigen retrieval did not result in any significant improvement and this lead to trying Method#2 with the subsequent patient samples. Method #2 As with Method #1, seven dura samples (numbered #7 to #13) were immersed in 0.5% PFA immediately after resection and was fixed overnight at 4 °C. One day prior to sectioning, they were cryoprotected in graded sucrose solutions, as mentioned above. Sectioning was performed as described for Method #1. The thickness of the sections was modified to 30 μm to prevent washing off the glass slide, which was frequent with Method #1 particularly after antigen retrieval. These sections were immediately mounted on Superfrost™ Plus microscopy slides (Thermo Fisher Scientific) and stored at -20 °C until IHC. This method resulted in a small labelling improvement (see ), however to improve even further we moved on to Method #3 with the rest of included patient samples. Method #3 The remaining 17 dura samples (numbered #14 to #30) included in the study were immediately frozen on dry ice in OCT medium (Richard-Allan Scientific™ Neg-50™, Thermo Fisher Scientific; Cat#6502) after resection and stored at -80 °C until sectioning. For sectioning, the frozen samples were transferred to the cryostat machine. Sectioning was performed similar to Methods #1 and #2, but without the sucrose steps. The sections obtained were either sagittal or coronal with a thickness of 30 μm. These sections were immediately mounted on Superfrost™ Plus microscopy slides (Thermo Fisher Scientific) and stored at -80 °C until IHC. On the day of IHC the slides were taken out from the − 80 °C, allowed to sit at room temperature for 10 min and the sections were fixed on the slide with 0.5% PFA for 10 min. Immunohistochemistry For immunolabeling of lymphatic endothelial cells (LECs) the sections were thawed at room temperature, rinsed 3 times for 5 min each in 0.01 M phosphate buffer saline (PBS). To reduce unspecific binding of antibodies the sections were blocked for 60 min with a blocking solution (10% normal donkey serum, 1% bovine serum albumin (BSA), 0.5% Triton X-100 in PBS). Afterwards they were incubated in blocking solution containing 0.01% sodium azide with the primary antibody against lymphatic vessel endothelial hyaluronan receptor 1 protein (LYVE-1) (polyclonal rabbit anti-LYVE1: Abcam cat#ab33682, RRID:AB_881387, 1:500 dilution) or podoplanin (PDPN) (GeneTex cat#50,043, RRID:AB_11166297, 1:1000 dilution) overnight at 4 °C. On the following day the sections were rinsed 3 times for 10 min each in 0.01 M PBS following incubation in the blocking solution with the secondary antibody (Cy3 donkey anti-rabbit; Jackson ImmunoResearch Labs; Cat#:711-165-152; RRID:AB_2307443, 1:250 dilution) for 30 min and another 30 min with the secondary antibody plus DyLight® 649 conjugated tomato lectin (LEL, TL; Vector labs; Cat#: DL-1178) to stain blood vessels. The sections were then rinsed again three times for 10 min each in 0.01 M PBS. Afterwards, Hoechst 33,258 was used for nuclear staining (Thermo Fisher Scientific; Cat#: H3569; RRID:AB_2651133, 1:5000 dilution) for five minutes. Lastly the sections were rinsed in distilled water, twice for five minutes, and coverslipped using ProLong™ Gold Antifade Mountant (Thermo Fisher Scientific; Cat#: P36934; RRID:SCR_015961). Prior to using DyLight® 649 conjugated tomato lectin, mouse antibodies against CD31 were used as blood vasculature marker incubated in conjunction with the antibody against LYVE-1 and the next day incubated with Cy2 donkey anti-mouse secondary antibody. However we opted for continuation of lectin instead of CD31 (see section). As negative controls, some sections were incubated in the same solution without the primary anti-LYVE-1/PDPN antibody. For positive controls, mouse liver and spleen samples of C57/BL6 genetic background were dissected immediately after cervical dislocation and frozen on dry ice. These mouse samples were prepared similarly to Method #3. Since liver is abundant in lymphatic vessels while the spleen lacks them, they serve adequately as controls for antibody specificity. Antigen Retrieval The following procedures were followed for antigen retrieval. The sections were washed in 0.01 PBS 3 times for 5 min followed by immersion in 0.2 M Tris-HCl for 30 min at 95 °C. Afterwards, the IHC protocol described above was done as described. The IHC protocol described above was done as described; however, all solutions were prepared with 1% Triton X-100 instead of 0.5%. The sections were washed in 0.01 PBS with Tween20 (PBST) 2 times for 5 min followed by application of 80% formic acid to each section on the glass slide for 2 min at 95 °C. Afterwards, the sections were rinsed in PBST 3 for 5 min and the IHC protocol described above was done as described. The sections were washed in 0.01 PBS 3 times for 5 min followed by immersion in 1 mM EDTA for 30 min at 90 °C. Afterwards, the IHC protocol described above was done as described. For sodium citrate buffer with heat induced antigen retrieval, the sections were washed in 0.01 PBS 3 times for 5 min followed by rinsing in distilled water. These were then immersed 0.1 M sodium citrate buffer pH: 6.0 and microwaved for 2:30 min at 700 W plus 5 min at 350 W. Sections were allowed to cool to room temperature rinsed again in distilled water and then blocked and incubated with primary antibody as described above. Microscopy and measurements High-resolution images were acquired using a LSM 710/700 confocal microscope (Carl Zeiss Microscopy) at magnification of 20x or 63x w/oil immersion objective. To compare the tissues processed with different fixation methods, the images were acquired using the same settings. They were processed using the Zen Blue software (Carl Zeiss Microscopy). In some instances for 3D visualization, a Z-stack image was acquire with a slice thickness between 13 and 19 μm depending on the thickness of the dura sample. The measurement of lymphatic vessels was performed using the same software in the sections were most surface area was visually labelled (see limitations in Discussion). To distinguish between cranial and arachnoid side of the dura mater, overall collagen appearance and cellular density was used (Fig. 1). Towards the cranial side of the dura, collagen fibers are oriented more uniformly with a low density of fibroblasts and cells while the arachnoid side exhibits less organized collagen fibers due to the high density of cells and structures like vessels . Statistics The statistical analyses were performed using SPSS version 27 (IBM Corporation, Armonk, NY, USA). Differences between groups were determined by analysis of variance (ANOVA) with post-hoc Bonferroni corrections for continuous variables and the Chi-square test for categorical variables. Two-tailed p-values of less than 5% were accepted as statistically significant.
We included patients with iNPH who underwent shunt surgery, and who followed the ordinary clinical routine for shunt surgery within the department of neurosurgery. Notably, it was beyond the scope of this study to compare meningeal lymphatic vessels in iNPH versus other diseases or control subjects.
The dura mater was sampled during ventriculoperitoneal (VP) shunt surgery based on clinical routine at the department of neurosurgery in Oslo University Hospital – Rikshospitalet. The trepanation necessary for the surgical procedure and tissue sampling was performed 12 cm caudally from the glabella and 1–2 cm lateral to the midline (Fig. ). The surgical procedure was performed using an operating microscope. A burr hole less than 1 cm in diameter was made and the underlying dura exposed. With a knife, the dura was incised circumferential (diameter about 5 mm) in its entire thickness (about 1 mm) and removed in one piece. Once extracted, the sample was prepared with one of three methods. Method #1 Six dura samples (numbered #1 to #6) were immersed in 4% paraformaldehyde (PFA). The following day the samples were transferred for preservation to a diluted (0.4%) PFA solution. One day prior to sectioning, they were cryoprotected in graded sucrose solutions (10% for 30 min, 20% for one hour and 30% sucrose overnight in 0.1 M phosphate buffer). For sectioning the samples were frozen on dry ice in OCT medium (Richard-Allan Scientific™ Neg-50™, Thermo Fisher Scientific; Cat#6502) and sectioning was performed in a Cryostat NX70 from Thermo Scientific with vacutome. It was not possible to determine the rostro-caudal or medio-lateral orientation in the horizontal plane of the tissue. Therefore, the sections obtained were either sagittal or coronal and 16 μm thick. Cranial (outer) vs. arachnoid (inner) orientation in the vertical plane of the tissue was deciphered through microscopy (See “Microscopy and measurements”). These sections were immediately mounted on Superfrost™ Plus microscopy slides (Thermo Fisher Scientific) and stored at -20 °C until immunohistochemistry (IHC). This fixation method did not produce any staining (see Discussion) and therefore the experiments with these samples were repeated with antigen retrieval methods (described below). Antigen retrieval did not result in any significant improvement and this lead to trying Method#2 with the subsequent patient samples. Method #2 As with Method #1, seven dura samples (numbered #7 to #13) were immersed in 0.5% PFA immediately after resection and was fixed overnight at 4 °C. One day prior to sectioning, they were cryoprotected in graded sucrose solutions, as mentioned above. Sectioning was performed as described for Method #1. The thickness of the sections was modified to 30 μm to prevent washing off the glass slide, which was frequent with Method #1 particularly after antigen retrieval. These sections were immediately mounted on Superfrost™ Plus microscopy slides (Thermo Fisher Scientific) and stored at -20 °C until IHC. This method resulted in a small labelling improvement (see ), however to improve even further we moved on to Method #3 with the rest of included patient samples. Method #3 The remaining 17 dura samples (numbered #14 to #30) included in the study were immediately frozen on dry ice in OCT medium (Richard-Allan Scientific™ Neg-50™, Thermo Fisher Scientific; Cat#6502) after resection and stored at -80 °C until sectioning. For sectioning, the frozen samples were transferred to the cryostat machine. Sectioning was performed similar to Methods #1 and #2, but without the sucrose steps. The sections obtained were either sagittal or coronal with a thickness of 30 μm. These sections were immediately mounted on Superfrost™ Plus microscopy slides (Thermo Fisher Scientific) and stored at -80 °C until IHC. On the day of IHC the slides were taken out from the − 80 °C, allowed to sit at room temperature for 10 min and the sections were fixed on the slide with 0.5% PFA for 10 min.
Six dura samples (numbered #1 to #6) were immersed in 4% paraformaldehyde (PFA). The following day the samples were transferred for preservation to a diluted (0.4%) PFA solution. One day prior to sectioning, they were cryoprotected in graded sucrose solutions (10% for 30 min, 20% for one hour and 30% sucrose overnight in 0.1 M phosphate buffer). For sectioning the samples were frozen on dry ice in OCT medium (Richard-Allan Scientific™ Neg-50™, Thermo Fisher Scientific; Cat#6502) and sectioning was performed in a Cryostat NX70 from Thermo Scientific with vacutome. It was not possible to determine the rostro-caudal or medio-lateral orientation in the horizontal plane of the tissue. Therefore, the sections obtained were either sagittal or coronal and 16 μm thick. Cranial (outer) vs. arachnoid (inner) orientation in the vertical plane of the tissue was deciphered through microscopy (See “Microscopy and measurements”). These sections were immediately mounted on Superfrost™ Plus microscopy slides (Thermo Fisher Scientific) and stored at -20 °C until immunohistochemistry (IHC). This fixation method did not produce any staining (see Discussion) and therefore the experiments with these samples were repeated with antigen retrieval methods (described below). Antigen retrieval did not result in any significant improvement and this lead to trying Method#2 with the subsequent patient samples.
As with Method #1, seven dura samples (numbered #7 to #13) were immersed in 0.5% PFA immediately after resection and was fixed overnight at 4 °C. One day prior to sectioning, they were cryoprotected in graded sucrose solutions, as mentioned above. Sectioning was performed as described for Method #1. The thickness of the sections was modified to 30 μm to prevent washing off the glass slide, which was frequent with Method #1 particularly after antigen retrieval. These sections were immediately mounted on Superfrost™ Plus microscopy slides (Thermo Fisher Scientific) and stored at -20 °C until IHC. This method resulted in a small labelling improvement (see ), however to improve even further we moved on to Method #3 with the rest of included patient samples.
The remaining 17 dura samples (numbered #14 to #30) included in the study were immediately frozen on dry ice in OCT medium (Richard-Allan Scientific™ Neg-50™, Thermo Fisher Scientific; Cat#6502) after resection and stored at -80 °C until sectioning. For sectioning, the frozen samples were transferred to the cryostat machine. Sectioning was performed similar to Methods #1 and #2, but without the sucrose steps. The sections obtained were either sagittal or coronal with a thickness of 30 μm. These sections were immediately mounted on Superfrost™ Plus microscopy slides (Thermo Fisher Scientific) and stored at -80 °C until IHC. On the day of IHC the slides were taken out from the − 80 °C, allowed to sit at room temperature for 10 min and the sections were fixed on the slide with 0.5% PFA for 10 min.
For immunolabeling of lymphatic endothelial cells (LECs) the sections were thawed at room temperature, rinsed 3 times for 5 min each in 0.01 M phosphate buffer saline (PBS). To reduce unspecific binding of antibodies the sections were blocked for 60 min with a blocking solution (10% normal donkey serum, 1% bovine serum albumin (BSA), 0.5% Triton X-100 in PBS). Afterwards they were incubated in blocking solution containing 0.01% sodium azide with the primary antibody against lymphatic vessel endothelial hyaluronan receptor 1 protein (LYVE-1) (polyclonal rabbit anti-LYVE1: Abcam cat#ab33682, RRID:AB_881387, 1:500 dilution) or podoplanin (PDPN) (GeneTex cat#50,043, RRID:AB_11166297, 1:1000 dilution) overnight at 4 °C. On the following day the sections were rinsed 3 times for 10 min each in 0.01 M PBS following incubation in the blocking solution with the secondary antibody (Cy3 donkey anti-rabbit; Jackson ImmunoResearch Labs; Cat#:711-165-152; RRID:AB_2307443, 1:250 dilution) for 30 min and another 30 min with the secondary antibody plus DyLight® 649 conjugated tomato lectin (LEL, TL; Vector labs; Cat#: DL-1178) to stain blood vessels. The sections were then rinsed again three times for 10 min each in 0.01 M PBS. Afterwards, Hoechst 33,258 was used for nuclear staining (Thermo Fisher Scientific; Cat#: H3569; RRID:AB_2651133, 1:5000 dilution) for five minutes. Lastly the sections were rinsed in distilled water, twice for five minutes, and coverslipped using ProLong™ Gold Antifade Mountant (Thermo Fisher Scientific; Cat#: P36934; RRID:SCR_015961). Prior to using DyLight® 649 conjugated tomato lectin, mouse antibodies against CD31 were used as blood vasculature marker incubated in conjunction with the antibody against LYVE-1 and the next day incubated with Cy2 donkey anti-mouse secondary antibody. However we opted for continuation of lectin instead of CD31 (see section). As negative controls, some sections were incubated in the same solution without the primary anti-LYVE-1/PDPN antibody. For positive controls, mouse liver and spleen samples of C57/BL6 genetic background were dissected immediately after cervical dislocation and frozen on dry ice. These mouse samples were prepared similarly to Method #3. Since liver is abundant in lymphatic vessels while the spleen lacks them, they serve adequately as controls for antibody specificity.
The following procedures were followed for antigen retrieval. The sections were washed in 0.01 PBS 3 times for 5 min followed by immersion in 0.2 M Tris-HCl for 30 min at 95 °C. Afterwards, the IHC protocol described above was done as described. The IHC protocol described above was done as described; however, all solutions were prepared with 1% Triton X-100 instead of 0.5%. The sections were washed in 0.01 PBS with Tween20 (PBST) 2 times for 5 min followed by application of 80% formic acid to each section on the glass slide for 2 min at 95 °C. Afterwards, the sections were rinsed in PBST 3 for 5 min and the IHC protocol described above was done as described. The sections were washed in 0.01 PBS 3 times for 5 min followed by immersion in 1 mM EDTA for 30 min at 90 °C. Afterwards, the IHC protocol described above was done as described. For sodium citrate buffer with heat induced antigen retrieval, the sections were washed in 0.01 PBS 3 times for 5 min followed by rinsing in distilled water. These were then immersed 0.1 M sodium citrate buffer pH: 6.0 and microwaved for 2:30 min at 700 W plus 5 min at 350 W. Sections were allowed to cool to room temperature rinsed again in distilled water and then blocked and incubated with primary antibody as described above.
High-resolution images were acquired using a LSM 710/700 confocal microscope (Carl Zeiss Microscopy) at magnification of 20x or 63x w/oil immersion objective. To compare the tissues processed with different fixation methods, the images were acquired using the same settings. They were processed using the Zen Blue software (Carl Zeiss Microscopy). In some instances for 3D visualization, a Z-stack image was acquire with a slice thickness between 13 and 19 μm depending on the thickness of the dura sample. The measurement of lymphatic vessels was performed using the same software in the sections were most surface area was visually labelled (see limitations in Discussion). To distinguish between cranial and arachnoid side of the dura mater, overall collagen appearance and cellular density was used (Fig. 1). Towards the cranial side of the dura, collagen fibers are oriented more uniformly with a low density of fibroblasts and cells while the arachnoid side exhibits less organized collagen fibers due to the high density of cells and structures like vessels .
The statistical analyses were performed using SPSS version 27 (IBM Corporation, Armonk, NY, USA). Differences between groups were determined by analysis of variance (ANOVA) with post-hoc Bonferroni corrections for continuous variables and the Chi-square test for categorical variables. Two-tailed p-values of less than 5% were accepted as statistically significant.
Patients Thirty iNPH patients who underwent shunt surgery were included in the study. Demographic data are shown in Table . The dura biopsy, sized about 5 mm, was obtained from the same place as the ventricular catheter was inserted during shunt surgery, on average 16.1 ± 4.5 mm lateral to the superior sagittal sinus (Table ). Although different fixation protocols for dura mater were used, patients did not differ significantly concerning age or gender group, and distance from superior sagittal sinus was not different between groups. The biopsy was not accompanied with adverse events. Impact of tissue processing and preparation In dura specimens fixated with PFA 4% (Method #1), we found no evidence of lymphatic structures despite trying several antigen retrieval methods (Table ). Reducing the strength of fixative to PFA 0.5% (Method #2) enabled visualization of lymphatic structures in 4/6 (67%) individuals; lymphatic vessels together with blood vessels in 1/6 (17%), lymphatic vessels remote from blood vessels in 2/6 (33%) and cluster-like lymphatic structures in 1/6 (17%) subjects (Table ). On the other hand, fixation with the freezing method (Method #3) demonstrated lymphatic structures in 16/17 (94%) subjects, with lymphatic vessels together with blood vessels in 10/17 (59%), lymphatic vessels remote from blood vessels in 10/17 (59%) and cluster-like lymphatic structures in 7/17 (41%) subjects (Table ). All the images were acquired using the same microscopic settings for comparing the different fixation methods. Hence, the fixation protocol is crucial for the identification of meningeal lymphatic vessels (Fig. ). The results produced by different immunohistochemistry protocols for lymphatic vessels for Methods #1–3 are further illustrated in Supplementary Fig. 1. Method #3, however, also produced an apparently weaker blood vessel staining that led us to use lectin instead of CD31 due to simplicity during IHC. Immunohistochemistry of meningeal lymphatic vessels The meningeal lymphatic vessels were found throughout the entire thickness of dura mater, but seem to be more predominant towards the arachnoid (Fig. ). Meningeal lymphatic vessels were seen close to either blood vessels (Fig. ; Supplementary Movie 1) or remote from blood vessels (Fig. ; Supplementary Movie 2). There were no apparent differences between these lymphatic vessels. The lymphatic vessels nearby blood vessels had a wall thickness 5.3 ± 2.8 μm, lumen diameter 27.6 ± 12.7 μm, and total width of 44.3 ± 19.7 μm. Similarly, the lymphatic vessels remote from blood vessels had a wall thickness 7.2 ± 5.0 μm, lumen diameter 27.3 ± 17.6 μm, and total width of 48.0 ± 27.2 μm (Fig. E, G). Expression of LYVE-1 was also found in aggregated cell clusters (Fig. ; Supplementary Movie 3), which we denote LYVE-1-expressing cells. These cell clusters had no apparent expression of vascular endothelial cells. We tested the specificity of LYVE-1 staining, by ruling out unspecific binding from the fluorophore-conjugated secondary antibody and auto-fluorescence (see Methods) (Supplementary Fig. ). To ensure the specificity of the antibody against LYVE-1, we tested for staining in mouse liver as positive control and mouse spleen as negative control (Supplementary Fig. ). The liver exhibited pronounced staining for LYVE-1 antibodies due to the abundance of lymphatic vessels, while no staining was observed in spleen due to lack of lymphatic vessels, validating the specificity of the antibody. In order to validate the lymphatic identity of the LYVE-1 positive immunoreactive structures, we confirmed co-labeling with both markers LYVE-1 and PDPN in lymphatic vessels (Fig. ). The control experiments of specificity of PDPN are shown in Supplementary Fig. 4. Furthermore, we show that the LYVE-1 receptor is expressed in lymphatic endothelial cells but not vascular endothelial cells or red blood cells. In Supplementary Fig. 5, we show blood vessels where vascular endothelial cells have been labeled with Lectin, but also LYVE-1 labeling intramurally (tunica media), highlighting the importance of control-staining not to misidentify blood vessels as lymphatic vessels. The wall thickness was larger in blood vessels (Supplementary Fig. 5).
Thirty iNPH patients who underwent shunt surgery were included in the study. Demographic data are shown in Table . The dura biopsy, sized about 5 mm, was obtained from the same place as the ventricular catheter was inserted during shunt surgery, on average 16.1 ± 4.5 mm lateral to the superior sagittal sinus (Table ). Although different fixation protocols for dura mater were used, patients did not differ significantly concerning age or gender group, and distance from superior sagittal sinus was not different between groups. The biopsy was not accompanied with adverse events.
In dura specimens fixated with PFA 4% (Method #1), we found no evidence of lymphatic structures despite trying several antigen retrieval methods (Table ). Reducing the strength of fixative to PFA 0.5% (Method #2) enabled visualization of lymphatic structures in 4/6 (67%) individuals; lymphatic vessels together with blood vessels in 1/6 (17%), lymphatic vessels remote from blood vessels in 2/6 (33%) and cluster-like lymphatic structures in 1/6 (17%) subjects (Table ). On the other hand, fixation with the freezing method (Method #3) demonstrated lymphatic structures in 16/17 (94%) subjects, with lymphatic vessels together with blood vessels in 10/17 (59%), lymphatic vessels remote from blood vessels in 10/17 (59%) and cluster-like lymphatic structures in 7/17 (41%) subjects (Table ). All the images were acquired using the same microscopic settings for comparing the different fixation methods. Hence, the fixation protocol is crucial for the identification of meningeal lymphatic vessels (Fig. ). The results produced by different immunohistochemistry protocols for lymphatic vessels for Methods #1–3 are further illustrated in Supplementary Fig. 1. Method #3, however, also produced an apparently weaker blood vessel staining that led us to use lectin instead of CD31 due to simplicity during IHC.
The meningeal lymphatic vessels were found throughout the entire thickness of dura mater, but seem to be more predominant towards the arachnoid (Fig. ). Meningeal lymphatic vessels were seen close to either blood vessels (Fig. ; Supplementary Movie 1) or remote from blood vessels (Fig. ; Supplementary Movie 2). There were no apparent differences between these lymphatic vessels. The lymphatic vessels nearby blood vessels had a wall thickness 5.3 ± 2.8 μm, lumen diameter 27.6 ± 12.7 μm, and total width of 44.3 ± 19.7 μm. Similarly, the lymphatic vessels remote from blood vessels had a wall thickness 7.2 ± 5.0 μm, lumen diameter 27.3 ± 17.6 μm, and total width of 48.0 ± 27.2 μm (Fig. E, G). Expression of LYVE-1 was also found in aggregated cell clusters (Fig. ; Supplementary Movie 3), which we denote LYVE-1-expressing cells. These cell clusters had no apparent expression of vascular endothelial cells. We tested the specificity of LYVE-1 staining, by ruling out unspecific binding from the fluorophore-conjugated secondary antibody and auto-fluorescence (see Methods) (Supplementary Fig. ). To ensure the specificity of the antibody against LYVE-1, we tested for staining in mouse liver as positive control and mouse spleen as negative control (Supplementary Fig. ). The liver exhibited pronounced staining for LYVE-1 antibodies due to the abundance of lymphatic vessels, while no staining was observed in spleen due to lack of lymphatic vessels, validating the specificity of the antibody. In order to validate the lymphatic identity of the LYVE-1 positive immunoreactive structures, we confirmed co-labeling with both markers LYVE-1 and PDPN in lymphatic vessels (Fig. ). The control experiments of specificity of PDPN are shown in Supplementary Fig. 4. Furthermore, we show that the LYVE-1 receptor is expressed in lymphatic endothelial cells but not vascular endothelial cells or red blood cells. In Supplementary Fig. 5, we show blood vessels where vascular endothelial cells have been labeled with Lectin, but also LYVE-1 labeling intramurally (tunica media), highlighting the importance of control-staining not to misidentify blood vessels as lymphatic vessels. The wall thickness was larger in blood vessels (Supplementary Fig. 5).
The present study demonstrates that successful immunohistochemistry of meningeal lymphatic vessels in human dura heavily relies on the fixation procedure. This suggests that the LYVE-1 receptor protein in meningeal lymphatic vessels is sensitive to both concentration and duration of fixation. The presently characterized meningeal lymphatic vessels were either located in close vicinity to blood vessels or remote from blood vessels, and we demonstrated LYVE-1-expressing cells gathered in clusters. We primarily used the lymphatic cell marker lymphatic vessel endothelial hyaluronan receptor 1 (LYVE-1) for immunohistochemistry of meningeal lymphatic vessels. This is a type I integral membrane glycoprotein that binds to soluble and immobilized hyaluronan. To confirm the lymphatic identity of the LYVE-1 reactive structures, we used podoplanin (PDPN) . Other lymphatic markers include vascular endothelial growth factor 3 (VEGFR-3) and chemokine ligand 21 (CCL21). Utilizing these available markers of lymphatic endothelial cells, the vast majority of data is derived from experimental studies in animals . Kutomi and Takeda studied cranial arachnoid granulations of dura in pigs and found the endothelial lining of granulations to be LYVE-1 immunoreactive, indicating they are lymphatic cells. These authors also referred to endothelial-like cell lined gaps in dura towards the subarachnoid space. To the best of our knowledge, this is the first study to report the successful staining of meningeal lymphatic vessels in tissue biopsies from patients. It is, however, highly warranted to get experience with assessment of lymphatic vessels in dura mater of patients, given the marked species differences of the dura mater . To this end, the available human data is derived exclusively from autopsy specimens though the number is low and the results somewhat variable. Louveau et al. reported LYVE-1 staining of lymphatic vessels nearby the superior sagittal sinus in one case, and Absinta reported in three autopsy specimens, dural lymphatic vessels immunoreactive to the antibodies against D2-40, PROX-1, CCL21 and LYVE-1. Visanji et al. in four autopsy specimens visualized dural lymphatic vessels nearby the superior sagittal sinus, utilizing a monoclonal primary antibody to podoplanin, D2-40. The same antibody previously was shown to visualize lymphatic vessels of the dura part of the optic nerve . Another study showed immunoreactive meningeal lymphatic vessels in ten autopsy specimens, with the help of the markers LYVE-1 and PDPN . Moreover, Godman et al. presented 21 autopsy specimens in whom 19/21 subjects showed positive staining for lymphatic vessels, using the antibodies against D2-40 and LYVE-1. These authors differentiated between two kinds of meningeal lymphatic vessels, namely lymphatic vessels comprising a single layer of endothelium with no smooth muscle, unoccupied lumen and LYVE-1 positive (Type 1) and another type of lymphatic vessels with material within the lumen being LYVE-1 negative. Type 1 lymphatic vessels were found in the periosteal and meningeal layers of the dura mater while Type 2 was distributed between the superior sagittal sinus and periosteal layer of the dura mater . However, despite these studies showing immunoreactive meningeal lymphatic vessels, the available literature about lymphatic vessels in humans is not consistent. It seems that when working with autopsy material, a protocol for antigen retrieval is essential in order to do immunostaining of meningeal lymphatic vessels, as more recently demonstrated by Mezey et al. . We were, however, not able to label lymphatic vessels in this way, despite having tried various methods of antigen retrieval. In contrast, using the freezing method, we were able to obtain good staining consistently without the need of antigen retrieval. The discrepancies may stem from variations in the fixation and autopsy procedures performed. Supporting this idea, Johnson examined dura mater of 113 autopsy specimens and reported microscopic fluid channels in 101 of 113 (90%) of the dura specimens; they were on the meningeal side of dura, variable in size, and podoplanin-negative. They observed D2-40 expression, indicative of lymphatic vessels, in only seven autopsies. Interestingly, the lymphatic structures were either linear or organized in clusters, either nearby or remote of blood vessels. More recently, Park et al. in nine autopsy specimens failed to demonstrate D2-40 immunoreactive lymphatic vessels in human dura, and reported a network of dural channels within the parasagittal dura not in association with lymphatic vessels. Based on our present observations, it may be speculated whether the absence or low frequency of immunoreactive lymphatic vessels in these two latter studies was due to the fixation procedure of the autopsy specimens. Alternatively, the microscopic dural channels are non-lymphoid structures, which may compare with the observation that the fluid channels are smaller than the previously reported dural lymphatic vessels. The present data unequivocally demonstrate that the fixation method is essential for the ability to demonstrate meningeal vessels in human dura. Evidently, fixation with PFA 4.0% is too strong and since the dimension of the sample is small, it most likely masks the epitopes in the LYVE-1 receptor and therefore prevents binding of antibodies. The antigen might as well become denatured by PFA 4%. This may be a bit surprising given the common clinical practice to demonstrate extra-cranial lymphatic vessels utilizing the antibody D2-40 . That begs the question whether lymphatic endothelium in the CNS exhibits surface protein isoforms that are more sensitive to fixation procedures than their counterparts in the rest of the body and if such were the case, what functional implications would this have for lymphatic vessels in the meninges? The present study only included patients with iNPH since we here primarily addressed methodological aspects of immunohistochemistry of meningeal lymphatic vessels. Future studies are needed to clarify whether meningeal lymphatic function is abnormal in iNPH, and whether this is reflected in smaller diameter of the meningeal lymphatic vessels. It is worthy to note that iNPH overlaps with Alzheimer’s regarding deposition of amyloid-β within the brain . In this regard, impaired meningeal clearance of amyloid-β might hypothetically be a key factor in iNPH. The present data is not suitable for deciphering the role of meningeal lymphatic function in iNPH, but if meningeal lymphatic dysfunction is a key factor in iNPH pathogenesis, an interesting approach would be lymphatic function enhancement, e.g. utilizing VEGFR3-specific recombinant VEGF-c . Likewise, iNPH is a disease of the elderly and it has been shown that meningeal lymphatic function declines with increasing age. For these reasons, dura biopsies during shunt surgery might provide diagnostic information in this disease. One interesting question is whether the dimensions of meningeal lymphatic vessels in iNPH are smaller than in healthy subjects. The diameter of meningeal lymphatic vessels previously reported in humans vary, and were ranging between 19 and 470 μm in one study , and 7 to 842 μm (average ± stdev: 125 ± 161 μm) in another study . In comparison, the diameter of rodent meningeal lymphatic vessels was in the range 20 to 30 μm . Considering lymphatic vessels both nearby and remote from blood vessels in the present study, the total width of lymphatic vessels was 46.7 ± 24.0 μm (ranges 10.1 to 87.7 μm) with wall thickness 6.6 ± 4.3 μm (ranges 1.7 to 17.1 μm) and lumen diameter 27.4 ± 15.5 μm (ranges 6.7 to 66.2 μm). Accordingly, the dimensions of lymphatic vessels in the present study are smaller than previously reported. It could be argued that the reason behind this is the location of the tissue examined and that lymphatic vessels vary greatly in their size, relative to the distance of the superior sagittal sinus. Lymphatic vessels in the rest of the body vary as well. For example, initial lymphatic capillaries have an approximate diameter of 10–60 μm . Therefore, it is possible that the lymphatic vessels observed in the present study are small collecting capillaries that transport their content into larger lymphatic vessels closer to the superior sagittal sinus. If so, this might suggest that lymphatic vessels become smaller in caliber as they move laterally from the superior sagittal sinus. We cannot decipher from the present data the function of meningeal lymphatic vessels in humans, but the literature is growing concerning the putative role of meningeal lymphatic vessels for normal brain function. An increasing body of evidence points at the role of meningeal lymphatic vessels in immune surveillance . In particular, the more recent observations of passage of fluid and cells between CSF, dura and skull bone marrow are intriguing as they provide for immunological peripheral-central cross talk at the meninges, both under physiological and pathological conditions. Moreover, the evidence is growing that meningeal lymphatic vessels serve a function in clearing CSF and solutes , including metabolic waste products such as soluble amyloid-β , tau and α-synuclein . Hence, impaired or abolished glymphatic/meningeal lymphatic waste clearance of the waste products may play a role in the pathophysiology of Alzheimer’s (soluble amyloid-β and tau) and Parkinson’s diseases (α-synuclein) . Some limitations with the present study should be noted. The statistical comparisons between groups was hampered by a rather low number of cases for Methods #1 and #2. Furthermore, the distance from the superior sagittal sinus differed a few millimeters between the groups (Table ). We do not think this would affect the results as the dura samples were obtained from the same location in all subjects. On the other hand, this also makes it impossible to conclude about variability in expression of lymphatic vessels throughout the dura. We here used the same location for the dura sample, as the aim of this study was to compare the effect of different tissue fixation protocols and their effect on antigenicity for immunohistochemistry. For that purpose, sampling from the same region was optimal. The present dura specimens were obtained from iNPH patients, making it impossible to conclude about meningeal lymphatic vessels in healthy subjects. In addition, we are not able to distinguish between CD45 + LYVE1 + macrophages and CD45- LYVE1 + lymphatic endothelial cells without the aforementioned marker, making it hard to make conclusions about the abundance of lymphatic meningeal vessels in these subjects. Moreover, we cannot conclude whether lymphangiogenesis is up- or down-regulated in iNPH patients. Further studies are required to determine how the distribution of meningeal lymphatic vessels differs throughout the dura in humans. Are they primarily distributed along the dural sinuses? An interesting point is whether the density of lymphatic vessels might differ between different disease categories. Lastly, due to the fixation method in order to preserve as much LYVE-1 antigenicity as possible, lymphatic vessel staining between adjacent sections was highly variable and in some cases the full extension of a lymphatic vessels’ surface in a single section was not labelled completely making it difficult to do systematic measurements and stereology was not possible to perform. Therefore, the measurements done in the present study were taken only from the sections where the vast majority a lymphatic vessel was visualized, which might introduce errors in width/lumen as the 3D structure changes throughout the tissue.
The present observations show that visualization of meningeal vessels in human dura is highly sensitive to the fixation method. We here for the first time demonstrate meningeal lymphatic vessels in patients, and show that human meningeal lymphatic vessels are located both in intimate vicinity to blood vessels as well as remote, and LYVE-1-expressing cells were organized in clusters. The present observations open for further studies on the characteristics and implications of human meningeal lymphatic vessels.
Below is the link to the electronic supplementary material. Supplementary movie 1. Meningeal lymphatic vessels in close proximity to blood vessels. Lymphatic vessels are visualized in green color and blood vessels in red color. See Fig. 4D. Supplementary movie 2. Meningeal lymphatic vessels in distance from blood vessels. Lymphatic vessels are visualized in green color. See Fig. 5D. Supplementary movie 3. Cluster of LYVE-1-expressing cells. Cluster of LYVE-1-expressing cells are visualized in green color. See Fig. 6D. Supplementary Material 4
|
Needs assessment for enhancing pediatric clerkship readiness | ed26f592-c3c0-42e4-9312-6f3b31e6ff78 | 10044806 | Internal Medicine[mh] | Early clinical exposure and clinical skill development in medical school training is important for students to be successful during clerkships. Inadequate preparation may impact students’ ability to learn key skills, knowledge, and behaviors. A study examining clerkship directors’ views regarding pre-clerkship preparation in core clinical competencies found that 80% of clerkship directors felt that proficiency is needed in communication skills, professionalism, and interviewing/physical exam skills before entering the clerkships . Historically, medical schools have taught a broad overview of basic sciences and clinical skills during the pre-clinical curricula. There is striking variability on how and when pediatric content and skills are taught to medical students . In a previously published study, we found that more than one third of 3 rd year medical students completing their pediatric clerkship at one of four U.S. medical schools did not feel well-prepared for the clerkship . In addition, one third of students specifically felt unprepared to perform pediatric physical exam skills. However, feeling unprepared is not limited to the pediatric clerkship . In qualitative studies, students report struggling with several tasks, including “understanding roles and responsibilities, adjusting to clinical cultures, performing clinical skills, learning the logistics of clinical settings, and encountering frequent changes in staff, settings, and content.” In addition to these challenges, clerkship directors also noted that students had difficulties applying their medical knowledge to clinical scenarios, demonstrating clinical reasoning, and being self-directed learners . In a quantitative assessment, Wenrich et al. compared the expectations of pre-clinical faculty, clerkship faculty, and third year medical students regarding clinical skills preparation for clerkships at a single North American medical school. Students had higher expectations than clerkship faculty for both basic and advanced clinical skills, as well as higher expectations than pre-clinical faculty for advanced clinical skills. Additionally, for many basic clinical skills, pre-clinical faculty had higher expectations of students compared to clerkship faculty . As medical schools undergo curricular reform, there are few data to guide what content or skills training should be included in the pre-clinical arena. Although previously published data showed that students did not feel well prepared for the pediatric clerkship including the competencies of knowledge, communication, and physical exam skills, we felt it was important to distinguish whether this applied to pediatrics in particular, or if students feel unprepared for clerkships in general. This study aims to compare how students perceive their preparedness for each of their clinical clerkships. Additionally, based on student responses regarding decreased preparedness in physical exam skills, we explored this further by surveying pediatric clerkship directors and clinical skills curriculum directors. We asked these two groups of educators representing the curricular experts who typically lead instruction on physical exam skills and pediatric clinical skills, assessing the expected competency levels of students’ pediatric physical exam skills prior to starting the pediatric clerkship.
This study was a product of the Pre-clinical and Clinical Skills Collaborative of the Council on Medical Student Education in Pediatrics (COMSEP). As part of the development.process for a standard pre-clinical pediatric curriculum, the authors reviewed the literature to identify pediatric physical exam skills as a particular area for study. Medical student clerkship preparedness survey development The survey was developed using previously published methods . Students were asked to “rate how the training in the first two years of medical school prepared them for: 1) the clerkship overall, 2) the communication skills needed for the clerkship, 3) the physical exam skills needed for the clerkship, and 4) the medical knowledge needed for this clerkship.” The survey used a 1–5 Likert (1 = poor, 5 = excellent). The clerkship directors in pediatrics, internal medicine, and obstetrics-gynecology at three participating institutions, and for family medicine and general surgery at two of these institutions, distributed the survey to all medical students from November 2013 to February 2015. All three participating institutions have current LCME accreditation; one is in a rural setting, one in a suburban/small city and one in a major city. All three schools had traditional curricular structures with a two-year pre-clinical phase followed by a two-year clinical education phase. Over the study period, School A had 120 medical students complete the third year, School B had 112 students and School C had 75 students. All surveys were administered on the last day of the rotation either as part of the electronic final evaluation of the rotation or a paper survey, per the school’s customary process for end-of-clerkship block student evaluation. All responses were anonymous and voluntary, and there were no reminders or incentives. A total of 307 students across three medical schools were surveyed during the study period. Due to the timing of the study and variability in student clerkship schedules, not all students completed all clerkships during the study period, and the family medicine and surgery surveys were only conducted at two of the three schools. This led to a different total number of students surveyed for each type of clerkship. IRB approval was obtained at each institution. Pediatric and clinical skills educator physical exam survey development A questionnaire was designed for pediatric clerkship and clinical skills curriculum directors to serve as a needs assessment on their perceptions of the competence expected of students in different components of the physical exam in pediatric patients, prior to the start of their pediatric clerkship. The survey was piloted by the authors, and subsequently by separate survey review teams of COMSEP, and of the Directors of Clinical Skills Education (DOCS). After incorporating recommendations from each of these teams, the survey was finalized and distributed to the listservs and member lists of COMSEP, representing 504 pediatric clerkship educators in the United States and Canada, and DOCS, representing 270 clinical skills educators (including a wide variety of medical specialties) in the United States and Canada at the time of distribution. (Supplemental Fig. ) We specifically invited the subset of these educators who are pediatric clerkship directors and clinical skills course directors, respectively, to complete the survey. Additionally, the first question of the survey asked participants to identify their role in medical student education. Only the responses of those who identified as a pediatric clerkship director or clinical skills course director were included in our study, and the responses of those who serve as other medical student educators were excluded. These educators were asked “What level of competence should medical students have in performing pediatric specific physical exam skills on the day they start their pediatric clerkship” for each of the organ system exams specified, as well as the newborn exam and development assessment skills. The rating scale ranged from “no knowledge of exam maneuver” to “ability to perform maneuver.” They were also asked, and to select all that apply, “When are pediatric specific exam skills taught for all students in your school” and “When should pediatric specific exam skills be taught for all students in your school.” Ethics approval for the survey of pediatrics clerkship and clinical skills directors was obtained at Queen’s University’s Health Sciences & Affiliated Teaching Hospitals Research Ethics Board (HSREB). Analysis Student survey responses were analyzed using Chi-Squared comparing total Fair-Poor responses (“inadequately prepared”) to total Good-Very Good–Excellent (“well prepared”) responses. Faculty survey responses were analyzed with paired t-test. p < 0.05 was considered significant.
The survey was developed using previously published methods . Students were asked to “rate how the training in the first two years of medical school prepared them for: 1) the clerkship overall, 2) the communication skills needed for the clerkship, 3) the physical exam skills needed for the clerkship, and 4) the medical knowledge needed for this clerkship.” The survey used a 1–5 Likert (1 = poor, 5 = excellent). The clerkship directors in pediatrics, internal medicine, and obstetrics-gynecology at three participating institutions, and for family medicine and general surgery at two of these institutions, distributed the survey to all medical students from November 2013 to February 2015. All three participating institutions have current LCME accreditation; one is in a rural setting, one in a suburban/small city and one in a major city. All three schools had traditional curricular structures with a two-year pre-clinical phase followed by a two-year clinical education phase. Over the study period, School A had 120 medical students complete the third year, School B had 112 students and School C had 75 students. All surveys were administered on the last day of the rotation either as part of the electronic final evaluation of the rotation or a paper survey, per the school’s customary process for end-of-clerkship block student evaluation. All responses were anonymous and voluntary, and there were no reminders or incentives. A total of 307 students across three medical schools were surveyed during the study period. Due to the timing of the study and variability in student clerkship schedules, not all students completed all clerkships during the study period, and the family medicine and surgery surveys were only conducted at two of the three schools. This led to a different total number of students surveyed for each type of clerkship. IRB approval was obtained at each institution.
A questionnaire was designed for pediatric clerkship and clinical skills curriculum directors to serve as a needs assessment on their perceptions of the competence expected of students in different components of the physical exam in pediatric patients, prior to the start of their pediatric clerkship. The survey was piloted by the authors, and subsequently by separate survey review teams of COMSEP, and of the Directors of Clinical Skills Education (DOCS). After incorporating recommendations from each of these teams, the survey was finalized and distributed to the listservs and member lists of COMSEP, representing 504 pediatric clerkship educators in the United States and Canada, and DOCS, representing 270 clinical skills educators (including a wide variety of medical specialties) in the United States and Canada at the time of distribution. (Supplemental Fig. ) We specifically invited the subset of these educators who are pediatric clerkship directors and clinical skills course directors, respectively, to complete the survey. Additionally, the first question of the survey asked participants to identify their role in medical student education. Only the responses of those who identified as a pediatric clerkship director or clinical skills course director were included in our study, and the responses of those who serve as other medical student educators were excluded. These educators were asked “What level of competence should medical students have in performing pediatric specific physical exam skills on the day they start their pediatric clerkship” for each of the organ system exams specified, as well as the newborn exam and development assessment skills. The rating scale ranged from “no knowledge of exam maneuver” to “ability to perform maneuver.” They were also asked, and to select all that apply, “When are pediatric specific exam skills taught for all students in your school” and “When should pediatric specific exam skills be taught for all students in your school.” Ethics approval for the survey of pediatrics clerkship and clinical skills directors was obtained at Queen’s University’s Health Sciences & Affiliated Teaching Hospitals Research Ethics Board (HSREB).
Student survey responses were analyzed using Chi-Squared comparing total Fair-Poor responses (“inadequately prepared”) to total Good-Very Good–Excellent (“well prepared”) responses. Faculty survey responses were analyzed with paired t-test. p < 0.05 was considered significant.
Medical students clerkship preparedness survey There were 97 student responses (42% response rate) from the three schools for obstetrics-gynecology, 229 responses for pediatrics (97% response rate), 208 responses for internal medicine (74% response rate), 73 for family medicine (45% response rate), and 63 for surgery (38% response rate). Each clerkship specialty had a different number of students surveyed, as students’ schedules varied within and among schools over the study duration, which included portions of 2 academic years, and the family medicine and surgery surveys were only conducted at two schools. Overall, thirty-three percent of respondents felt their pre-clinical education was “poor” or “fair” in preparing them for their pediatric clerkship. This compared to 4% for their family medicine clerkship, 11% for internal medicine, 30% for surgery, and 31% for their obstetrics-gynecology rotation (Table ). When comparing the percent of student responses of “fair or “poor” from pediatrics with other core clerkships, there were statistically significant differences in all areas assessed (overall, communication skills, physical exam skills and medical knowledge) for pediatrics compared to both internal medicine and family medicine ( p < 0.005). For the physical exam skills, there were statistically significant higher numbers of “fair” and “poor” responses for pediatrics compared to all other core clerkships surveyed ( p < 0.05) (Fig. ). Pediatric and clinical skills educator physical exam survey A total of 44 Clinical Skills Course directors and 91 Pediatric Clerkship directors completed the survey. For the majority of organ system physical exams, both clinical skills and pediatric educators felt students should have both knowledge of and some ability to perform the exam maneuvers on pediatric patients prior to the start of their pediatric clerkship (Fig. ). Greater competence was expected by both groups of educators for the lung, cardiovascular, and abdominal exams, in which many educators felt students should have the full ability to perform these exams. With respect to the fundoscopic and genitourinary exam, many educators felt a lower competence was needed and that only knowledge of the exam in children was needed rather than some ability to perform the exam. There were no significant differences between pediatric clerkship and clinical skills directors. With respect to the newborn specific physical exam and childhood development assessment skills, both pediatric and clinical skills educators felt students should have some knowledge of the physical exam prior to their pediatric clerkship, but that students did not need to demonstrate an ability to perform the skill pre-clerkship (Fig. ). Notably, there was a statistically significant difference in that more clinical skills directors felt that students ought to have some ability to perform development assessment skills as compared to pediatric clerkship directors ( p < 0.0001). We asked both groups when pediatric specific exam skills are taught for all students in their school, as well as when they should be taught for all students in their school (Fig. ). Most clinical skills directors (80%) felt that pediatric specific exam skills were taught in the pre-clinical clinical skills course, whereas only 49% of pediatric clerkship directors identified this ( p < 0.002). On the other hand, both groups felt the longitudinal clinical skills course was a curriculum where these skills should be taught (85% of clinical skills course directors and 70% of pediatric clerkship directors, no significant differences). Of note, the minority of both sets of educators (17% of pediatric clerkship, 5% of clinical skills directors) identified a transition time or orientation before beginning any clerkships as a place where these skills were being taught, but close to half of pediatric clerkship directors (47%) thought this would be an appropriate time compared to 25% of clinical skills directors ( p < 0.002). Both groups felt pediatric physical exam skills should continue to be taught during the pediatric clerkship through a mixture of a skills session during clerkship orientation or during the clerkship, and through experiential learning with patients and preceptors.
There were 97 student responses (42% response rate) from the three schools for obstetrics-gynecology, 229 responses for pediatrics (97% response rate), 208 responses for internal medicine (74% response rate), 73 for family medicine (45% response rate), and 63 for surgery (38% response rate). Each clerkship specialty had a different number of students surveyed, as students’ schedules varied within and among schools over the study duration, which included portions of 2 academic years, and the family medicine and surgery surveys were only conducted at two schools. Overall, thirty-three percent of respondents felt their pre-clinical education was “poor” or “fair” in preparing them for their pediatric clerkship. This compared to 4% for their family medicine clerkship, 11% for internal medicine, 30% for surgery, and 31% for their obstetrics-gynecology rotation (Table ). When comparing the percent of student responses of “fair or “poor” from pediatrics with other core clerkships, there were statistically significant differences in all areas assessed (overall, communication skills, physical exam skills and medical knowledge) for pediatrics compared to both internal medicine and family medicine ( p < 0.005). For the physical exam skills, there were statistically significant higher numbers of “fair” and “poor” responses for pediatrics compared to all other core clerkships surveyed ( p < 0.05) (Fig. ).
A total of 44 Clinical Skills Course directors and 91 Pediatric Clerkship directors completed the survey. For the majority of organ system physical exams, both clinical skills and pediatric educators felt students should have both knowledge of and some ability to perform the exam maneuvers on pediatric patients prior to the start of their pediatric clerkship (Fig. ). Greater competence was expected by both groups of educators for the lung, cardiovascular, and abdominal exams, in which many educators felt students should have the full ability to perform these exams. With respect to the fundoscopic and genitourinary exam, many educators felt a lower competence was needed and that only knowledge of the exam in children was needed rather than some ability to perform the exam. There were no significant differences between pediatric clerkship and clinical skills directors. With respect to the newborn specific physical exam and childhood development assessment skills, both pediatric and clinical skills educators felt students should have some knowledge of the physical exam prior to their pediatric clerkship, but that students did not need to demonstrate an ability to perform the skill pre-clerkship (Fig. ). Notably, there was a statistically significant difference in that more clinical skills directors felt that students ought to have some ability to perform development assessment skills as compared to pediatric clerkship directors ( p < 0.0001). We asked both groups when pediatric specific exam skills are taught for all students in their school, as well as when they should be taught for all students in their school (Fig. ). Most clinical skills directors (80%) felt that pediatric specific exam skills were taught in the pre-clinical clinical skills course, whereas only 49% of pediatric clerkship directors identified this ( p < 0.002). On the other hand, both groups felt the longitudinal clinical skills course was a curriculum where these skills should be taught (85% of clinical skills course directors and 70% of pediatric clerkship directors, no significant differences). Of note, the minority of both sets of educators (17% of pediatric clerkship, 5% of clinical skills directors) identified a transition time or orientation before beginning any clerkships as a place where these skills were being taught, but close to half of pediatric clerkship directors (47%) thought this would be an appropriate time compared to 25% of clinical skills directors ( p < 0.002). Both groups felt pediatric physical exam skills should continue to be taught during the pediatric clerkship through a mixture of a skills session during clerkship orientation or during the clerkship, and through experiential learning with patients and preceptors.
To our knowledge, this is the first study to compare student perceptions of preparedness for their pediatric clerkship to other core clerkships based on their pre-clinical training. Close to one-third of medical students felt their pre-clinical education did not prepare them well for their pediatric clerkship overall. In the areas of communication skills, physical exam, and medical knowledge, medical students felt less prepared for pediatrics compared to family medicine and internal medicine. This may reflect the focus on adult medicine in the areas of medical knowledge, clinical skills, and communication skills that are typically taught in the early medical school curricula. Medical students specifically reported feeling less prepared for pediatric physical exam skills compared to all other specialties surveyed. A similar number of students felt “unprepared” overall for pediatrics, obstetrics-gynecology and surgery. This is not surprising since a substantial component of learning for the medical student in obstetrics-gynecology and surgery relates to the operating room and surgical skills. Assisting in the labor and delivery suite is also a novel experience for most students . The surgical literature also notes that a lack of any exposure to surgical skills in pre-clinical education can negatively impact students in various career choices. Exposure to select skills prior to the clerkship experiences may have a positive effect on early skill acquisition and enable more time-efficient training in the clerkship in an era where there is compression of clerkship clinical experiences . In 2005, the Association of American Medical Colleges (AAMC) released a report titled “Recommendations for Clinical Skills Curricula for Undergraduate Medical Education.” In this report, the committee stated that the competencies for undergraduate medical education should provide fundamental clinical competencies that provide the foundation for later and more sophisticated levels of clinical practice . One recommendation in particular states that students should demonstrate “The ability to provide clinical care within the practical context of a patient’s age, gender, personal preferences, family, health literacy, culture, religious perspective, and their economic circumstances.” Some may feel that most pediatric-specific physical exam skills can be taught during the actual clerkship and do not need significant pre-clerkship attention. However, as the duration of clerkships has continued to shorten over the past decade, the time to develop these experiences within the context of the clerkship becomes more challenging. According to the AAMC, the average number of weeks for the pediatric clerkship at LCME accredited medical schools has declined, from 7.4 weeks in 2010–2011 to 6.6 weeks in 2018–2019 ( https://www.aamc.org/data-reports/curriculum-reports/report/curriculum-reports ). In response to this and with ongoing limited resources, some schools have gone to a longitudinal integrated clerkship approach with favorable results . This application as a way to improve overall clinical skills, knowledge and overall experience in specialties like pediatrics, obstetrics-gynecology or general surgery is unclear but suggests further study. When comparing the expectations of pediatric clerkship directors and clinical skills course directors regarding pediatric physical exam skill preparation, we found that for a wide array of organ system physical exam skills, pediatric clerkship directors and clinical skills course directors agreed that students should have knowledge of and some ability to perform the physical exam on children. On the other hand, both sets of directors felt that students ought to have knowledge of the newborn specific exam and developmental assessment skills, but on average did not feel it was necessary for them to have ability to perform these skills prior to the pediatrics clerkship. Notably, clinical skills directors believed pre-clinical students should demonstrate more competence in these areas as compared to pediatric clerkship directors. These differences may relate to their contexts in clinical education, where clinical skills educators aim for a certain level of comprehensive competence across many disciplines in their curricula, pediatric clerkship directors could be identifying specific skills achieved particularly well in the context of their specialty. Of note, Wenrich et al. also found that pre-clinical faculty had higher expectations of students’ preparedness in advanced clinical skills compared to clerkship directors . These differences in expectations, as well as those of third year medical students, could be an area for future exploration. Interestingly, only ~ 50% of pediatric clerkship directors felt that pediatric clinical skills were being taught in longitudinal clinical skills courses, as compared to ~ 80% of clinical skills course directors. Meanwhile, 70 and 85%, respectively, believe these skills should be taught in the pre-clinical clinical skills course. This discrepancy could be related to previously observed challenges in integrating and collaboration among pre-clinical and clinical curricular leaders. Mechanisms for collaborative sharing and curricular integration may reconcile this . Additionally, about half of pediatric clerkship directors believe pediatric physical exam skills ought to be taught in a transition to clerkship curriculum, whereas only ~ 25% of clinical skills course directors feel that way. If these skills can be successfully integrated into the longitudinal clinical skills course, then this timing could serve as an efficient opportunity for spaced learning and deliberately reinforcing these skills in a context closer to the pediatric clerkship. Both sets of educators identify the pediatric clerkship itself as an optimal context for pediatric physical exam skills education, recognizing that certain exams and achievement of greater competence to perform these skills is best achievable in the clinical specialty setting. An intriguing future direction could be a collaboration among clinical skills and pediatric medical student educators on a shared pre-clinical pediatric physical exam skills curriculum with an evaluation of the student experience and student performance by students and faculty. Since both sets of educators value this experience, a major challenge could be finding and prioritizing curricular time, especially given trends over the last decade for schools to shorten the duration of their pre-clinical curriculum ( https://www.aamc.org/data-reports/curriculum-reports/report/curriculum-reports ). Another major challenge remains identifying children for medical student physical exam skills education. Different approaches to this challenge may work best at different medical schools. Among the possible solutions include scheduled pre-clinical experiences at pediatric outpatient, inpatient and/or pediatric specialty sites with goals for students to learn and practice physical exam skills under supervision by pediatric providers . Another opportunity could be through collaboration with local daycare and school settings, where medical students could provide service and volunteer support (e.g. at school based clinics, vaccine clinics, health screenings sessions), while gaining opportunity to work with children of different ages and practice supervised physical exams. Simulated experiences provide a rich opportunity for students to learn and practice clinical skills on adult simulated patients. Creating analogous simulated opportunities with assented children could be another strategy. There are several limitations to our study. Only three medical schools participated in the student survey. Schools were well sampled as far as size and location (city vs. rural) but all schools are geographically in the northeast part of the United States. The student survey results are also based on student perception and do not necessarily reflect student ability. We also only surveyed at schools who have traditional curricula with two pre-clinical education years followed by two clinical education years. With many schools adjusting to less time in the pre-clinical curriculum, this could modify the findings at these schools. Though the educator survey sampled from a group of clinical skills and pediatric clerkship educators across the United States and Canada, response rate varied and there were not mechanisms to compare responses by region or within institutions.
Our study confirms that students feel less prepared for their clinical experiences in pediatrics, in particular, pediatric physical exam skills, compared to their other clerkship specialties. It also demonstrates that clinical educators in clinical skills courses and pediatric clerkships both believe students should have some level of competence in pediatric physical exam skills prior to starting their pediatric clerkship. Medical schools should provide learning opportunities in a variety of healthcare settings that enable students to achieve specified pre-clerkship objectives. As medical schools move towards curriculum reform including optimization of earlier clinical education, we would suggest that pediatric content and skills development be included in the “pre-clinical” curriculum.
Additional file 1. Pediatric and clinical skills educator physical exam survey.
|
Effect of interactive, multimedia-based home-initiated education on preoperative anxiety inchildren and their parents: a single-center randomized controlled trial | 369582af-9802-4f6c-8a44-8632f207b8b3 | 10045252 | Patient Education as Topic[mh] | Preoperative anxiety in pediatric patients usually begins at admission when they know that they would undergo a surgical operation . Up to 50–60% of children undergoing surgical procedures experience significant preoperative anxiety . An unpleasant surgical or anesthesia experience, younger age, short preoperative preparation time, and parental anxiety are all risk factors for preoperative anxiety in children . Negative behaviors including separation anxiety, nightmares, aggression toward authorities, nocturnal enuresis, and eating disorders, can all increase preoperative anxiety in children, which also increases the postoperative analgesic requirements, prolongs the postoperative recovery process, induces emotional trauma in children and their parents, or affects the long-term cognitive and emotional development of children [ – ]. Pharmacological methods can often be used to address preoperative anxiety. Midazolam as a premedication is considered to be a reliable strategy for reducing preoperative anxiety . However, it can be a stress source itself, and therefore strict compliance and administration timing are needed, especially in children, because it may cause high-level impulsiveness and delayed emergence from anesthesia . Non-pharmacological interventions are widely supported for use in reducing preoperative anxiety in children due to their advantage of improving children’s cooperation without causing adverse effects . Studies have reported limitations of some strategies in reducing perioperative anxiety in children . For instance, transport in a ride-on toy car can relieve preoperative anxiety, but this measure is suitable and effective only for preschool children aged 2–5 years old . Some researchers also suggest the use of clown doctors, video games, and other distraction tools to temporarily relieve anxiety in children, but whether these methods can help relieve the anxiety of their parents remains unknown . Another study reported that parental presence at induction of anesthesia (PPIA) was ineffective in reducing the anxiety level of children ; rather, it increased the heart rate and skin conductance level of the parents . Given the large number of surgeries performed in children, it is imperative to optimize the surgical outcome by reducing perioperative anxiety in both children and their parents. One of the interventional categories effective for reducing anxiety in children is by providing preoperative information to them in a manner appropriate to their developmental stage. The present study aimed to explore whether the use of comic booklets, videos, and coloring books for preoperative education could reduce anxiety in young pediatric patients.
This randomized controlled clinical trial was carried out from November 2020 to July 2021 upon approval from the Institutional Review Board of Shanghai Children’s Medical Center on 17/06/2020 (SCMCIRB-Y2020095), and was registered at the Chinese Trial Registry (ChiCTR2000039622) on 03/11/2020. Informed consent was obtained from the parents or legal guardians ofall participants. The trial was carried out following the Declaration of Helsinki, and the authors guaranteed the accuracy and completeness of the data and analysis of this paper.
Children who underwent strabismus surgery at Shanghai Children’s Medical Center were recruited in this study. The inclusion criteria were children of either sex aged 4–9 years old with the American Society of Anesthesiologists (ASA) Physical Status I-II. The exclusion criteria were children with a previous history of surgery, developmental delay and verbal communication problems, hearing impairment or a history of mental illness, and who refused to participate in this study. The termination criteria were children or guardians who failed to cooperate with medical measures or withdrew from the study due to various unexpected events such as family changes or missing pets.
The study used a block randomization method with a block length of four. Eligible children (and their parents) were equally randomized to a control group and an intervention group by using a random assignment table (with randomization information including random seed, length, and the number of blocks) generated by an independent statistical professional based on the block randomization methods. Patient assessment and data collection were completed by a trained full-time research nurse. The investigators who implemented the intervention based on randomized information and performed all the assessments and the statistical analysts were blinded to the grouping information.
After making the appointment for strabismus surgery in the ophthalmology clinic, all participating children and their parents were given preoperative education about hospital admission, preoperative preparation, fasting, anesthesia procedures and pain management, strabismus surgery, and postoperative recovery procedures.
The children and their parents in the control group received a question-and-answer (Q&A) introduction to preoperative education from the researchers lasting for approximately 15–30 min, and no other preoperative education was given during the period before surgery.e.
The researchers prepared the preoperative educational information in the form of a comic booklet (S-Fig. 1), a video (S-Video), and an interactive coloring book (S-Fig. 2). The comic booklet was entitled “I’m not afraid of surgery and anesthesia”, which included admission registration, preoperative preparation, fasting, anesthesia procedures, pain management, and post-anesthesia recovery procedures. The children were allowed to choose the intervention method according to their interests, and the researchers made the content of the comic booklet into a ten-minute animated video entitled “I’m not afraid of surgery and anesthesia”. The interactive coloring game book covered admission, preoperative examination, strabismus surgery, and recovery procedures. The first intervention stage was implemented in the ophthalmology clinic. The researchers distributed the comic booklets to the children and their parents, guided them to read the comic booklets, and then showed them the video. After that, they were given interactive coloring books to read while researchers gave them explanations about the surgical and recovery procedures, and this intervention lasted for aproximately 15–30 min. The second intervention stage was implemented before hospital admission, during which the children and their parents were required to read the comic booklets or watch the video at home at least twice. The intervention was considered a success when the children and their parents read the comic booklets and/or watched the video at least twice and the children completed the interactive coloring book games; otherwise, the intervention was considered a failure and the cases involved were excluded from the study.
The anxiety level of the children was scored by using the modified Yale Preoperative Anxiety Scale-Short-Form (mYPAS-SF) by the blinded research nurse at four different time points: (1) in the Ophthalmology outpatient clinic before intervention as the baseline (T0); (2) in the preoperative waiting area (T1); (3) at the time of separating from their parents and moving to the operation room (T2); and (4) at the time of anesthesia induction (T3) (Fig. ). mYPAS-SF is a simplified version of mYPAS, originally developed by Kain et al. . The scale contains 18 items in four dimensions (activity, emotional expressivity, state of arousal, and vocalization), and the score range is 22.92–100. According to the children’s behavior at T3, the Induction Compliance Checklist (ICC) was used to evaluate compliance by the researcher. Eleven items were rated as poor (ICC > 4), moderate (ICC = 1–4), or perfect (ICC = 0) . Parental anxiety at T0 and T2 was measured by the SAS , which includes 20 questions (the higher the score, the higher the anxiety level), and the visual analog scale (VAS), which ranges from 1 (no anxiety) to 10 (extreme anxiety) . The primary outcome variablewas the children’s anxiety level, scored by the mYPAS-SF at T3. The secondary outcome variables included children’s anxiety at other time points, children’s induction compliance, and parental anxiety levels. Other basic information and intervention completion data were collected at T1 from children and parents through questionnaires.
According to the results of several existing studies in the literature and a preliminary study in our center , the mean mYPAS-SF score at T3 was approximately 60 ± 18 in the control group. With a two-sided significance level of 5% and an efficacy of 80%, 36 participants per group were required to detect a 20% decrease between the intervention and control groups (12 on the rating scale). The estimated rate of missed visits was 10%, and 84 participants were finally recruited. Statistical analysis Data were analyzed with the ITT population. Normality was analyzed by the Shapiro-Wilk test. Data of normal distribution are reported as the mean ± standard deviation (SD) and compared with the Student’s t-test. Non-normally distributed data or ordinal data are presented as the median (interquartile range) and were compared with the Mann–Whitney U test. Categorical data are presented as numbers (percentages) and compared with the χ2 test or Fisher’s exact test. Changes in mYPAS-SF over time were analyzed using a mixed-effect model with repeated measurement (MMRM) analysis using m-YPAS scores at all follow-up time points (T1, T2, and T3) as the dependent variable, treatment as the main factor, m-YPAS scores at T0 as a covariate, and a random intercept to model within-subject correlation. A p value < 0.05 was considered statistically significant. R for Windows version 4.1.2 was used for statistical analysis.
Data were analyzed with the ITT population. Normality was analyzed by the Shapiro-Wilk test. Data of normal distribution are reported as the mean ± standard deviation (SD) and compared with the Student’s t-test. Non-normally distributed data or ordinal data are presented as the median (interquartile range) and were compared with the Mann–Whitney U test. Categorical data are presented as numbers (percentages) and compared with the χ2 test or Fisher’s exact test. Changes in mYPAS-SF over time were analyzed using a mixed-effect model with repeated measurement (MMRM) analysis using m-YPAS scores at all follow-up time points (T1, T2, and T3) as the dependent variable, treatment as the main factor, m-YPAS scores at T0 as a covariate, and a random intercept to model within-subject correlation. A p value < 0.05 was considered statistically significant. R for Windows version 4.1.2 was used for statistical analysis.
Demographic and clinical characteristics Between November 2020 and June 2021, 528 children and their parents were screened for this study. Of them, 443 did not meet the inclusion criteria, one declined to participate, and 84 were enrolled, with 42 in each group. Of the 84 included children, the operation was postponed beyond the study period in four children due to personal reasons, intervention failed in the other two children, and the remaining 78 (40 in the control group and 38 in the intervention group) were available for analysis. The flow chart of the study in terms of patient enrollment, allocation, follow-up observation, and analysis is shown in Fig. . There was no significant difference in the baseline sociodemographic and clinical variables between the two groups (Table ).
Between November 2020 and June 2021, 528 children and their parents were screened for this study. Of them, 443 did not meet the inclusion criteria, one declined to participate, and 84 were enrolled, with 42 in each group. Of the 84 included children, the operation was postponed beyond the study period in four children due to personal reasons, intervention failed in the other two children, and the remaining 78 (40 in the control group and 38 in the intervention group) were available for analysis. The flow chart of the study in terms of patient enrollment, allocation, follow-up observation, and analysis is shown in Fig. . There was no significant difference in the baseline sociodemographic and clinical variables between the two groups (Table ).
There was no significant difference in the baseline m-YPAS-SF scores at T0 between the control group and the intervention group (41.68 [31.26, 45.85] vs. 41.68 [37.51, 45.85], p = 0.859). The mean m-YPAS-SF score in the intervention group was significantly lower than that in the control group at T1, T2, and T3 (all p < 0.001). By using MMRM after adjusting m-YPAS scores at T0 as a covariate, the intervention effect in terms of the mYPAS-SF score was also significant over time (p < 0.001). The percentage of children with perfect induction compliance (ICC = 0) in the intervention group was higher than that in the control group [18.4% vs. 7.5%], and poor induction compliance (ICC>4) was lower (2.6% vs. 17.5%, p = 0.048) (Table ). As shown in Fig. , the anxiety level in the intervention group was significantly lower than that in the control group at T1, T2, and T3 (all P < 0.001). The anxiety level was the lowest at T1 in both groups, and the anxiety level reached the highest at T3 in the control group.
There was no significant difference in the mean VAS score and SAS score at T0 between the two groups. The mean VAS score at T2 in the intervention group was significantly lower than that in the control group (4.0 and 2.3 vs. 5.2 and 2.1, 95% CI: -1.2 (-2.20, -0.20), P = 0.021). The difference in the mean SAS score at T2 was not significant between the two groups (36.51 and 9.65 vs. 37.86 and 11.04, 95% CI: -1.34 (-6.02, 3.32), P = 0.568) (Table ).
The evaluation of the intervention completion surveys and the information received from the children through questionnaires are shown in Table . Of the 40 children in the intervention group, 38 (95%) completed the intervention, and the remaining two (5%) failed the intervention, including one who neither finished the coloring book, watched the video, nor read the comic book, and the other one who did not complete the intervention because he did not like the coloring book.
In this study, we investigated an approach that integrated several previous relatively independent approaches to provide children with an active participatory [ – ], multimedia-based home-initiated educational intervention prior to admission. Since this experiment incorporates parents and includes three aspects of children’s audio-visual stimulation, active painting, and reading, it is more effective than other experiments. We found that the anxiety level in children in the intervention group was significantly lower than that in children in the control group at all designated time points (during preoperative waiting, at the time of separation from their parents, and at the time of induction of anesthesia).During anesthesia induction, children in the control group had higher levels of anxiety and were less cooperative. It was also found that the parents of children in the intervention group had a lower VAS anxiety score when they were separated from their children. These results can serve as a complement to the literature concerning non-pharmacological approaches for the prevention and treatment of preoperative anxiety in pediatric patients, including education, behavioral techniques, PPIA, and complementary and alternative medicine (CAM) techniques . The strengths of this study include the randomized controlled design, utility, and use of existing and validated behavioral-based scales to reduce observer bias. A similar study showed that reading educational comic leaflets approximately one week before surgery could effectively alleviate anxiety levels in children aged 6–17 years old, but they did not address the effect on parental anxiety. It is commonly believed that children’s preoperative anxiety positively correlates with parental anxiety . We did not perform a related analysis, but we found that parental anxiety scores measured by the SAS and VAS were inconsistent. The possible reason is that SAS is a clinical tool to analyze the subjective anxiety of the client, which evaluates the client’s emotional state over a period, while VAS measures the immediate emotional feeling. Li et al. observed how parental anxiety influenced children to feel more frightened and be less cooperative. Thus, efforts should be made to decrease parents’ preoperative stress behaviorally or by using other interventions that could reduce their children’s anxiety . Another study performed by Hilly et al. reported that one-hour study two weeks before theoperation could reduce preoperative anxiety or even prevent the occurrence of postoperative adverse behavioral changes in children. However, this strategy often requires extra time and additional labor for support, while our approach is simpler and more effective. Our study found that preoperative education could effectively reduce preoperative anxiety in children. Kim et al. also showed that the use of handheld tablet devices with interactive functions may be the most effective intervention strategy for reducing preoperative anxiety in children, which is consistent with our findings. More importantly, our broad inclusion criteria and uniformity of case selection may add credibility to the study and therefore be more applicable to pediatric patients with ophthalmological diseases. Children older than four years old usually have already had a more developed sense of self and potential harm . They are also better able to cooperate. In children, the perception of anxiety depends on the developmental stage and cognitive potential, and different responses can be observed among those facing the same stressor agent . Studies have shown that most children who undergo surgical procedures wish to know more detailed information about what they will encounter and what will happen in the operating room. They usually prefer to have comprehensive information concerning their surgery, including information about pain, anesthesia, perioperative procedures, and potential complications . Additionally, the more anxious the children feel, the more eagerly they want to know information about pain . Children’s processing of preparatory information may be affected by multiple factors, including previous experience, developmental level, and comprehension . The timingof providing preparatory information may also influence how much can be retained, although the optimum timing is not known. Some earlier work suggests that information should be given at least five days in advance for children aged six years or older and no more than a week in advance for children below six years of age . From this point of view, proactive anxiolytic measures should be taken early even before hospital admission , so the intervention in our study was initiated in the outpatient clinic before admission to the hospital. The results of our questionnaire investigation demonstrated that the intervention failed in a four-year-old boy who could fully understand the comic booklet because his parent interfered too much while watching the video, so he refused to engage with the coloring book. The other failure occurred in a seven-year-old boy because he did not like the coloring game, although he did well in reading the comic booklet and watching the animated video, and behaved cooperatively throughout the operation period. This interactive multidimensional early intervention provided additional help to pediatric patients, and effectively reduced their preoperative anxiety, which is consistent with our previous expectations. Although there have been numerous clinical studies on how to reduce preoperative anxiety in children and a series of measures have also been proposed , specific factors affecting preoperative anxiety inchildren of different age groups are still not fully understood. For this reason, there are limited targeted interventions with controversial outcomes. The Home-Initiated-Programme-to-Prepare-for-Operation (HIPPO) intervention did not achieve any significant effect in diminishing children’s anxiety before an operation .Eijler et al. also reported that the provision of virtual reality exposure as a form of distraction therapy for children undergoing day surgery had no significant beneficial effect on anxiety and pain . Fincher et al. instituted comprehensive play interventions but found no alleviation in perioperative anxiety in children . The discrepancy may be due to the multifactorial etiology of preoperative anxiety in pediatric patients. Therefore, it is increasingly recognized that addressing preoperative anxiety should be a multimodal effort. It was also found in our study that the anxiety level was the lowest at T1 in the preoperative waiting area in children of both groups, which was lower than the level at T0 in the same group. For this reason, we transformed the preoperative waiting area into a children’s play area where parents can accompany their children to play with toys, read picture books, watch cartoons, and play games with other children awaiting surgery. As a result, both children and their parents feel more relaxed and less anxious. In contrast, children in the control group all became vigilant when they were about to enter the operating room and separate from their parents and their anxiety level peaked at the time of anesthesia induction. This phenomenon shows that simple distraction does not solve the source of anxiety in children, while interactive and multidimensional preoperative education is the key to relieving children’s anxiety.
There are some limitations in this study. First, our multimedia products including comic booklets, videos, and coloring books were created based on the environment of our hospital and are personalized for children who come to the hospital for treatment and surgery, and thus may not be suitable for use in other hospitals. For instance, decorations and culture of our hospital are different from those of other hospitals. However, at any rate, our experience and practice may serve as a useful reference. In addition, we excluded children younger than four years old because they may not be able to read and write, so more studies are required to explore how to relieve anxiety in children younger than four-year old. Finally, we did not follow up with the children when they were returned to the ward, so it is unclear whether these measures can also help reduce postoperative maladaptive behaviors over time.
The results of this practical randomized controlled trial demonstrated that participatory, multimedia-based home-initiated preoperative intervention could effectively relieve preoperative anxiety in children and reduce parental anxiety. It is an effective and noninvasive way to treat anxiety in pediatric patients.
Below is the link to the electronic supplementary material. Supplementary Material 1
|
Canada’s First Joint Oncology-Allergy Clinic: Successful Desensitization to Trastuzumab Following Severe Anaphylactic Reaction in Which Epinephrine Was Inappropriately Withheld | b9b089e1-8991-44ee-9919-9748cbeb66f2 | 10046925 | Internal Medicine[mh] | Trastuzumab is a humanized recombinant monoclonal antibody agent targeting the proto-oncogene human epidermal growth factor receptor 2 (HER2) . HER2 is overexpressed in approximately 20% of breast cancers . Neoadjuvant trastuzumab has been shown to increase pathologic complete response rates when combined with chemotherapy, can cause regression of disease, and improve overall survival in patients with metastatic HER2-positive breast cancer . While up to 40% of patients may experience trastuzumab-associated infusion reactions, severe hypersensitivity reactions occur in only 1–3% of patients . Various mechanisms can cause acute reactions to monoclonal antibody therapies, such as trastuzumab, including IgE-mediated anaphylactic and anaphylactoid reactions, serum sickness, cytokine release syndrome, and tumor lysis syndrome . Symptoms can range from local skin reactions to potentially fatal anaphylaxis . Anaphylaxis is an acute multi-system allergic reaction—it is clinically diagnosed using one of three accepted diagnostic criteria . Other tests can be used to support diagnosis in patients with atypical symptoms. Current guidelines recommend measuring serum tryptase from one-half to two hours after the onset of anaphylaxis, in addition to measurement of baseline tryptase 24 h after resolution of all anaphylaxis symptoms . During anaphylaxis, serum tryptase concentrations increase as a result of mast cell degranulation, and while measuring levels during a clinical emergency will not help diagnose anaphylaxis, if elevated above baseline, it can be helpful in confirming the diagnosis of anaphylaxis retrospectively . A serum tryptase 1.2 times higher than baseline tryptase +2 μg/L supports a diagnosis of anaphylaxis . The baseline tryptase also permits the exclusion of other conditions associated with an elevated tryptase that can contribute to severe anaphylactic reactions, such as Mastocytosis and Hereditary Alpha Tryptasemia . Skin-prick or intradermal testing post-reaction can also be conducted to evaluate for IgE-mediated disease when used in correlation with clinical history . Epinephrine is the recommended treatment for anaphylaxis and should be administered immediately to patients in whom anaphylaxis is suspected . This recommendation stands even if the diagnosis is uncertain, as there are no absolute contraindications to epinephrine administration in anaphylaxis . Supportive therapy may also be indicated, including volume replacement with intravenous fluids, oxygen therapy, bronchodilators (inhaled β2-agonists), and antihistamines for the treatment of cutaneous symptoms . While corticosteroids are often given following anaphylaxis, there is little evidence of immediate benefit or prevention of delayed biphasic reaction . However, supportive therapy should never delay the administration of epinephrine and cannot replace epinephrine as the definitive treatment for anaphylaxis . Delayed epinephrine remains the greatest risk factor for anaphylaxis mortality and should always be given with a low threshold to treat . Trastuzumab has potentially tremendous benefits, and when hypersensitivity reactions occur, rechallenge with desensitization protocols has become more common and has been included in small numbers in several case series of desensitization to oncology treatments . There are currently no formal guidelines or recommendations concerning trastuzumab rechallenge, but desensitization protocols have been published in the literature . A recent European Academy of Allergy and Clinical Immunology (EAACI) position paper proposed an algorithm for the evaluation of hypersensitivity reactions to chemotherapeutic agents . It recommends serum tryptase levels and immunologic skin testing for patients with immediate hypersensitivity reactions in order to assess their risk of continuing treatment . If deemed low risk, patients can undergo a challenge in which the drug is administered in a monitored environment . If hypersensitivity is suspected and the patient is deemed to have a moderate-to-high risk of reaction on further exposure, they can undergo rapid drug desensitization . During desensitization, the drug is administered in diluted amounts according to a multi-step protocol that increases the target dose over time, resulting in a temporary state of tolerance . While the mechanism is incompletely understood, this immunologic tolerization of the offending drug is induced by inhibiting mast cell activation responses and thus preventing anaphylaxis . Oncology presents a unique situation in which repeat drug exposure after serious adverse reactions is often warranted due to the mortality risk of untreated cancer—allergists can assist with both symptom assessment and risk mitigation.
We report successful desensitization in a 43-year-old female with locally advanced HER2-positive breast cancer following a severe anaphylactic reaction to trastuzumab infusion for which epinephrine was not administered. At initial assessment, the patient had biopsy-confirmed HER2-positive, ER/PR-negative, and right breast invasive mammary carcinoma measuring 1.8 cm in diameter. Imaging demonstrated multiple axillary lymph nodes (largest 3.0 × 2.1 cm) and a right supraclavicular node (1 cm). No metastases were identified. Her oncologist assessed this as stage III and recommended neoadjuvant Docetaxel, Carboplatin, and trastuzumab every 3 weeks for a total of 6 cycles, followed by mastectomy. Chemotherapy was administered at a local clinic in a northern region of the province of Manitoba. The patient was otherwise healthy, took no other medications, with no history of allergies or allergic disease. Cycle 1 of neoadjuvant therapy was administered without issue. However, during Cycle 2, the patient experienced an adverse reaction approximately 2 min after trastuzumab infusion was initiated. She became flushed and reported evolving symptoms that quickly culminated in hypotension (BP 80/49), tachycardia (HR 147), dyspnea (O 2 sat 91%), and nausea. She had not yet received Docetaxel or Cyclophosphamide. The trastuzumab infusion was discontinued, and her IV was flushed with normal saline (NS). O 2 was applied at 6 L/min via a face mask. The patient received 50 mg Diphenhydramine IV. SpO 2 continued to decrease, at which point O 2 was increased to 10 L/min. Hydrocortisone, 100 mg IV, was then administered. The patient became unresponsive and began vomiting (total 500 mL emesis). Vital signs were persistently unstable, and multiple NS boluses were administered via pressure bag to a total of 5 L. She did not receive epinephrine. SpO 2 stabilized on 10 L/min face mask and remained stable on a decrease to 5 L/min. The patient was observed for 2 additional hours. Each decrease in NS IV rate resulted in hypotension with BP of 80/52 and tachycardia of 120–150 bpm. The patient was unconscious for 1.5 h and was weaned off O 2 , at which point she was awake and walked to the bathroom with assistance from the nursing staff. On return to bed, her vital signs quickly deteriorated again. The local emergency department (ED) was contacted, and the patient was transferred for further management. After a presentation to the ED, the patient was given norepinephrine IV. This decision was made based on the large volume of IV fluid she had received prior to arrival. She was weaned off the norepinephrine and was stable and discharged home later that evening. A tryptase level was not drawn due to patient refusal. Manitoba has recently established a multidisciplinary Medical Oncology and Allergy Clinic aimed at allowing oncology patients to receive treatments despite adverse reactions. Re-administration with desensitization was pursued, as the anti-HER2 component to her therapy was felt to carry significant potential benefit. Cycle 2 was administered two weeks later with a desensitization protocol. Desensitization involves administering the drug at increasing concentrations and infusion rates with set time intervals to allow those with an anaphylactic allergy to receive the medication safely . The patient tolerated the desensitization protocol for Cycle 2 and has since completed up to 5 cycles using desensitization without issue.
In this case, epinephrine was not administered. The patient developed a severe anaphylactic reaction and remained persistently hypotensive and hypoxic to the point of unresponsiveness despite the repeated use of supportive measures. The use of Diphenhydramine—a first-generation antihistamine—was likely detrimental, as it may have reduced consciousness and worsened hypotension; antihistamines do not save lives as they only treat mild pruritus and hives. Epinephrine was required to treat the bronchoconstriction and vasodilation secondary to anaphylactic shock, and the patient continued to deteriorate when she did not receive it. This patient’s reaction to trastuzumab was anaphylaxis. While it may not have been absolutely clear at the time, the clinical uncertainty should not have prevented epinephrine from being administered. She was likely sensitized on first exposure when she tolerated the initial dose of trastuzumab and then reacted within minutes of starting her Cycle 2 infusion. Anaphylaxis is a clinical diagnosis, and she met the first diagnostic criteria for anaphylaxis based on her acute development of generalized flush, hypotension, bronchospasm and hypoxia, and gastrointestinal symptoms. Furthermore, she failed to improve clinically despite receiving antihistamines, corticosteroids, and over 5 L of crystalloid intravenous fluids. This suggests vasodilation and distributive shock secondary to anaphylaxis mediators. After the reaction, the patient was referred to the Medical Oncology and Allergy Clinic for evaluation. The patient declined intradermal skin testing to confirm Trastuzumab hypersensitivity, as she understood that regardless of positive or negative results, she would proceed with desensitization, given the severity of her reaction. Immunologic skin-prick or intradermal testing is always recommended but is sometimes unavailable or does not change management. In the case of some chemotherapy agents, it may not be feasible due to cutaneous toxicity . The tests are also poorly validated as there is limited information about their sensitivity and specificity . There are currently no standardized skin-prick or intradermal testing doses for many chemotherapy agents and monoclonal antibody therapies ; some concentrations have been recommended, but the evidence is limited , including for trastuzumab . Therefore, even when available, immunologic skin tests are not always reliable in confirming hypersensitivity after a reaction occurs. In this particular case of severe anaphylaxis, desensitization is the appropriate intervention for further oncologic therapy regardless of intradermal test results. The fact that she also successfully tolerated rechallenge with desensitization and was able to continue treatment reinforces anaphylaxis as the most likely etiology. Serum tryptase can also be retrospectively used to confirm anaphylaxis in the event of clinical uncertainty and is recommended by current guidelines . While the patient did meet the clinical criteria for anaphylaxis , a serum tryptase was unfortunately not drawn due to the patient’s refusal, which is a disadvantage. The lack of baseline tryptase also means that confounding conditions, such as Mastocytosis and Hereditary Alpha Tryptasemia, could not be evaluated, especially given the severity of the patient’s reaction. However, genetic testing for Hereditary Alpha Tryptasemia is not currently available in Manitoba, locally, or as an external test. Furthermore, even if a Tryptase were drawn, a normal level would not have ruled out Mastocytosis, and the patient ultimately would have required a bone marrow biopsy to diagnose the condition . The authors were reassured by the patient’s absent history of prior anaphylaxis or other symptoms indicative of mast cell disorder and felt that the condition was unlikely in this case. Delay in epinephrine administration is a common issue worldwide, particularly with drug-induced anaphylaxis. There is often failed or delayed recognition in hospital and emergency room settings, and even when recognized, epinephrine is under-used, and patients are inappropriately treated with only fluids and antihistamines . One Canadian study postulated that healthcare provider concerns about epinephrine side effects were sometimes a barrier to its administration , but there are no absolute contraindications to the administration of epinephrine when treating anaphylaxis . There are limited studies informing trastuzumab-specific guidelines around treatment protocols when adverse reactions occur. However, the current anaphylaxis guidelines for the diagnosis and treatment of suspected anaphylactic reactions apply to all instances of anaphylaxis . This includes reactions to monoclonal antibody therapies, such as trastuzumab, chemotherapy agents, and other drugs . The treatment guidelines should therefore be implemented in oncology settings, such as the one described in this case, and epinephrine should always be the recommended treatment. Multidisciplinary involvement in this and several similar cases has prompted a change in the CancerCare Manitoba treatment room protocols to ensure prompt administration of epinephrine intramuscularly at the onset of possible anaphylaxis. When hypersensitivity reactions occur in response to chemotherapy agents or biologics, it can become a serious impediment to patients receiving first-line therapies. Drug desensitization is a means by which patients can tolerize agents that have previously caused hypersensitivity reactions . The drug is typically administered using a series of gradually increasing dose increments that eventually total to equal the target drug dose . Rechallenge with desensitization should be considered before therapy is discontinued or switched to potentially sub-optimal alternatives. The EAACI gave Task Force status to the interdisciplinary field of AllergoOncology in 2014, which aims to encourage the exchange of expertise in order to further understanding of both topics . The field focuses on the relationship between allergic responses and cancer, immunomodulation, including IgE-mediated responses, and aims to provide new insights into treatments, such as cancer immunotherapies . While allergic reaction to anti-cancer agents is not a new problem, the management of allergy and allergic disease in daily clinical oncology is an emerging area that falls under this interdisciplinary umbrella . Reactions to chemotherapy agents and monoclonal antibody therapies are increasingly common in the last several decades and have become more clinically significant as the number and efficacy of therapies have increased . While there is an abundant collaboration between allergists and oncologists in many European jurisdictions as a result of the EAACI’s efforts in the expanding field of AllergoOncology , there are only a handful of examples in North America. Only two such clinics have been established in the United States, out of the Brigham and Women’s Hospital in Boston and the Albert Einstein College of Medicine in New York. The Medical Oncology and Allergy Clinic in Winnipeg, Manitoba, is the first formal collaboration in Canada with one of its goals to provide recommendations regarding acute management of hypersensitivity reactions and prevention of their recurrence (e.g., with the administration via desensitization). When patients have adverse reactions to their cancer therapies, they are promptly referred to the clinic for urgent assessment, immunologic skin testing, and consideration of rechallenge or desensitization. The clinic deals with all allergy-related issues experienced by current oncology patients. This includes not only infusion-related or anaphylactic reactions and subsequent skin testing and desensitization but also suspected allergic reactions to oral anti-cancer drugs and contrast dye. Patients with penicillin or other antibiotic allergies are also seen, especially those being considered for allogenic stem cell transplant. Prior to the clinic’s establishment, local allergists and oncologists noticed there was a high frequency of often time-sensitive consultations for oncology patients. This led to the involvement of the hospital’s Chief Medical Officer, department heads of medical oncology and allergy, and directors of the Systemic Therapies program and Provincial Oncology Drug Program. The data supporting allergist involvement in medical oncology was reviewed, along with the regular consults, leading to the agreement that a dedicated clinic would then provide the time, place, and staff to prioritize the management of oncology patients’ allergy-related problems. The clinic was established in September 2022 and runs one half-day per week, and approximately five patients are seen per clinic. The timing of referral is generally quick, and patients are seen within 5 business days. In the case of this patient, she was able to continue Cycle 2 of trastuzumab after a mere 2-week delay. She went on to complete the full course without further adverse reactions and an apparent positive clinical response to her treatment regimen. She may have had worse outcomes if she had not undergone desensitization and if trastuzumab had been discontinued.
Treatment for anaphylaxis is epinephrine: the largest risk factor for anaphylaxis-related death is delayed epinephrine. Antihistamines are not life-saving, and steroids have limited utility in anaphylaxis treatment. Authors are currently changing Manitoba-wide treatment protocols to ensure early recognition and appropriate treatment of anaphylaxis during treatment administration. The authors eagerly encourage oncology programs across the country to reassess their protocols for the treatment of adverse reactions. Hypersensitivity reaction to anti-cancer drugs should prompt urgent allergy referral for assessment and consideration of risk-mitigating procedures (e.g., desensitization) prior to discarding life-saving treatments. We propose this multidisciplinary clinic model as a treatment framework moving forward, with the goal of continuing first-line therapies in cancer patients who develop drug-hypersensitivity. This case highlights the unmet need for a multidisciplinary approach to the management of oncology patients who experience hypersensitivity reactions.
|
B7H4 Expression Is More Frequent in MSS Status Colorectal Cancer and Is Negatively Associated with Tumour Infiltrating Lymphocytes | f5b247a6-fbae-48bb-a35b-f1607df35c2f | 10046962 | Anatomy[mh] | The B7H4 (aliases VTCN1, B7S1, B7x) protein belongs to the B7 family . The B7 family ligands (present on APCs-Antigen presenting cells) bind to its counter receptor from the CD28 family (present on the T cell), which plays a central role in fine-tuning the antigen-specific immune response. This immune response belongs to cell-mediated adaptive immunity, which is particularly important in the proper antitumour response. B7H4 is a co-inhibitory ligand of the B7 family . The general function of B7H4 is to downregulate immune reactions by inhibiting T-cell activation, proliferation, and cytokine production. The expression of B7H4 is limited in normal tissues; however, its overexpression has been confirmed in a wide selection of solid malignancies, including lung, liver, kidney, ovary, stomach, skin, pancreas, colorectal, and breast cancer . The members of the B7 family include, among others, PD-L1, PD-L2, PD-1, and CTLA-4. These molecules have a highly important place in oncological clinical practice as targets for immunotherapy. An immune checkpoint blockade against them is used in a broad range of human malignancies . However, they are not universal. In colorectal cancer, their usefulness is limited to patients with damage mismatch repair genes (dMMR), for whom microsatellite instability (MSI) is its marker. The group of patients with MSI status represents only approximately 15% of all patients suffering from colorectal cancer . The selective upregulation of PD-1, CTLA-4, LAG-3, Tim-3, and killer immunoglobulin-like receptors is often present in MSI-H tumours. The presence of these co-inhibitory receptors may explain the phenomenon. MSI-H tumours are not eliminated naturally, despite high immune activation in this type of cancer. Furthermore, tumour-infiltrating cells (TILs) express high levels of PD-1 in MSI colorectal cancer, which is absent in microsatellite stable tumours (MSS); this explains why a checkpoint blockade is effective in MSI status tumours . For dMMR/MSI-H colorectal cancer, there are approved anti-programmed cell death protein 1 (PD-1) antibodies called pembrolizumab and nivolumab for treating patients who have previously received chemotherapy. Additionally, the combination of nivolumab with ipilimumab (a CTLA-4 inhibitor) to treat MSI-H colorectal cancers that progressed prior to chemotherapy is available . The tumour microenvironment (TME) consists of stromal, extracellular components, and immune cells. The immune part creates only immunological tumour microenvironment (iTME). The iTME can vary across the cancer subtypes and the disease stage . The iTME is a dynamic system in which the combination of cell types, location, and functional orientation leads to the creation of an effective anti-tumour barrier and further tumour rejection or a tumour-promoting environment. Cytokines are some of the most important components within the iTME and are involved in a conflict between tumour cells and tumour-infiltrating immune cells . However, there are limited data on how the local cytokinome landscape influences the expression of B7H4 in colorectal cancer . This study aimed to evaluate the expression of the B7H4 to MSI/MSS status and other clinicopathological features of CRC. Furthermore, another goal was to explain the immunological context in which the expression of B7H4 occurs, through the 48-cytokine screening panel of cancer tissues homogenates and immunogenic features, immune composition, and functional annotations analysis of online available datasets.
2.1. Characteristics of the Patient Group The samples from 167 patients obtained during surgery due to CRC were used in the study. Patients were treated in the 1st Specialist Hospital in Bytom, Poland (approval of the Research Ethics Committee No. KNW/0022/KB1/42/III/14/16/18, 14 July 2020). The collected specimens included colorectal tumour tissues and adjacent normal tissue. Patients were enrolled based on the inclusion and exclusion criteria described in our previous paper . Research group characteristics are presented in . 2.2. Evaluation of the B7H4 Expression by ELISA Fragments of the tumour tissue and surgical tissue margin were weighted and homogenized according to the standard homogenization protocol already described in our previous paper . To assess the levels of the B7H4 protein, an enzyme-linked immunosorbent assay (ELISA) was used, following the manufacturer’s instructions. B7H4 levels were evaluated by a human B7H4 ELISA kit (Cloud Clone, Wuhan, China) with a sensitivity of 56 pg/mL. The absorbance of the samples was determined using a Universal Microplate Spectrophotometer (μQUANT, Biotek Inc., Winooski, VT, USA). The measurement was conducted at a wavelength of 450 nm. The obtained results were recalculated to the corresponding total protein level and presented as pg/mg of protein. 2.3. Evaluation of the B7H4 Expression by IHC For B7H4 expression, the B7H4 immunostaining was performed in 76 cases in which the MSI/MSS statuses were also established. Tissue samples were obtained from formalin-fixed paraffin-embedded tissue blocks with primary CRC and tumour-free margin samples. Then, the samples were deparaffinized and rehydrated. In the next step, antigen retrieval was performed by incubating slices in EnVision Flex Target Retrieval Solution High pH (Dako, Carpinteria, CA, USA) for 20 min at 95 °C. Prepared samples were incubated with Peroxidase-Blocked Reagent (Dako) and then incubated with antibody: B7-H4 Polyclonal Antibody, Invitrogen, incubation time: 40 min.; dilution: 1:1500. After this process, they were put in EnVision FLEX HRP (Dako). Then, antigen–antibody complexes were stained using 3,3′-diaminobenzidine. Finally, tissue sections were counterstained with hematoxylin, dehydrated, and covered with coverslips for further analysis. 2.4. Assessment of the MSI/MSS Status For MSI/MSS status evaluation in 101 cases, the IHC staining for MSH2, MSH6, PMS2, and MLH1 was performed on 4 µm thick sections of a representative formalin-fixed, paraffin-embedded (FFPE) tumour tissue block on a Dako Autostainer Link 48. Samples underwent deparaffinization and rehydratation. In the next step, antigen retrieval was performed by incubating slices in EnVision Flex Target Retrieval Solution High pH (Dako, Carpinteria, CA, USA) for 20 min at 95 °C. Prepared samples were incubated with Peroxidase-Blocked Reagent (Dako) and then incubated with one of the following antibodies: Mouse Monoclonal antibody MSH2 (G219-1129), Cell Marque, incubation time: 30 min.; dilution: 1:400; Mouse Monoclonal antibody MSH6 (44), Cell Marque, incubation time: 45 min.; dilution: 1:100; Mouse Monoclonal antibody PMS2 (MRQ-28), Cell Marque, incubation time: 40 min.; dilution: 1:50; Mouse Monoclonal antibody MLH1 (G168-728), Cell Marque, incubation time: 40 min.; dilution: 1:100. After this process, they were put in EnVision FLEX HRP (Dako). Then, antigen–antibody complexes were stained using 3,3′-diaminobenzidine. Finally, tissue sections were counterstained with hematoxylin, dehydrated, and covered with coverslips for further analysis. Tumours were assessed as to whether nuclear staining of invasive tumour cells for MSH2, MSH6, PMS2, and MLH1 was seen in the presence of positive internal control (inflammatory and stromal cells). Tumours with nuclear staining for markers in at least 1% of invasive tumour cells were considered to have positive marker staining. The algorithm on which the interpretation of immunohistochemistry testing was based is presented in the article of Olave and Graham . The MSI status was recognized if one of the following marker layouts were present: MLH1 and PMS2 loss, PMS2 loss, MSH2, and MSH6 loss, or MSH6 loss. We assumed negative staining as the loss. 2.5. Assessment of the Tumour-Infiltrating CD8+ T Cells Briefly, 4 µm thick tissue sections were used for immunohistochemical (IHC) analysis. They were deparaffinized with xylene, rehydrated in graded alcohol, and washed in deionized water. In the next step, antigen retrieval was performed by incubating slices in EnVision Flex Target Retrieval Solution High pH (Dako, Carpinteria, CA, USA) for 20 min at 95 °C. Prepared samples were incubated with Peroxidase-Blocked Reagent (Dako) and then incubated with antibody: (CD8+/144B) Mouse Monoclonal Antibody diluent, incubation time: 40 min.; dilution: 1:100. After this process, they were put in EnVision FLEX HRP (Dako). Then, antigen–antibody complexes were stained using 3,3′-diaminobenzidine. Finally, tissue sections were counterstained with hematoxylin, dehydrated, and covered with coverslips for further analysis. 2.6. Assessment of the TILs and Budding Tumour-infiltrating lymphocytes (TILs) assessment was performed in 102 specimens. The percentage of tumour-associated lymphatic infiltration was estimated semi-quantitatively on a five-grade scale on the same H&E-stained slides by the two pathologists, according to the criteria defined by Salgado et al. in breast cancer . These include intratumoural lymphocytes with cell-to-cell contact between lymphocytes and tumour cells, and stromal TILs in tumour tissue located dispersed in the stroma within the tumour cells without direct contact, including TILs at the invasive margin. According to the recommendations, stromal TILs were scored as a percentage of the stromal area alone, excluding areas occupied by carcinoma cells. Lymphatic infiltrates outside the tumour borders were not included in the evaluation. A lymphocyte infiltration area lower than 5% was considered TILs 1, whereas 5–25%, 25–50%, and 50–75% of lymphocytes in the stroma were defined as TILs 2, TILs 3 and TILs 4, respectively. More than 75% was defined as TILs 5. Tumour budding was assessed in the same 101 specimens. Tumour buds were estimated in one FOV at a hotspot area in the invasive front under ×20 magnification. The number of buds was adjusted by the normalization factor (1.210). Budding was reported in the following manner: low budding: 0–4 buds; intermediate budding: 5–9 buds; high budding: >10 buds. The mean number of buds per FOV was also used in the statistical analysis. 2.7. Assessment of the Cytokines Screening Panel Homogenates’ supernatants were collected after they were stored at −80 °C. The concentrations of cytokines/chemokines/growth factors were measured in 77 homogenates by the Bio-Plex Pro Human cytokines screening panel 48 cytokines assay (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer’s instructions. In brief, 50 µL aliquot of the sample was diluted 1:2 with sample diluent, incubated with antibody-coupled beads and biotinylated secondary antibodies, and followed by streptavidin-phycoerythrin. Standard curves for each studied parameter were performed using respective cytokine standard solutions. The beads were analysed in the Bio-Plex Array Reader (Bio-Plex Manager 6.2 software from the Bio-Plex 200 System). The intra-assay %CV varied up to 15%, and the inter-assay %CV varied up to 25%, depending on the analysed parameter. The obtained results were then normalised to the corresponding total protein level. This method has been previously used in analysing cytokines in lymphocyte cell culture supernatants and in blood serum samples . 2.8. Exploration of Biological Characteristics of B7H4 We performed functional annotations analysis based on mRNA expression profiles in the CRC online dataset from the FieldEffectCrc Package among cohort A, consistent with 311 CRC samples . We normalized the matrix data using the DESeq2 package . Then, we divided the cohort into high versus low expressions of VTCN1(B7H4). Gene set enrichment analysis was used to elucidate the potential Hallmarks pathways from The Molecular Signatures Database (h.all.v7.5.symbols.gmt) of B7H4 in CRC in R Studio with fgsea package. The genes with significant differences in expression, including high vs. low B7H4 expression, were screened for GO enrichment analyses (|logFC| > 0.5 and p.adj. < 0.05). Then, we used TCGA-COAD data to explore the association of B7H4 expression with immunogenicity and immune landscape in colorectal cancer. Results of the immune-related scores, mutation analysis, and immune cell infiltration scores, were all obtained from the CAMOIP website ( http://220.189.241.246:13838/#shiny-tab-188 home (accessed on 15 December 2022)). 2.9. Statistical Analysis Data distribution was determined using the Shapiro–Wilk test. The log transformation of the concentrations of the studied proteins provided a better fit for the Gaussian distribution. The data are presented as mean ± SD for the variables with normal distribution, and as median with interquartile range for the variables with non-normal distribution. To compare the tumour and margin concentrations, the paired Student’s t -test (for variables with a normal distribution) and Mann–Whitney U test (for variables with non-normal distribution) were used. Independent variables were also compared using the Student’s t -test and Mann–Whitney U test. Pearson’s coefficient or Spearman coefficient were used for assessing the relationships between the examined variables (for variables with normal and non-normal distribution, respectively). Tau-Kendalls’ Tau Rank Correlation Coefficient was used to determine the association between the levels of the examined proteins, T, and N parameters. p values < 0.05 were considered significant. Hierarchical clustering and principal component analysis were performed to reduce the number of variables and clarify the influence of B7H4 expression on the cytokines profile. The factors obtained in PCA were used as new variables named Dim 1 and Dim 2 according to their proportions of the explained variance. Next, the Dim 1 and Dim 2 values were used in further analyses. p values ≤ 0.05 were considered significant. Statistical analysis was performed using STATISTICA 13 software (Statsoft) and the R Studio (Integrated Development for R. RStudio, PBC, Boston, MA, USA).
The samples from 167 patients obtained during surgery due to CRC were used in the study. Patients were treated in the 1st Specialist Hospital in Bytom, Poland (approval of the Research Ethics Committee No. KNW/0022/KB1/42/III/14/16/18, 14 July 2020). The collected specimens included colorectal tumour tissues and adjacent normal tissue. Patients were enrolled based on the inclusion and exclusion criteria described in our previous paper . Research group characteristics are presented in .
Fragments of the tumour tissue and surgical tissue margin were weighted and homogenized according to the standard homogenization protocol already described in our previous paper . To assess the levels of the B7H4 protein, an enzyme-linked immunosorbent assay (ELISA) was used, following the manufacturer’s instructions. B7H4 levels were evaluated by a human B7H4 ELISA kit (Cloud Clone, Wuhan, China) with a sensitivity of 56 pg/mL. The absorbance of the samples was determined using a Universal Microplate Spectrophotometer (μQUANT, Biotek Inc., Winooski, VT, USA). The measurement was conducted at a wavelength of 450 nm. The obtained results were recalculated to the corresponding total protein level and presented as pg/mg of protein.
For B7H4 expression, the B7H4 immunostaining was performed in 76 cases in which the MSI/MSS statuses were also established. Tissue samples were obtained from formalin-fixed paraffin-embedded tissue blocks with primary CRC and tumour-free margin samples. Then, the samples were deparaffinized and rehydrated. In the next step, antigen retrieval was performed by incubating slices in EnVision Flex Target Retrieval Solution High pH (Dako, Carpinteria, CA, USA) for 20 min at 95 °C. Prepared samples were incubated with Peroxidase-Blocked Reagent (Dako) and then incubated with antibody: B7-H4 Polyclonal Antibody, Invitrogen, incubation time: 40 min.; dilution: 1:1500. After this process, they were put in EnVision FLEX HRP (Dako). Then, antigen–antibody complexes were stained using 3,3′-diaminobenzidine. Finally, tissue sections were counterstained with hematoxylin, dehydrated, and covered with coverslips for further analysis.
For MSI/MSS status evaluation in 101 cases, the IHC staining for MSH2, MSH6, PMS2, and MLH1 was performed on 4 µm thick sections of a representative formalin-fixed, paraffin-embedded (FFPE) tumour tissue block on a Dako Autostainer Link 48. Samples underwent deparaffinization and rehydratation. In the next step, antigen retrieval was performed by incubating slices in EnVision Flex Target Retrieval Solution High pH (Dako, Carpinteria, CA, USA) for 20 min at 95 °C. Prepared samples were incubated with Peroxidase-Blocked Reagent (Dako) and then incubated with one of the following antibodies: Mouse Monoclonal antibody MSH2 (G219-1129), Cell Marque, incubation time: 30 min.; dilution: 1:400; Mouse Monoclonal antibody MSH6 (44), Cell Marque, incubation time: 45 min.; dilution: 1:100; Mouse Monoclonal antibody PMS2 (MRQ-28), Cell Marque, incubation time: 40 min.; dilution: 1:50; Mouse Monoclonal antibody MLH1 (G168-728), Cell Marque, incubation time: 40 min.; dilution: 1:100. After this process, they were put in EnVision FLEX HRP (Dako). Then, antigen–antibody complexes were stained using 3,3′-diaminobenzidine. Finally, tissue sections were counterstained with hematoxylin, dehydrated, and covered with coverslips for further analysis. Tumours were assessed as to whether nuclear staining of invasive tumour cells for MSH2, MSH6, PMS2, and MLH1 was seen in the presence of positive internal control (inflammatory and stromal cells). Tumours with nuclear staining for markers in at least 1% of invasive tumour cells were considered to have positive marker staining. The algorithm on which the interpretation of immunohistochemistry testing was based is presented in the article of Olave and Graham . The MSI status was recognized if one of the following marker layouts were present: MLH1 and PMS2 loss, PMS2 loss, MSH2, and MSH6 loss, or MSH6 loss. We assumed negative staining as the loss.
Briefly, 4 µm thick tissue sections were used for immunohistochemical (IHC) analysis. They were deparaffinized with xylene, rehydrated in graded alcohol, and washed in deionized water. In the next step, antigen retrieval was performed by incubating slices in EnVision Flex Target Retrieval Solution High pH (Dako, Carpinteria, CA, USA) for 20 min at 95 °C. Prepared samples were incubated with Peroxidase-Blocked Reagent (Dako) and then incubated with antibody: (CD8+/144B) Mouse Monoclonal Antibody diluent, incubation time: 40 min.; dilution: 1:100. After this process, they were put in EnVision FLEX HRP (Dako). Then, antigen–antibody complexes were stained using 3,3′-diaminobenzidine. Finally, tissue sections were counterstained with hematoxylin, dehydrated, and covered with coverslips for further analysis.
Tumour-infiltrating lymphocytes (TILs) assessment was performed in 102 specimens. The percentage of tumour-associated lymphatic infiltration was estimated semi-quantitatively on a five-grade scale on the same H&E-stained slides by the two pathologists, according to the criteria defined by Salgado et al. in breast cancer . These include intratumoural lymphocytes with cell-to-cell contact between lymphocytes and tumour cells, and stromal TILs in tumour tissue located dispersed in the stroma within the tumour cells without direct contact, including TILs at the invasive margin. According to the recommendations, stromal TILs were scored as a percentage of the stromal area alone, excluding areas occupied by carcinoma cells. Lymphatic infiltrates outside the tumour borders were not included in the evaluation. A lymphocyte infiltration area lower than 5% was considered TILs 1, whereas 5–25%, 25–50%, and 50–75% of lymphocytes in the stroma were defined as TILs 2, TILs 3 and TILs 4, respectively. More than 75% was defined as TILs 5. Tumour budding was assessed in the same 101 specimens. Tumour buds were estimated in one FOV at a hotspot area in the invasive front under ×20 magnification. The number of buds was adjusted by the normalization factor (1.210). Budding was reported in the following manner: low budding: 0–4 buds; intermediate budding: 5–9 buds; high budding: >10 buds. The mean number of buds per FOV was also used in the statistical analysis.
Homogenates’ supernatants were collected after they were stored at −80 °C. The concentrations of cytokines/chemokines/growth factors were measured in 77 homogenates by the Bio-Plex Pro Human cytokines screening panel 48 cytokines assay (Bio-Rad Laboratories, Hercules, CA, USA) according to the manufacturer’s instructions. In brief, 50 µL aliquot of the sample was diluted 1:2 with sample diluent, incubated with antibody-coupled beads and biotinylated secondary antibodies, and followed by streptavidin-phycoerythrin. Standard curves for each studied parameter were performed using respective cytokine standard solutions. The beads were analysed in the Bio-Plex Array Reader (Bio-Plex Manager 6.2 software from the Bio-Plex 200 System). The intra-assay %CV varied up to 15%, and the inter-assay %CV varied up to 25%, depending on the analysed parameter. The obtained results were then normalised to the corresponding total protein level. This method has been previously used in analysing cytokines in lymphocyte cell culture supernatants and in blood serum samples .
We performed functional annotations analysis based on mRNA expression profiles in the CRC online dataset from the FieldEffectCrc Package among cohort A, consistent with 311 CRC samples . We normalized the matrix data using the DESeq2 package . Then, we divided the cohort into high versus low expressions of VTCN1(B7H4). Gene set enrichment analysis was used to elucidate the potential Hallmarks pathways from The Molecular Signatures Database (h.all.v7.5.symbols.gmt) of B7H4 in CRC in R Studio with fgsea package. The genes with significant differences in expression, including high vs. low B7H4 expression, were screened for GO enrichment analyses (|logFC| > 0.5 and p.adj. < 0.05). Then, we used TCGA-COAD data to explore the association of B7H4 expression with immunogenicity and immune landscape in colorectal cancer. Results of the immune-related scores, mutation analysis, and immune cell infiltration scores, were all obtained from the CAMOIP website ( http://220.189.241.246:13838/#shiny-tab-188 home (accessed on 15 December 2022)).
Data distribution was determined using the Shapiro–Wilk test. The log transformation of the concentrations of the studied proteins provided a better fit for the Gaussian distribution. The data are presented as mean ± SD for the variables with normal distribution, and as median with interquartile range for the variables with non-normal distribution. To compare the tumour and margin concentrations, the paired Student’s t -test (for variables with a normal distribution) and Mann–Whitney U test (for variables with non-normal distribution) were used. Independent variables were also compared using the Student’s t -test and Mann–Whitney U test. Pearson’s coefficient or Spearman coefficient were used for assessing the relationships between the examined variables (for variables with normal and non-normal distribution, respectively). Tau-Kendalls’ Tau Rank Correlation Coefficient was used to determine the association between the levels of the examined proteins, T, and N parameters. p values < 0.05 were considered significant. Hierarchical clustering and principal component analysis were performed to reduce the number of variables and clarify the influence of B7H4 expression on the cytokines profile. The factors obtained in PCA were used as new variables named Dim 1 and Dim 2 according to their proportions of the explained variance. Next, the Dim 1 and Dim 2 values were used in further analyses. p values ≤ 0.05 were considered significant. Statistical analysis was performed using STATISTICA 13 software (Statsoft) and the R Studio (Integrated Development for R. RStudio, PBC, Boston, MA, USA).
3.1. The Expression of B7H4 Is Upregulated in Tumour Tissues In this study, to elucidate the link between B7H4 expression, tumour microsatellite status, and the immunological background, the B7H4 expression level was first compared between tumour tissues and normal tissues adjacent to CRC. A total of 159 pairs of cancer tissues and normal adjacent tissues’ homogenates were analysed for B7H4 level, measured by ELISA. B7H4 was elevated in tumour tissues’ homogenates compared to normal adjacent tissues ( p < 0.0001, ). We also analysed the expression of B7H4 by IHC in 76 tumour section slides. The percentage of B7H4 positive tumours was 78% ( , A). 3.2. The Expression of B7H4 Is Associated with the MSS Status of the Tumour Further, we analysed whether the expression of B7H4 is related to tumour microsatellite status. To elucidate this link we assessed the MSS/MSI status in 77 cases. B7H4 expression in the tumour was more frequent in MSS cases ( p = 0.005, ). Moreover, the tumour-infiltrating lymphocytes score was negatively correlated with the percentage of IHC B7H4 expression level (R = −0.18, p = 0.049, ). No association was found between CD8+ T cell infiltration and B7H4 expression. The B7H4 level was also weakly positively associated with the N parameter (R = 0.16, p = 0.013, ) and negatively with the T feature (R = −0.15, p = 0.023, ). There were no significant associations between B7H4 and other clinicopathological features . 3.3. B7H4 Expression Is Negatively Associated with Pro-Inflammatory Cytokines Next, we investigated the immunological background of B7H4 expression, the cytokinome composition in cancer tissue homogenates, to elucidate the cytokine differential expression pattern between B7H4 IHC positive and B7H4 negative tumours ( A). We divided our 48 cytokine panel by hierarchical clustering to create heatmaps of cytokine expression patterns in B7H4 positive and negative tumours. Heatmaps created two distinct expression patterns. The expressions of the following cytokines were inversely correlated with B7H4. ELISA measurements in B7H4 staining positive tumours: IL-9, IL-18, IP-10(CXCL10), MIG (CXCL9) and SDF-1a (CXCL12) ( B). Moreover, IL-5 alone correlated positively with B7H4 tumour staining percentage and inversely with TILs score ( and A,B). In B7H4 positive tumours, a cluster consisting of IL-5, IL-4, VEGFA, M-CSF, IFN-γ, IL-Ra, IL-8, and IL-1β was chosen for further analysis ( A). We conducted Principal Component Analysis (PCA) with this cluster, obtaining two principal components that explained 66.28% of the overall variance (the sum of PC1 variance was 43.66% and PC2 variance was 22.62%) ( C,D,F). We found a positive correlation between PC2/Dim.2 and B7H4 IHC expression ( E). In the B7H4 positive group, no correlation between panel cytokines and CD8+ T cells infiltration was found; however, in the B7H4 negative group we found that IL-5, IL-10, and MIG positively correlated with CD8+ T cells (respectively, p = 0.0032, p = 0.03, p = 0.015 E–G). In the same group, TILs were positively correlated with IL-2Ra, MIG, and MIP-1b (respectively, p = 0.018, p = 0.035, p = 0.025 A–C). 3.4. Exploration of Immunogenicity and Immune Infiltration Landscape of B7H4 Based on TCGA-COAD Data Subsequently, we further explored the effect of B7H4 on tumour immunity with online available tools. We juxtaposed the B7H4 immune landscape with the landscape of a known immune checkpoint in MSI type CRC PD-1 ( PDCD1 ). A shows the mutational landscape of gene mutations in B7H4 high vs. low expression and PD-1 high vs. low expression of CRC patients, indicating that PD-1 high has a higher frequency of mutations than B7H4 high. Except for the APC and TP53 genes, the other top 20 genes had higher mutation frequencies in the PD-1 high group. The types of mutations were mainly missense and frameshift mutations. The mutation frequencies of the APC and TP53 genes were higher in the B7H4 high group; in contrast, the other genes had higher mutation frequencies in the PD-1 high. In addition, Tumour Mutational Burden (TMB), and Neoantigen Loads (NAL) were significantly higher both in PD-1 high and B7H4 high groups ( B,C). However, the elevation of TMB and NAL scores in the B7H4 high was not as substantial as in PD-1 high group. Finally, the MANTIS score, which is a score that predicts a patient’s MSI status, was significantly higher in PD-1 high group. The higher the Mantis score, the more likely a patient will have an MSI-H status. These findings indicate that B7H4 expression is related to a moderately immunogenic landscape in CRC ( D). The immune scores of Intratumour Heterogeneity, TIL Regional Fraction, Number of Segments, Fraction Altered, and Aneuploidy Score were significantly up-regulated in the B7H4 low expression group ( A), unlike CTA Score, which was decreased in this group. When we compared immune cell infiltration scores calculated by the CIBERSORT algorithm in PD-1 high and B7H4 groups, we noticed that immune composition is in a greater portion shaped by PD-1 immune checkpoint expression. The diversity of iTME was slightly dependent on B7H4 expression. However, the fraction of CD4+ T cells was associated with B7H4 expression. In B7H4 high group, there was a significantly higher percentage of CD4+ memory resting T cells population and, on the other hand, the lower B7H4 expression was associated with a higher percentage of CD4+ memory activated T cells ( B,C). 3.5. Functional Annotations and Predicted Signalling Pathways for VTCN1 (B7H4) Expression For further investigation of the B7H4 relation with transcriptome activity in CRC, we conducted GSEA between low and high B7H4 expression datasets to identify the B7H4-related pathways activated in CRC. Significant differences ( p.adjusted < 0.05) in the enrichment of the Molecular Signature Database Collection for Hallmark gene sets are shown in . The results showed that the top three upregulated pathways for B7H4 were myogenesis, adipogenesis, and oxidative phosphorylation. Furthermore, among upregulated pathways, there were genes upregulated by STAT5 in response to IL-2, genes encoding components of the complement system, which is part of the innate immune system, genes upregulated by IL-6 via STAT3, IFN-γ response, and genes downregulated by KRAS activation. These results showed that a high expression of B7H4 was closely associated with immune responses and malignancy in CRC. Conversely, a high expression of B7H4 has related downregulated pathways associated with the control of the cell cycle. Precisely, the first four most impacted and downregulated pathways were E2F Target, c-MYC, and G2/M Checkpoint, which suggested that the main effect is an impaired cell cycle block, particularly a G2/M phase transition arrest. Functional analyses of B7H4 demonstrated that the significant GO terms are mainly associated with the regulation of the muscle and related to immunological processes. We found that, in the biological process category, B7H4 expression was tightly associated with the regulation of B cell activation, immunoglobulin production, cell recognition, complement activation, and phagocytosis recognition, and was associated in the molecular function category with antigen binding .
In this study, to elucidate the link between B7H4 expression, tumour microsatellite status, and the immunological background, the B7H4 expression level was first compared between tumour tissues and normal tissues adjacent to CRC. A total of 159 pairs of cancer tissues and normal adjacent tissues’ homogenates were analysed for B7H4 level, measured by ELISA. B7H4 was elevated in tumour tissues’ homogenates compared to normal adjacent tissues ( p < 0.0001, ). We also analysed the expression of B7H4 by IHC in 76 tumour section slides. The percentage of B7H4 positive tumours was 78% ( , A).
Further, we analysed whether the expression of B7H4 is related to tumour microsatellite status. To elucidate this link we assessed the MSS/MSI status in 77 cases. B7H4 expression in the tumour was more frequent in MSS cases ( p = 0.005, ). Moreover, the tumour-infiltrating lymphocytes score was negatively correlated with the percentage of IHC B7H4 expression level (R = −0.18, p = 0.049, ). No association was found between CD8+ T cell infiltration and B7H4 expression. The B7H4 level was also weakly positively associated with the N parameter (R = 0.16, p = 0.013, ) and negatively with the T feature (R = −0.15, p = 0.023, ). There were no significant associations between B7H4 and other clinicopathological features .
Next, we investigated the immunological background of B7H4 expression, the cytokinome composition in cancer tissue homogenates, to elucidate the cytokine differential expression pattern between B7H4 IHC positive and B7H4 negative tumours ( A). We divided our 48 cytokine panel by hierarchical clustering to create heatmaps of cytokine expression patterns in B7H4 positive and negative tumours. Heatmaps created two distinct expression patterns. The expressions of the following cytokines were inversely correlated with B7H4. ELISA measurements in B7H4 staining positive tumours: IL-9, IL-18, IP-10(CXCL10), MIG (CXCL9) and SDF-1a (CXCL12) ( B). Moreover, IL-5 alone correlated positively with B7H4 tumour staining percentage and inversely with TILs score ( and A,B). In B7H4 positive tumours, a cluster consisting of IL-5, IL-4, VEGFA, M-CSF, IFN-γ, IL-Ra, IL-8, and IL-1β was chosen for further analysis ( A). We conducted Principal Component Analysis (PCA) with this cluster, obtaining two principal components that explained 66.28% of the overall variance (the sum of PC1 variance was 43.66% and PC2 variance was 22.62%) ( C,D,F). We found a positive correlation between PC2/Dim.2 and B7H4 IHC expression ( E). In the B7H4 positive group, no correlation between panel cytokines and CD8+ T cells infiltration was found; however, in the B7H4 negative group we found that IL-5, IL-10, and MIG positively correlated with CD8+ T cells (respectively, p = 0.0032, p = 0.03, p = 0.015 E–G). In the same group, TILs were positively correlated with IL-2Ra, MIG, and MIP-1b (respectively, p = 0.018, p = 0.035, p = 0.025 A–C).
Subsequently, we further explored the effect of B7H4 on tumour immunity with online available tools. We juxtaposed the B7H4 immune landscape with the landscape of a known immune checkpoint in MSI type CRC PD-1 ( PDCD1 ). A shows the mutational landscape of gene mutations in B7H4 high vs. low expression and PD-1 high vs. low expression of CRC patients, indicating that PD-1 high has a higher frequency of mutations than B7H4 high. Except for the APC and TP53 genes, the other top 20 genes had higher mutation frequencies in the PD-1 high group. The types of mutations were mainly missense and frameshift mutations. The mutation frequencies of the APC and TP53 genes were higher in the B7H4 high group; in contrast, the other genes had higher mutation frequencies in the PD-1 high. In addition, Tumour Mutational Burden (TMB), and Neoantigen Loads (NAL) were significantly higher both in PD-1 high and B7H4 high groups ( B,C). However, the elevation of TMB and NAL scores in the B7H4 high was not as substantial as in PD-1 high group. Finally, the MANTIS score, which is a score that predicts a patient’s MSI status, was significantly higher in PD-1 high group. The higher the Mantis score, the more likely a patient will have an MSI-H status. These findings indicate that B7H4 expression is related to a moderately immunogenic landscape in CRC ( D). The immune scores of Intratumour Heterogeneity, TIL Regional Fraction, Number of Segments, Fraction Altered, and Aneuploidy Score were significantly up-regulated in the B7H4 low expression group ( A), unlike CTA Score, which was decreased in this group. When we compared immune cell infiltration scores calculated by the CIBERSORT algorithm in PD-1 high and B7H4 groups, we noticed that immune composition is in a greater portion shaped by PD-1 immune checkpoint expression. The diversity of iTME was slightly dependent on B7H4 expression. However, the fraction of CD4+ T cells was associated with B7H4 expression. In B7H4 high group, there was a significantly higher percentage of CD4+ memory resting T cells population and, on the other hand, the lower B7H4 expression was associated with a higher percentage of CD4+ memory activated T cells ( B,C).
For further investigation of the B7H4 relation with transcriptome activity in CRC, we conducted GSEA between low and high B7H4 expression datasets to identify the B7H4-related pathways activated in CRC. Significant differences ( p.adjusted < 0.05) in the enrichment of the Molecular Signature Database Collection for Hallmark gene sets are shown in . The results showed that the top three upregulated pathways for B7H4 were myogenesis, adipogenesis, and oxidative phosphorylation. Furthermore, among upregulated pathways, there were genes upregulated by STAT5 in response to IL-2, genes encoding components of the complement system, which is part of the innate immune system, genes upregulated by IL-6 via STAT3, IFN-γ response, and genes downregulated by KRAS activation. These results showed that a high expression of B7H4 was closely associated with immune responses and malignancy in CRC. Conversely, a high expression of B7H4 has related downregulated pathways associated with the control of the cell cycle. Precisely, the first four most impacted and downregulated pathways were E2F Target, c-MYC, and G2/M Checkpoint, which suggested that the main effect is an impaired cell cycle block, particularly a G2/M phase transition arrest. Functional analyses of B7H4 demonstrated that the significant GO terms are mainly associated with the regulation of the muscle and related to immunological processes. We found that, in the biological process category, B7H4 expression was tightly associated with the regulation of B cell activation, immunoglobulin production, cell recognition, complement activation, and phagocytosis recognition, and was associated in the molecular function category with antigen binding .
In this study, we provide the data of B7H4’s role in the immunological response in CRC. Firstly, we performed extensive research on the literature, and based on it we have chosen the B7H4 as the most probable immune checkpoint to be associated with the MSS type CRC. We assessed that B7H4 expression is higher in tumour tissue than in normal adjacent tissue, and is associated with MSS type CRC. Further, we investigated the relation of B7H4 positive and negative tumours to cytokines and we found out that the B7H4 is inversely correlated with antitumour cytokines. Subsequently, the analysis of the immune cell composition associated with a high expression of B7H4 showed the association with CD4+ T cells population. Finally, the GSEA and GO analyses were performed to elucidate what pathways the expression of B7H4 is related to. The B7H4 in recent years, as well as other immune checkpoints, has gained significant attention. There are a few reasons for this fact. Although the receptor for B7H4 remains unrecognized, B7H4, with the growing knowledge about its function, becomes an attractive therapeutic potential not only in the cancer field , but also in the autoimmune diseases field . In normal tissues, human B7H4 mRNA is commonly expressed; however, B7H4 protein exhibits minimal or negative expression, e.g., in the intestine . In contrast, its elevated expression has been observed in many tumour tissues . In line with these findings, we also observed a higher expression of B7H4 in tumour tissue, and 78% positive stained tumours for B7H4 in our cohort. Similarly, Yan et al. reported 76.38% B7H4 positive tumour tissues in CRC . The upregulation of B7H4 was also demonstrated in colorectal cancer in other studies . Moreover, we found that the B7H4 positive tumours were significantly more frequent in microsatellite stable tumours. We further investigated the association between clinicopathological features and B7H4. The results showed a negative correlation between B7H4 expression and immune infiltrating cells (TILs), as we had expected. Tumour-Infiltrating Lymphocytes and CD8+ T lymphocyte density has been reported to be inversely correlated with B7H4 expression in many solid tumours . However, we did not find an association between CD8+ T cell infiltration and B7H4 expression. On the other hand, Yan et al. found that B7H4 positively correlated with CD8+ T cells and negatively correlated with M2 macrophages and Treg cells. We found further contradictory results, as the B7H4 level was weakly negatively correlated with the T feature and positively associated with the N parameter. Another study reported that B7H4 was positively correlated with lymph node metastasis, advanced TNM stage, and poor tumour differentiation in CRC . Although there are a few studies regarding clinicopathological features and their relation to B7H4, due to the inconclusiveness of their results we need to conduct further research on a larger cohort. Understanding the tumour type-specific cytokine profile could provide insights into iTME targeted therapy using the plasticity of pro-tumour or anti-tumour polarised immune cells. The iTME targeted drugs could change the immunosuppressive environment of “cold tumours” to “hot” tumours and restore proper antitumour response . To elucidate the cytokinome relation to the B7H4 tumour phenotype, using the hierarchical clustering method, we found out that the B7H4 tumour positive and B7H4 negative cytokinomes were significantly different. In a subsequent analysis of the B7H4 positive group, we observed a negative correlation between IL-9 and IL-18. IL-9, a Th2 cell cytokine, has been reported as one of the cytokines which are downregulated by B7H4 in lung cancer . In CRC, its relation to B7H4 has not been yet investigated; however, studies show that the level of IL-9 decreased in CRC, along with the progression of the disease, and recent experiments confirmed the antitumour role of IL-9 . Like IL-9, IL-18 also exhibits antitumour properties and its lower expression in CRC has been shown by Feng et al. They also observed that the upregulation of IL-18 leads to the inhibition of colon cancer cell proliferation . CXCL10, CXCL9, and CXCL12 were also negatively correlated with B7H4. CXCL10 and CXCL9 are antitumour chemokines that can inhibit cancer cell proliferation as well as regulate immunity to recruit a variety of immune cells to kill tumour cells . The role of CXCL12 in CRC is, on the other hand, contradictory and its interplay in iTME is more context-dependent; however, current knowledge is more prone to its pro-tumoral effect . Another interesting finding was a positive correlation of IL-5 with B7H4 tumour staining percentage and its negative association with TILs score. The IL-5 main production sources are CD4+ T cells and type 2 innate lymphoid cells (ILC2s); the others are innate immune cells, such as mast cells and eosinophils . In CRC, Th2 cells are considered to be the main source of IL-5 in iTME . In breast cancer, an immune checkpoint blockade increased IL-5 production by CD4+ T cells . We conducted PCA in the B7H4 positive group on the chosen cluster and we found a positive correlation between PC2 and B7H4 IHC expression. PC2 had the main contribution of IL-5, IL-4, and VEGF-A. Th2 immunosuppressive effect on iTME is exhibited by, among others, the secretion of IL-5 and IL-4 . In the B7H4 positive group, we did not find any positive correlation between the cytokines panel, CD8+ T cells infiltration, and TILs. However, in the B7H4 negative group, we found a few cytokines that positively correlated with CD8+ T cells and TILs score. Thus, we may suspect that the main immunosuppressive effect of B7H4 on iTME comes from the tumour cells that express B7H4. After exploring the effect of B7H4 on tumour immunity, we noticed that the mutational landscape of B7H4 high tumours was significantly different compared to the PD-1 related gene mutation load. Moreover, B7H4 expression exhibits a relation to the high mutation frequencies of the APC and TP53 genes, which were not observed in the PD-1 group. A highly similar mutational phenotype was observed by Lin et al. in the MSS/MSI-L CRC group . Immunogenic factors such as Tumour Mutational Burden (TMB) and Neoantigen Loads (NAL) were slightly higher in the B7H4 high group, but not as high as in the PD-1 high group. These findings indicate that the expression of both immune checkpoints is related to a more immunogenic environment, despite the substantially more immunogenic tumour characterizing the PD-1 expression. Like in our results, the overall lower TMB and NAL scores were also reported in the MSS/MSI-L CRC group . Intratumour Heterogeneity, Number of Segments, Fraction Altered, and Aneuploidy Score, were significantly lower in the B7H4 high expression group. Intratumour Heterogeneity (ITH) reversely association with B7H4 could be explained by results that link ITH levels with MSI status and, conversely, its lower score with CMS2 CRC subtype, which corresponds to the canonical epithelial tumour harbouring high chromosomal instability in a microsatellite stable (MSS) context . Along with the Number of Segments, Fraction Altered, and Aneuploidy Score, the results for B7H4 expression are likely related to genomic stable tumours. According to Thorsson et al., the scores measuring DNA damage are related to immune infiltration across many cancer types and subtypes, including CRC . Then, we investigated the immune composition of TME, beyond the TILs score and CD8+ T cell fraction measured in our cohort, on the TCGA-COAD dataset with the CIBERSORT algorithm. The results of that analysis revealed that the B7H4 expression influence on shaping the iTME is not very impactful, especially when compared to the role of PD-1 expression on immune composition. However, the B7H4 high group had a significantly higher percentage of T cells, CD4+ memory resting population and, on the contrary, a lower percentage of T cells, CD4+ memory activated. B7H4 suppressive effects on CD4+ and CD8+ T cells have been demonstrated in in vitro studies, where they exhibited the ability to suppress T cell effector functions, including inflammatory cytokine production and cytolytic activity . Similar effects have been reported in a few cancer studies . We have found that B7H4 is involved in the regulation of the CD4+ T cells effect in iTME, but we have not found its link to the CD8+ T cells population. It could be due to the overall lower tumour lymphocyte infiltration in B7H4 high expression tumours. However, along with the results from the cytokines panel, we suspect that the immunosuppressive role of B7H4 is the effect of the interplay with Th2 cells and the Th2 cytokines. Nevertheless, further investigation of iTME in B7H4 positive tumours is required to elucidate its exact effect on the immunological landscape in CRC. Finally, we used GSEA to survey the transcriptome of CRC cases to compare B7H4 high-expressed CRC versus low-expressed. The results showed that among the upregulated immune-related pathways, there were genes upregulated by STAT5 in response to IL-2, genes encoding components of the complement system, which are part of the innate immune system, and genes upregulated by IL-6 via STAT3, and IFN-γ response. IL-6/STAT3 as well as the IL-2/STAT5 pathway activates downstream target genes to protect tumour cells from apoptosis, increase tumour cell proliferation, cell cycle progression, invasion, and metastasis, and are involved in drug resistance . Similar results for the B7H4 high expression group were observed in ovarian cancer . Additionally, the downregulated pathways associated with the control of cell cycle at E2F Target and G2/M Checkpoint are related to the high expression of B7H4. In detail, E2F Target, c-MYC, and G2/M Checkpoint were the first four most impacted and downregulated pathways, suggesting that the main effect is an impaired cell cycle block, in particular a G2/M phase transition arrest. E2F is an essential protein that regulates the cell cycle with a block at the G2/M phase . On the other hand, downregulated c-MYC target pathways could seem contradictory as c-MYC is an oncogene; however, this pathway includes many genes involved in restraining cell growth . Functional analyses of B7H4 also confirmed its association with immunological-related processes. Significantly enriched GO terms referred to a wide range of immunological functions, from adaptive, with a predominance of the humoral response, to innate immunity, such as the regulation of B cell activation, immunoglobulin production, cell recognition, complement activation, phagocytosis recognition, phagocytosis engulfment, and antigen binding. These results indicate that B7H4 expression is linked to several pro-tumorigenic immunological processes and impaired cell cycle regulatory machinery.
In conclusion, the findings of this study revealed that B7H4 expression is upregulated in CRC and is associated with the MSS status of tumours. Furthermore, B7H4 may play a pivotal role in the shaping of iTME in the microsatellite stable consistent immune landscape. B7H4 is positively correlated with antitumour immunosuppressive iTME. The iTME in the presence of B7H4 positive tumours might be mainly shaped by Th2 cells and their cytokine network. Additionally, the B7H4 positive tumour’s genomic landscape is similar to the one of MSS type CRC, which supports our finding that B7H4 expression is associated with microsatellite stable tumour types with a preserved mismatch repair gene mechanism. Taken together, the data suggest that B7H4 might be a potentially promising therapeutic target in MSS type CRC.
|
Impact of the COVID-19 Pandemic on Medical Oncology Workload: A Provincial Review | f5a8a594-0e3f-42c2-9788-b6a4c0802209 | 10047010 | Internal Medicine[mh] | Cancer is the leading cause of death in Canada, with increasing incidence and mortality . Almost half of Canadians will be diagnosed with cancer in their lifetime and incidence is projected to increase a further 79% by 2032 . Eastern and Central Canada are found to have the highest predicted incidence rates, with the second highest age-standardized provincial rates observed in the province of Nova Scotia . Nationally, Cancer Care programs face significant resource limitations, including personnel, infrastructure, and escalating treatment expenditures, all of which were exacerbated due to the COVID-19 pandemic . Workforce evaluation and planning, as well as adaptability, is critically important to the provision of optimal patient care and current metrics do not reflect the full scope of medical oncology practice, nor the lasting changes from the COVID-19 pandemic. Currently, there are very limited published data on Medical Oncology workload benchmarks and even less on actual workload. The most often cited metric in the Canadian literature is defined as number of new cancer patient consultations per oncologist per year . A recent global study, which included a Canadian cohort, was conducted to address medical oncology workload and describe allocation of resources . Two other reports published by Canadian groups have attempted to identify flaws in the current medical oncology workforce assessment model, suggesting a change is needed . The authors emphasized the importance of developing a comprehensive workload metric to account for escalating complexity of care and evolving treatment landscapes more accurately . These studies represent the only data available in the Canadian context since the Cancer Care Ontario Systemic Therapy Task Force report was published in 2000 . While the current demands are already straining the system—secondary to Canada’s aging population—including longer cancer patient survival on systemic treatment and availability of novel treatments, the COVID-19 pandemic has exacerbated this further. COVID-19 has significantly impacted all healthcare systems and delivery of care, with federal and provincial governments recommending variable safety measures, including limitations on in-person encounters when possible, and a shift toward provision of virtual care. Both surgical and radiation oncology have faced significant limitations in care delivery at times throughout the pandemic, including cancelations and operating room closures due to over-capacity hospital admissions, physician discretion, staff shortages due to COVID-19, and cancelation of regular radiation consultation and treatment, in an effort to minimize exposure and maximize patient safety . Across Canada, there was a 20% reduction in cancer surgeries in the first 6 months of the pandemic, as per data collected by the Canadian Institute for Health Information . This increased the demand on medical oncology, for example, with the need for introduction or continuation of chemotherapy, as opposed to surgical resection in some cases . Though numeric data reflecting that specific impact has been difficult to collect, this has been reported by most Medical Oncologists practicing in Canada and has been noted globally . With the lack of published information on medical oncology workload, we devised a study to review what the current workload for medical oncologists is in the Nova Scotia Central Zone, which accounts for the greatest portion of the province’s cancer care delivery. We ask whether the workload has evolved over time and what the impact of the COVID-19 pandemic was on the workload, if any. We hope this may inform a comprehensive framework to capture workload that could aid in future discussions on resource allocation and workforce planning in Canada.
The Nova Scotia Cancer Centre within the Nova Scotia Health Central Zone (NSHCZ), affiliated with Dalhousie University, is the academic center for the province of Nova Scotia, Canada, providing consultative and ongoing cancer care for approximately 72% of the cancer patients in the province, a total population of just over 1,000,000 people. It accounts for 80% of the provincial medical oncologists (MO), while the remaining 20% practice in the 2 other geographically separated centers in the province. All are under the jurisdiction of the Cancer Care Nova Scotia Program (CCNSP) and have a similar clinical practice. Data were extracted for all patient encounters from the Oncology Patient Information System (OPIS). OPIS is the electronic booking system for all patient encounters at the Nova Scotia Cancer Center. Encounter types included are new patient consultations, follow-up visits, which could be undertaken in person or virtually, and outpatient treatment visits. Additionally, we captured telephone toxicity assessments (nursing-led phone calls to patients of approximately 10 min duration to review treatment toxicities, then reviewed and approved by the Medical Oncologist to authorize subsequent treatment cycles) and chart reviews (scheduled opportunities for the physician to review, discuss, and follow-up on patient investigations or clinical status; these often include patient phone calls, but are not intended as full follow-up appointments). Both in-person and virtual encounters are captured through OPIS and data were included for the full complement of 12–16 full time Medical Oncologists (the number of staff increased over the study period) and one physician extender (in this case a nurse practitioner with extra training in oncology, who provides clinical support), as well as the resident trainees in oncology, in the outpatient setting over the study period. Encounter data were collected per the Nova Scotia Health fiscal year. The fiscal year begins 1 April and ends 31 March of the following calendar year. We included data from 1 April 2014 to 31 March 2022 (thus recorded as fiscal years 2014–2021), including the duration of the first provincially mandated pandemic-associated restrictions in March 2020. The total number of encounters was extracted and sorted by encounter type for each year. As the OPIS system does not categorize patient encounters by tumor type, to assess for this, each encounter was categorized into disease site based on the tumor sites treated by the individual physician. Thus, a physician who treats 50% breast and 50% GI had their patient encounters split in half, as an approximation of the workload for each disease site.
3.1. Overall Clinical Workload The total number of patient encounters, including new consults, follow up visits, telephone toxicities, and chart reviews were analyzed from 2014 to 2021. Overall, there were 17,670 patient encounters in 2014, compared to 33,310 in 2021, translating to an average annual increase in total encounters of 9.5%. In 2014, there were 2228 new patient consults, increasing to 3254 by 2021 (a 46% increase). The number of follow-up visits increased from 8902 to 13,485 (52% increase). Chart reviews and telephone toxicities increased by 146% and 200%, respectively, (5625 chart reviews in 2014 compared to 13,822 in 2021 and 915 telephone toxicities in 2014 compared to 2749 in 2021. To consider the change in the number of medical oncologists located in the NSHCZ over time, we analyzed the number of patient encounters per medical oncologist (PEMO). In 2014, there were 12 medical oncologists (MO) and 1 nurse practitioner (NP) in the NSHCZ, leading to a 1359.2 PEMO. By 2021, there were 15 MOs and 1 NP in the NSHCZ, which led to a 53.2% increase in the PEMO (2081.9). This increase is primarily the result of the increase in chart checks per MO (432.7 in 2014 and 863.9 in 2021; 99.6% increase) and telephone toxicities per MO (70.4 in 2014 and 171.8 in 2021; 144% increase). The number of new consults per MO increased by 18.7% (171.4 in 2014 and 203.4 in 2021), with an increase in follow-up visits per MO of 23.1% (684.8 in 2014 and 842.8 in 2021). presents the annualized data for both in-person and virtual care encounters, demonstrating increases across all encounter types. In-person encounters increased by 45.6%, from 10,968 to 15,965, between 2014 and 2021. Within the same timeframe, virtual care encounters, including virtual follow-up appointments, telephone toxicities, and chart reviews, grew more substantially, from 6702 to 17,345 (158.8% increase). Of note, even before the pandemic, virtual encounters were rising annually (2017–2018 = 16.5% increase, 2018–2019 = 27.7% increase, 2019–2020 = 26.9% increase) when compared to in-person encounters (2017–2018 = 6.1% increase, 2018–2019 = 3.6% increase, 2019–2020 = 2.2% increase). This clearly suggests that virtual care provision was increasing even prior to the COVID-19 pandemic. 3.2. Clinical Workload during the COVID-19 Pandemic There was a 7.9% increase in overall patient encounters between 2019 and 2020 (29,447 in 2020 compared to 27,295 in 2019). Although virtual encounters appeared to be on the rise prior to the pandemic, there was a 46% increase in virtual care encounters from 2019–2020, with a corresponding decrease in in-person encounters over this period (27%). As the height of the first and second waves of the pandemic settled in Nova Scotia, we saw a return of in-person encounters (15,965 in 2021, 10,208 in 2020, and 14,057 in 2019), with a small reduction in virtual care when compared to 2020 (17,345 in 2021 and 19,238 in 2020). There was still a 31.8% increase in virtual care when compared to 2019 (17,345 in 2021 to 13,164 in 2019). Overall, the greatest number of patient encounters in 2020 were chart reviews, followed by follow-up visits, new consultations, and telephone toxicities. All types of encounter had increased, with the greatest percentage increase noted in chart reviews (virtual) and follow-ups (some in-person, some virtual) across the study timespan, regardless of tumor/disease site. This was maintained in 2021. 3.3. Clinical Workload by Tumor Site To look at the changes in workload over time by disease site, we separated each patient encounter by attending physician and subsequently categorized them into four major tumor sites (breast, gastrointestinal (GI), genitourinary (GU), and lung). Note, encounters for skin, central nervous system, and sarcoma were excluded from this analysis. Additionally, it should be emphasized that these are estimates based on the treating physician’s typical patient load. When you look at total volume, no matter which disease site is looked at, there is an increase in all four of the workload metrics. GI cancer treaters saw an increase from 2014 to 2021 in total consults (669.2 to 975.5), return visits (2284 to 3227.4), telephone toxicities (257.2 to 440), and chart checks (1121.3 to 2749.2) ( A). Breast cancer treaters saw an increase from 2014 to 2021 in total consults (712.3 to 949), return visits (2740.8 to 3370.8), telephone toxicities (343.5 to 577), and chart checks (1343.5 to 3382) ( B). GU cancer treaters saw an increase over the same time frame in total consults (247.75 to 298), return visits (1228.25 to 1876.8), telephone toxicities (104 to 443), and chart checks (1130 to 1682) ( C). Lung cancer treaters saw an increase in total consults (392.8 to 558.3), return visits (1773.8 to 2324.7), telephone toxicities (343.5 to 525.8), and chart checks (1343.5 to 2931.1). As the number of MOs changed from 2014 to 2021, we determined the number of each encounter type seen per MO by each disease site. In both breast and GU disease sites, we saw a decrease in 2021 in the consults seen per MO and a decrease in follow-up visits in breast cancer seen, likely reflective of the COVID-19 pandemic . There were increases in consult and follow-up visits in both GI (consults 9.3%, follow-up visits 6.0%) and lung disease sites (consults 13.7%, follow-up visits 4.8%). Moreover, three disease sites saw an increase in chart checks (GI: 83.9%; Breast:76.2%; Lung: 72.6%), while GU saw a slight dip in chart checks per MO (0.8%). All four disease sites saw an increase in telephone toxicities per MO (GI: 28.4%; Breast:17.5%; GU: 184.0%; Lung: 224.6%.). To estimate the work one consult generates, we compared the number of follow-up visits, telephone toxicities, and chart reviews to the number of consults in 2014 and 2021. A GI consult led to a 1.9 chart review increase in 2021, when compared to a GI consult in 2014, but no change in follow-up visits or telephone toxicities. A breast consult saw a similar 1.7 chart review increase in 2021, with no change in the number of follow-up visits or telephone toxicities. A lung consult led to a 1.9 increase in chart reviews and a 0.6 increase in telephone toxicities per consult in 2021, but no change in follow-ups. GU on the other hand, saw an increase of 1.3 visits, 1.1 telephone toxicities, and 1.0 chart reviews per consult.
The total number of patient encounters, including new consults, follow up visits, telephone toxicities, and chart reviews were analyzed from 2014 to 2021. Overall, there were 17,670 patient encounters in 2014, compared to 33,310 in 2021, translating to an average annual increase in total encounters of 9.5%. In 2014, there were 2228 new patient consults, increasing to 3254 by 2021 (a 46% increase). The number of follow-up visits increased from 8902 to 13,485 (52% increase). Chart reviews and telephone toxicities increased by 146% and 200%, respectively, (5625 chart reviews in 2014 compared to 13,822 in 2021 and 915 telephone toxicities in 2014 compared to 2749 in 2021. To consider the change in the number of medical oncologists located in the NSHCZ over time, we analyzed the number of patient encounters per medical oncologist (PEMO). In 2014, there were 12 medical oncologists (MO) and 1 nurse practitioner (NP) in the NSHCZ, leading to a 1359.2 PEMO. By 2021, there were 15 MOs and 1 NP in the NSHCZ, which led to a 53.2% increase in the PEMO (2081.9). This increase is primarily the result of the increase in chart checks per MO (432.7 in 2014 and 863.9 in 2021; 99.6% increase) and telephone toxicities per MO (70.4 in 2014 and 171.8 in 2021; 144% increase). The number of new consults per MO increased by 18.7% (171.4 in 2014 and 203.4 in 2021), with an increase in follow-up visits per MO of 23.1% (684.8 in 2014 and 842.8 in 2021). presents the annualized data for both in-person and virtual care encounters, demonstrating increases across all encounter types. In-person encounters increased by 45.6%, from 10,968 to 15,965, between 2014 and 2021. Within the same timeframe, virtual care encounters, including virtual follow-up appointments, telephone toxicities, and chart reviews, grew more substantially, from 6702 to 17,345 (158.8% increase). Of note, even before the pandemic, virtual encounters were rising annually (2017–2018 = 16.5% increase, 2018–2019 = 27.7% increase, 2019–2020 = 26.9% increase) when compared to in-person encounters (2017–2018 = 6.1% increase, 2018–2019 = 3.6% increase, 2019–2020 = 2.2% increase). This clearly suggests that virtual care provision was increasing even prior to the COVID-19 pandemic.
There was a 7.9% increase in overall patient encounters between 2019 and 2020 (29,447 in 2020 compared to 27,295 in 2019). Although virtual encounters appeared to be on the rise prior to the pandemic, there was a 46% increase in virtual care encounters from 2019–2020, with a corresponding decrease in in-person encounters over this period (27%). As the height of the first and second waves of the pandemic settled in Nova Scotia, we saw a return of in-person encounters (15,965 in 2021, 10,208 in 2020, and 14,057 in 2019), with a small reduction in virtual care when compared to 2020 (17,345 in 2021 and 19,238 in 2020). There was still a 31.8% increase in virtual care when compared to 2019 (17,345 in 2021 to 13,164 in 2019). Overall, the greatest number of patient encounters in 2020 were chart reviews, followed by follow-up visits, new consultations, and telephone toxicities. All types of encounter had increased, with the greatest percentage increase noted in chart reviews (virtual) and follow-ups (some in-person, some virtual) across the study timespan, regardless of tumor/disease site. This was maintained in 2021.
To look at the changes in workload over time by disease site, we separated each patient encounter by attending physician and subsequently categorized them into four major tumor sites (breast, gastrointestinal (GI), genitourinary (GU), and lung). Note, encounters for skin, central nervous system, and sarcoma were excluded from this analysis. Additionally, it should be emphasized that these are estimates based on the treating physician’s typical patient load. When you look at total volume, no matter which disease site is looked at, there is an increase in all four of the workload metrics. GI cancer treaters saw an increase from 2014 to 2021 in total consults (669.2 to 975.5), return visits (2284 to 3227.4), telephone toxicities (257.2 to 440), and chart checks (1121.3 to 2749.2) ( A). Breast cancer treaters saw an increase from 2014 to 2021 in total consults (712.3 to 949), return visits (2740.8 to 3370.8), telephone toxicities (343.5 to 577), and chart checks (1343.5 to 3382) ( B). GU cancer treaters saw an increase over the same time frame in total consults (247.75 to 298), return visits (1228.25 to 1876.8), telephone toxicities (104 to 443), and chart checks (1130 to 1682) ( C). Lung cancer treaters saw an increase in total consults (392.8 to 558.3), return visits (1773.8 to 2324.7), telephone toxicities (343.5 to 525.8), and chart checks (1343.5 to 2931.1). As the number of MOs changed from 2014 to 2021, we determined the number of each encounter type seen per MO by each disease site. In both breast and GU disease sites, we saw a decrease in 2021 in the consults seen per MO and a decrease in follow-up visits in breast cancer seen, likely reflective of the COVID-19 pandemic . There were increases in consult and follow-up visits in both GI (consults 9.3%, follow-up visits 6.0%) and lung disease sites (consults 13.7%, follow-up visits 4.8%). Moreover, three disease sites saw an increase in chart checks (GI: 83.9%; Breast:76.2%; Lung: 72.6%), while GU saw a slight dip in chart checks per MO (0.8%). All four disease sites saw an increase in telephone toxicities per MO (GI: 28.4%; Breast:17.5%; GU: 184.0%; Lung: 224.6%.). To estimate the work one consult generates, we compared the number of follow-up visits, telephone toxicities, and chart reviews to the number of consults in 2014 and 2021. A GI consult led to a 1.9 chart review increase in 2021, when compared to a GI consult in 2014, but no change in follow-up visits or telephone toxicities. A breast consult saw a similar 1.7 chart review increase in 2021, with no change in the number of follow-up visits or telephone toxicities. A lung consult led to a 1.9 increase in chart reviews and a 0.6 increase in telephone toxicities per consult in 2021, but no change in follow-ups. GU on the other hand, saw an increase of 1.3 visits, 1.1 telephone toxicities, and 1.0 chart reviews per consult.
This study has attempted to provide an account of the changes in MO workload over time and to review the impact of the COVID-19 pandemic on the workload in hopes of aiding in future discussions on resource allocation planning in Canada. The Systemic Therapies Task Force was established in 2000 by Cancer Care Ontario (CCO) and determined the workload benchmark for Canadian Medical Oncologists based on an optimal average annual target of new patient consults per practitioner. This target was originally derived by calculating the number of hours of patient care available per academic Oncologist per year divided by the number of hours estimated for patient care required per specific tumor type . This has been the metric used nationally since then to measure workload and inform cancer program hiring practices. According to the study by Fundytus et al., high-income countries, including Canada, seemed to fall within the target range for new consults. However, it is important to note that although the median reported clinical volume was consistent with the proposed annual target, half of the survey participants exceeded it . A recent snapshot assessment of medical oncology workload in Europe showed Western European countries would see on average 175 consults per year, similar to the Canadian standard, while Eastern European countries saw an average of 225 consults in a year . We believe we are the first to look at the different encounter types for medical oncologists in Canada. Our data suggest an increase in volume of patient encounters over time. No matter which metric was reviewed, the total volume of consults, follow-up visits, telephone toxicities, and chart reviews increased over time. Over the 8-year period, five additional MOs were hired, two replacing two retiring MOs and three who were completely new. To adjust for the increased number of MOs, we normalized the total volumes based on the total number of cancer treaters (MOs and NPs). We interestingly saw that the increase was primarily due to telephone toxicities and chart reviews, while consults and follow-up visits remained relatively stable with some marginal change depending on the disease site. Marhold et al. have previously published an analysis of patient consult volumes in a large Austrian center, showing a similar rise in consults from 2006 to 2018 . One thought on the lack of proportional growth in consults and follow-ups per Medical Oncologist is that space limitations, including clinic space and chemotherapy chair space, limit the number of patients that can be seen in each clinic. We know treatment in a cancer center is a multifaceted process, requiring personnel (physicians, nursing, pharmacists), space (clinic and chemotherapy chair space), and supportive care (occupational and physiotherapists, dieticians, social workers) provision . Moreover, access to cancer care is directly dependent on the number of providers available . Once space and providers are full, no other patient encounters can occur beyond telephone toxicities and chart checks, as these do not require physical space to be performed. This is also reflected in the encounters seen by an MO, broken down by disease site. No matter which disease site was evaluated, the increase in workload metrics were primarily seen in the chart reviews and telephone toxicities, likely as a result of the evolution of patient assessment during the pandemic. With the rapidly evolving treatment landscape and associated complexities, using new consults as the only metric, without consideration of patient encounters in the form of chart reviews, follow-ups, and telephone toxicities (CFT), seems extremely limited. New patient consultations accounted for only 9.8% of all patient encounters in 2021, illustrating the limitation of this one-dimensional metric. Another way of viewing these data is to look at the time spent on patient encounters. For general approximation, consults are a minimum of 1 h in length, follow up visits 30 min, telephone toxicities 10 min, and chart reviews 5 min. Based on the data from 2021, this would equate to 3255 h spent on consults, 6742.5 h spent on follow up visits, 459.8 h on telephone toxicities, and 1151.8 h on chart reviews (a total of 11,609.1). Using those approximations, new consults only accounted for 28% of the total encounter time spent on/with patients in 2021, highlighting the importance of considering other metrics beyond new consults to encompass actual workload. A study by a group in Spain attempted to better-capture medical oncologist workload by assigning a target percentage value for various academic activities (including clinical, research, and educational activities), as well as a proposed time valuation for clinical encounters (for example, time spent for new encounters versus follow-ups) ; thus, offering an example of one such potential strategy for workload metric diversification and inclusion. Compared to when the CCO was issued, the field of Medical Oncology has undergone revolutionary changes. Many metastatic disease sites with previously limited available treatments now have numerous therapeutic options, with longer survivorship, including melanoma , lung, , and hepatocellular carcinoma as examples. The advent of both immune checkpoint inhibitors and targeted molecules has expanded the treatment and toxicity landscape far beyond chemotherapy . Additionally, an increasing number of malignancies have more complex treatment regimens relevant to a broader population, for example chemoimmunotherapy in lung cancers regardless of PD-L1 status or multi-agent chemotherapy as standard of care first line treatment in many pancreatic cancers . This was clearly reflected in the patient encounters seen at our individual site over the study period. We have seen both lung and genitourinary disease sites translate into an increase in patient encounters over time, potentially reflecting the availability of new treatments such as immunotherapy, which were not yet regularly integrated into the treatment landscape for colorectal or breast cancers. With newer treatments, physicians may watch their patients more closely for potential toxicities and could employ check-ins with the patient through chart checks and telephone toxicities. Furthermore, with the current national shortage in primary care providers , an increase in primary care support required for medical oncology patients will likely continue to contribute to additional encounters. While our data focus specifically on clinical workload, it would be inappropriate not to acknowledge the additional demands on a Medical Oncologist’s time, including time needed for teaching of trainees, research, guideline development, interaction with other institutions, and patient advocacy, to list a few. Patient care is just the tip of the iceberg of a Medical Oncologist’s workload. . Despite pandemic restrictions, where patients were relegated to a larger proportion of virtual care, there was still an 7.9% increase in overall patient encounters, with 35% of these occurring in-person. COVID-19 highlighted a trend toward significantly more virtual care provision, a trend that was becoming apparent pre-pandemic. Virtual oncology care was occurring in Nova Scotia before the onset of the COVID-19 pandemic, although its role was limited. Reliance on virtual care increased during the pandemic in the 2020 and 2021 fiscal years, aiming to deliver care whilst maintaining patient and workplace safety. Survey data collected by the Nova Scotia Department of Health and Wellness on the patient experience of virtual care in Nova Scotia during the first months of the COVID-19 pandemic (April to August 2020) found that patients appeared to be highly satisfied with their virtual care experience . Moreover, virtual care has been shown to be highly successful in various oncologic settings and sites, thus supporting an important adaptation suitable to many pre-existing barriers to cancer care access for Canadians, including geographic distance from centers of care, patient morbidity, and cost of travel . With current space and personnel limitations, one of the only avenues for growth is through virtual care. The COVID-19 pandemic created new pressures on the system as well, with significant decreases in uptake of screening and surgical delays across the country . In Nova Scotia, breast cancer screening decreased 62% and colorectal cancer screening decreased 65% in 2020 compared to 2019, over the 6-month period from March to December . These interruptions and delays will impact cancer care in the months and years to come, as screening resumes, with more new diagnoses and more advanced diagnoses . A workforce planning model has been developed in Radiation Oncology , but to our knowledge, nothing similar exists in Medical Oncology. Others are working to create workforce models to help build a stronger future for Canadian Cancer Care, but robust information accounting for all aspects of clinical workload is lacking . This is a critical limitation that needs to be overcome, as development of a robust and meaningful workload assessment and care planning continues. A study from Spain, attempting to better-capture medical oncologist workload, assigned a target percentage value for various academic activities (including clinical, research, and educational activities), as well as a proposed time valuation for clinical encounters (for example, new consults versus follow-ups) . These types of strategies, with percentages or points assigned to clinical and academic tasks, may allow us to capture the full scope of Medical Oncologists’ practice, as well as create a metric that can be adjusted and adapted over time, to meet the ever-evolving demands of cancer care provision. We must acknowledge the limitations within our study. As this was a retrospective analysis, we are dependent on the patient encounters logged into OPIS. Any missed or improperly labelled encounters would not have been captured, potentially underrepresenting workload. Moreover, there are workload elements (some mentioned above) that are not directly included in this analysis. Patient phone calls to the nursing/physician team, physician contact with other patient care members, physician pages (the paging system is the primary mode of formal healthcare provider communication at NSHCZ), and others are missing. In February 2020, the Nova Scotia Cancer Centre (and OPIS) began tracking phone calls from patients to their respective cancer care teams (nursing/medical oncologist). For the 2-month period from February through March 2020, 1261 calls were received. This increased from 1 April 2020–31 March 2021, to 8616, and, subsequently, from 1 April 2021 to 31 March 2022, to 13,843. The additional phone calls would increase the total encounters at the cancer center to 47,153. This will be important to capture in the future development of a robust workload metric. As patient encounters are not coded based on disease site, there is no good way of tracking the role of disease sites and treatment options on patient encounters. With the potential adoption of a new electronic patient recording system, we hope to be able to stratify patient encounters to address this more thoroughly in the future. We must finally acknowledge that other centers in Canada may practice differently than Nova Scotia. The size of catchment area, number of visits, the length of visits, and the presence of physician extenders would likely play a role in the outcomes of that province’s analysis. As we are the first to have taken this approach, it would be very interesting to see if similar outcomes are seen in some of the other larger provinces. We would expect a similar trend as space utilization is already currently maximized at many centers, leading to longer wait times, and a reliance on virtual care is becoming more standardized.
As the incidence of cancer increases and treatments continue to expand and evolve, Medical Oncologists globally can expect a continued increase in workload. This trend has been identified through the national Canadian Cancer statistics, as well as locally with patient encounter data obtained at our individual site. In addition to increasing cancer incidence, the changes instituted at the onset of the COVID-19 pandemic highlight some further proportion of lasting change in the delivery of cancer care, in particular, virtual care delivery. New patient consultation metrics, taken in isolation, do not accurately reflect the current trends in Medical Oncology clinical workload. Creating and implementing a new algorithm that is adaptable to the changing therapeutic landscapes would be a new way to capture workload and could translate into the resource needs required to inform healthcare hiring practices and allocation of space and other resources to optimize patient care. Our data are the first to break down MO workload in a Canadian province. The hope is that these data will start the conversation about how to take a critical look at the metrics and resource planning in the country to provide creative solutions to deal with the ever-growing demands on medical oncologists.
|
Variation in general practice referral rate to acute medicine services and association with hospital admission. A retrospective observational study | 95087cde-ebf9-4653-a91f-ed134cbb69a7 | 10047615 | Family Medicine[mh] | A small fraction of people seen in primary care each year are referred to hospital services for acute medical care. Acute medical assessment units are designed to provide urgent assessment and treatment for patients with potentially life-threatening conditions for whom primary care physicians (general practitioners) deem outpatient or community management impractical or unsafe. The majority of patients referred to acute medical care are appropriate and either hospital admission is arranged, urgent care given with safe discharge or serious pathology excluded. Referrals continue to increase , and, when they exceed the capacity of an assessment unit, can negatively affect patient experience and treatment outcome. The “NHS Long Term Plan,” a UK government strategy launched in January 2019 aspires to treat patients where possible in the community and primary care setting. Similarly, the national clinical strategy for health in Scotland aims to reduce variation in practice with a much greater focus on supported self-management, anticipatory care planning, and community-based medical treatment. Historically, variation in general practice (GP) referral rates to outpatient secondary care has been widely reported, however data relating to referral rates for acute medical assessment remain sparse. Duffy et al. described acute medical admission rates in Tayside, Scotland, in 1996/1997 and reported a variation in emergency admission rates to hospital of 1.8-fold between the top and bottom deciles of practices. The majority of this variance (64%) was attributed to deprivation status and age of the practice population. Blatchford et al. described a similar study in Glasgow, Scotland, also examining admission rates, and reported a near 2-fold variation between top and bottom deciles after adjustment for age, sex, and deprivation score of each practice population. Both these studies addressed the question of variation in admission rate of the practice population rather than referral rate by the general practitioners for acute hospital medical assessment. Further they included patients admitted via ambulance emergency and self-attendance and as such variation in active GP referral behaviour for acute medical assessment remains unclear. In most large UK centres, triage of individual urgent referrals to acute medicine by secondary care medical staff is not feasible and so acute medicine assessment workload is influenced substantially by clinical decision-making in primary care. While the decision to refer a patient for acute medical assessment rests with the assessing general practitioner, multiple variables feed into this decision including patient actual or perceived acuity of illness, presence of social support, patient and practitioner preference and expectation, distance from hospital, patient comorbidity, and presence of accessible alternatives to admission. Similar factors will influence the decision to admit a patient after initial hospital clinical assessment when coupled with initial examination and investigations and the preference of the assessing acute clinician . This retrospective study aimed to describe GP referral variation for acute medical assessment and to evaluate the extent to which this variation was explained by these complex socioeconomic and demographic factors and subsequently to understand how referral variation related to need for inpatient hospital care. Therefore, 2 distinct analyses were performed. 1) Individual GP-level analysis of referral rates for acute medical assessment and associations with practice population-level variables of age, deprivation, hospital proximity, care home residence, and practice list size. 2) Patients referred for acute medical assessment arrive at the receiving hospital and undergo examination and baseline laboratory and radiological examination in an assessment area. If inpatient admission is subsequently needed for ongoing clinical care then the patient is admitted. If inpatient care is not required the patient is discharged for community-based follow-up. Need for inpatient care was used as a marker of “appropriateness” of referral. Patient-level analysis was performed for this cohort including socioeconomic factors, markers of comorbidity, biochemical markers of clinical acuity, and the referral behaviour of the source practice.
Data were obtained from the TRAKcare inpatient clinical management system (Intersystems, USA), which is in use in all hospital sites in NHS Lothian, a health board providing care for around 800,000 patients in South East Scotland. Data were examined from secondary care hospitals over a 2 year 9 month period from April 2017 to January 2020 (Royal Infirmary of Edinburgh and Western General Hospital, Edinburgh). The study period was truncated at January 2020 due to the reported changes in health seeking behaviour during the COVID-19 pandemic lockdown. There are 121 practices in NHS Lothian. Twenty-six practices admit to an alternative hospital site for which data were unavailable and were thus excluded. The university practice and homeless access practice were excluded as were 2 practices which merged during the study period and 3 practices with incomplete practice population data leaving 88 practices in total with a combined patient population of 666,112. For practice-level analysis, practice list size, percentage of practice population over 65, percentage in the 2 most deprived Scottish Index of Multiple Deprivation (SIMD) quintiles, and percentage patient population in care home residency were obtained from Public Health Scotland Information Services Division ( www.isdscotland.org ) and examined as continuous variables. SIMD uses 7 domains (housing, skills and training, income, employment, health, education, geographic access and crime) to produce an area-based ranking index including standardized mortality ratio, mental health prescribing data and hospital admissions for drug and alcohol use and is used as an estimate of comorbidity and deprivation. Distance from the major hospital admission site for each practice was calculated using the geographical coordinates of each practice postcode and “geosphere” package in R. To assess GP referral rates, patients referred from GPs in Lothian for acute medical assessment were calculated (referrals per 100 patient-years) for each practice. Only patients from routine in hours GP service were included (Monday to Friday 0800–1800). Referrals from other sources (emergency medicine, GP out of hours service, self-presentation) and to other specialties (e.g. surgical assessment, minor injury assessment) were excluded. For patient-level analysis individual SIMD quintile was obtained from the SIMD database and examined as an ordinal variable, Charlson score was calculated using “comorbidity” package in R based on hospital discharge codes from the preceding 5 years and examined as an ordinal variable (0, 1–2, >2) based on population frequency. Care home residence was identified from individual patient address and examined as a dichotomous variable. AKIN acute kidney injury score was calculated based on admission creatinine as previously described and examined as a dichotomous variable (no AKI = “0,” AKIN stage 1–3 = “1–3”). For practice-level data, quasi-Poisson regression modelling was used to manage over dispersion within the dataset. For patient-level data, mixed model binary logistic regression was performed with hospital admission as the dependent variable and hospital included as a random effect. Model assumptions were checked visually including colinearity between dependent variables and were deemed acceptable. Practice referral rate for acute medical assessment was expressed as referrals per 100 patient-years and entered as a continuous dependent variable with age, deprivation, care home residence, distance from hospital site and practice list size as explanatory variables. When interpreting the model output a P value of <0.05 was deemed significant. Model performance was assessed using observed versus expected counts with adjusted R 2 to express goodness of fit. Patient-level analysis and graphical outputs of the referred population were performed ( n = 42,424) using “finalfit” package. It is recognized that the presence of absence of an acute kidney injury is not available to primary care clinicians on initial assessment although clinical features associated with sepsis or dehydration may be present. As such AKIN score was not included in the reported model but sensitivity analysis was performed demonstrating inclusion had minimal effect on statistical outcome. The study received local Caldicott Guardian approval and was undertaken in line with local information governance procedures. Analyses were performed using R versions 4.1.2 “Bird Hippie.” R script of full analysis can be found at https://github.com/marcus-lyall/GP-referral-variance.git .
There were 42,424 referrals for acute medical assessment from 88 practices during the study period. The median crude rate of referral per practice was 2.26 per 100 patient-years (min = 0.34, max = 5.33). This was equivalent to a 15.7-fold variation between the lowest and highest referring practice, a 4.9-fold variation between the median of the first and tenth deciles of referring practices (median first decile = 0.88, tenth decile = 4.36 referrals per 100 patient-years per practice) and a 2.53-fold variation between first and fourth quartiles (first quartile = 1.29, fourth quartile = 3.27 referrals per 100 patient-years) . Practice population age and deprivation were significantly associated with acute medical referral rate ( t statistic 5.26 and 5.11, P < 0.01). There was a trend to negative association with care home residency ( t statistic −1.98, P = 0.051) but there was no significant association with distance to hospital or practice list size . The completed model explained 67% of the variation seen (observed/expected adjusted R 2 0.67) . Consistent with this, adjustment for confounders reduced but did not remove substantial variation in referral rate which remained 2.15-fold (1.43 and 3.07 referrals per 100 patient-years per practice between first and fourth quartiles) . Examining patient-level data for referred and clinically assessed patients demonstrated that increasing age, being out with the most affluent deprivation quintile and increasing Charlson comorbidity score were independently associated with need for medical admission . Care home residency demonstrated an over 2-fold risk of admission with only 130/993 (13%) of patients referred from a care home being discharged following initial assessment (odds ratio [OR] 2.32 [1.92–2.81, P < 0.001]). Patients from the highest referring quartile of GP practices were over 50% less likely to require hospital admission than those referred by the lowest referring practice quartile (OR 1.53 [1.41–1.66, P < 0.001]) and ~30% less likely to require admission than patients from practices within the interquartile range (third referral quartile = OR 1.28 [1.21–1.36, P < 0.01], second referral quartile = OR 1.30 [1.23–1.37, P < 0.001]). AKIN score, while an important marker of clinical acuity, is not available to primary care practitioners at the time of referral decision-making and was therefore included in a secondary sensitivity analysis. AKIN stages 1–3 were associated with increased likelihood of admission (adjusted OR 3.77 [3.45–4.13, P < 0.001]) but did not substantially change the effect of source practice referral quartile .
This large retrospective observational study demonstrates a high variation in referral rates between primary care GPs in Lothian for acute medical assessment which is incompletely explained by the available demographic variables of the practice populations studied. To our knowledge this is the first study to examine in detail GP referral practice for acute medical assessment. At the practice level, referral variance was explained to a significant degree by the reported model ( R 2 0.67) however significant variation persists after correction for established factors of high clinical need. It has been well described that age and deprivation are key correlates of emergency secondary healthcare access and so the finding that correction for these factors partially attenuates the degree of variation is expected. , , There was no significant effect of distance to hospital where a negative association may be expected. This variable was examined in a continuous fashion with the median distance or only 4.3 km (IQR 0.48). It may be that an effect does exist with rural (30–40 km) versus urban practices however this may be confounded by differences in rural practice organization. The finding that considerable variation continues to exist after correction for these variables is likely in part due to the lack of patient-level population wide data allowing detailed adjustment for deprivation, age, and comorbidity on an individual basis but perhaps also due to the influence of unmeasurable primary and secondary care factors identified in the conceptual framework . Interestingly, practice list size was not associated with an increase in referral rate. We hypothesized that larger practice list sizes would result in less individual patient care continuity, recently demonstrated to increase unscheduled care utilization, hospital admission, and mortality in the Norwegian population. It may be that increasing practice list size is not significantly associated with a change in care continuity or alternatively that reductions in continuity result only in an increase in patient emergency care self-presentation, not measured in these analyses. There was a highly significant sequential negative association between the referral rate of the source GP and the likelihood for admission after adjustment for confounders. It is accepted that variation in admitting behaviour will exist between secondary care clinicians however approximately 30 individual senior physicians overseeing >100 individual junior clinicians assessed the referred patient populations at both hospital sites and as such there is no rationale this would bias 1 practice over another throughout the study period. Individual hospital site may influence referral rate due to hospital infrastructure referral pathways and ambulatory care availability and this was adjusted for in the mixed model methodology. Detailed patient-level data were available for patients attending for acute medical assessment. These data demonstrate an expected increase in need for hospital admission with age, deprivation, comorbidity, care home residency, and the presence of an acute kidney injury. While age, comorbidity, and acute kidney injury are likely to reflect acuity of presenting pathology, the independent association between deprivation and need for admission is consistent with previous works and may also indicate reduced levels of available social support hindering early discharge. , Interestingly, our analysis demonstrates a 2.3-fold independent association with care home residence and need for admission after medical assessment with only 13% of care home residents immediately discharged. The authors suggest a major barrier to early discharge in these patients is the need for ambulance transport which is not available until to the following day. Patients in the care home setting already have domiciliary support in place and in many cases may be better served community hospital at home services. Widening access to and improving utilization of these services as well as ensuring appropriate individual care plans provides an opportunity to reduce referrals and improve patient experience in the care home population. Taken together these findings suggest that variation in primary care clinician risk threshold, patient population preference or expectation and perhaps GP culture plays an important role in the adjusted variance observed. Previous studies have examined outpatient referral and also total medical admission rates from all sources. , , , However, these studies were not specific to the acute medicine/primary care interface, which is under extreme pressure in the UK health service. O’Donnell et al. reviewed the literature up to the year 2000 and concluded that the measured patient demographics, individual practice, and individual general practitioner factors were responsible for only a minority of the observed variance in referrals with the majority of variance unexplained. This review extensively focussed on nonurgent referrals for outpatient assessment with a minor focus on acute assessment and only 8 of 29 studies included inpatient referrals. Further these included referrals for all specialties including surgery, mental health, and paediatrics and the majority of included data are now over 20 years old and may not accurately reflect current practice in the rapidly changing health care landscape. A more recent analysis of ~130,000 patient referrals from the health improvement network database in primary care demonstrated that age, gender, and deprivation significantly affected referral behaviour in a disease-specific manner. Referring clinician gender and experience factors may also be important , and targeted education with rolling feedback to general practitioners may significantly reduce variation in practice. High variation in referral rates to the emergency department and direct specialty admission at the clinician level in the out of hours GP setting as also been reported, which may be due to differences in clinical trainee status, locality of in hours work or by GP attitudes and beliefs around referral threshold. Strengths of this study include its size and duration of study period. GP personnel fluctuate with periods of leave and locum use, which could affect referral patterns. The examination of a 2 year 9 month period should have ameliorated the impact of these factors. The use of a single computer-based clinical management systems across the region allowed a highly specific dataset to be curated for patients referred for medical assessment by their GP practice with the exclusion of self-presenting patients, out of hours GP referrals, ambulance service referrals and those patients immediately triaged to minor injury, surgical and other specialty services. In addition, the practice where the patient was currently registered was recorded on referral, compensating for those moving practice during the study. A key limitation in this study is the use of group practice as an independent variable and the population-level explanatory variables assigned to each practice to explain variation. Analysis at the practice-level measures the collective referral behaviour of up to twenty general practitioners which may vary widely and reflect not only the referral practice of the clinician but the organizational infrastructure of the practice. Clinician-level factors such as experience, skill mix, and gender known to be associated with differences in referral rate are unaccounted for in this study. , , Practice-level factors such as access to appointments, training practice status, the presence and utilization of extended evening or weekend hours service and reliance on locum staff may also have significant effect and cannot be controlled for in this study. Community factors such as hospital at home access, community hospital access, crime rates, and transport links may play an important role and data to adjust for these are lacking. In addition, individual patient-level factors such as prior healthcare contact, prescribing burden, educational level, marital status, and household support may also play a key role not adequately represented in the SIMD variable. Increasing granularity and linkage of routinely collected healthcare and national disease databases coupled with advances machine learning to assign relative weight to these factors shows promise at predicting hospital admission and may help clarify key dependent variables in future. We have used need for medical admission as a measure of GP practice or patient population risk tolerance and, in essence, assumed that practices where a high proportion of patients were immediately discharged, demonstrated a lower tolerance for management in the community setting. It is important to note that need for admission is not an absolute indicator of an appropriate referral. Patients with complex home support are often admitted due to transport issues and departmental crowding when they may not require inpatient management. Furthermore, referral for exclusion of potentially serious pathology such as myocardial infarction, pulmonary embolus, and subarachnoid haemorrhage, following which a patient is discharged, is often entirely appropriate. The optimal admission to discharge ratio for a given population is currently unknown. Further work is now required to identify the population of patients referred to acute medicine who could conceivably have been managed in the community and to identify modifiable community and GP-level factors associated with high acute medicine utilization. Populations for whom acute medical referral could be avoided are patients with a low risk of pathology for whom outpatient investigation or rapid access clinic (chest pain, respiratory, neurovascular, ambulatory care) would suffice and patients with a high level of frailty or physical dependence for whom urgent social support is not available. Future research in this field must include more detailed patient- and consultation-level data to examine both likelihood of referral for specific complaint and subsequent clinical outcome. Point of care testing, strengthening primary and secondary care communication with high referring practices and extending ambulatory care services and alternatives to acute medicine referral may all play a role in reducing variation in referral rate and relieving highly congested acute unscheduled care pathways. ,
There is a high variation in referral rates for acute medical assessment between GPs in this study, a significant proportion of which may be due to practice level, patient or community factors that are as yet unexplained. When this is achieved there may be an opportunity to optimize the use of acute medical assessment units and advance the current strategic vision of community-based care.
cmac097_suppl_Supplementary_Material Click here for additional data file. cmac097_suppl_Supplementary_Checklist Click here for additional data file.
|
Future of Artificial Intelligence Applications in Cancer Care: A Global Cross-Sectional Survey of Researchers | f571ddc1-3d47-4a6f-b24e-a599fb90f430 | 10047823 | Internal Medicine[mh] | Cancer is one of the leading causes of death in every country globally . World Health Organization (WHO) data for the year 2022 show that cancer is the second leading cause of death among noncommunicable diseases, resulting in 9.3 million deaths per year, second only to cardiovascular diseases (17.9 million) . This prominence as a leading cause of death places cancer as the target of major innovative efforts by academia and the pharmaceutical industry. These efforts have brought several new diagnostic and treatment technologies to the market, such as immunotherapy and precision medicine . More recently, great interest has been observed in the search for new diagnostics and treatments involving artificial intelligence (AI) . AI is a broad field that comprises various technologies such as deep learning, machine learning, natural language processing, neural networks, and rule-based systems . As is the case in many other fields of healthcare, the integration of AI in cancer care is expected to reshape the existing scenario in the future . For example, as a predictive modeling and early detection, AI could be used to analyze data from a variety of sources, such as electronic health records, genetic information, and environmental data, to predict an individual’s risk of developing cancer and to tailor prevention strategies accordingly . AI-related applications may reduce screening costs , provide more reliable diagnostics , improve prognostics , and aid in the discovery of new drugs . Several areas of cancer care are expected to benefit from AI-related applications, including cancer radiology and clinical oncology . In the United States alone, more than 70 AI-related applications for different specialties and tumors had received approval from the Food and Drug Administration (FDA) by 2021 . However, different inhibiting factors may affect the creation and adoption of these AI-related applications in cancer care. It may be hampered by ethical and regulatory issues associated with legal uncertainty about responsibility and accountability for AI-supported decisions or the lack of improvements in medical applications . The difficulty of incorporation into clinical practice itself and the lack of standardization in cancer-related health data may also hamper these new developments. While the potential applications of AI in cancer care are promising, the barriers to their widespread adoption, such as those discussed, create uncertainty around the success of these technologies in the fight against cancer. This study aims to provide a glimpse into the future of AI in cancer care by gathering the perspectives of researchers involved in this field. To do so, we conducted a cross-sectional survey of authors of recent peer-reviewed articles related to cancer and AI retrieved from the Web of Science Core Collection (WoS). Previous studies sought to anticipate the future of AI use in cancer care through a literature review . Overall, most of them focused on specific areas, such as precision medicine , clinical oncology , diagnosis , and cancer target identification . Other studies focused on broader aspects of cancer care, such as current applications and future perspectives . Although our study also conducted a literature review, it did so mainly to identify the relevant aspects posed for the future of the topic and, thus, to design the survey questionnaire applied to the authors above. More comprehensive in scope, our study differs from the others in its method (survey research) and in presenting a common vision about the future of AI in cancer care based on the expectations of more than a thousand researchers in the field. According to the expectations of the survey participants, in this study, we present (a) the likelihood of the occurrence of future events pointed out in the scientific literature (such as reducing screening costs and improving diagnostics), (b) the AI applications most likely to be successful in the future (cancer diagnostics and early cancer detection, for example), (c) the areas of cancer care that are most likely to benefit from AI in the future (e.g., pathology and cancer radiology), and (d) the factors most likely to hamper the use of AI applications in cancer care in the future (such as incorporating AI applications into clinical practice and ethical or regulatory issues).
2.1. Literature Review and Questionnaire We conducted a literature review to identify pressing issues of future AI applications in cancer care. To do so, we selected review articles recently published in WoS-indexed journals. The identification of publications was made with the following query: TI = ((“Artificial Intelligence” OR “Computational Intelligence” OR “Machine Intelligence” OR “Computer Reasoning” OR “Computer Vision System*” OR “Machine learning” OR “Transfer Learning” OR “Deep Learning” OR “Hierarchical Learning”) AND (Tumor OR Tumors OR Neoplasm OR Neoplasms OR Neoplasia OR Neoplasias OR Cancer OR Cancers)) Timespan: 20 September 2020 to 20 September 2022 (Index Date) SCI-EXPANDED Document Types: Review Article Languages: English The search strategy combined thesaurus terms related to AI and cancer collected in the Medical Subject Headings (MeSH), US National Library of Medicine ( https://www.ncbi.nlm.nih.gov/mesh , acessed on 10 September 2022). In WoS advanced search mode, we used the tag Title (TI) to search for these terms in the titles of review articles published in the last two years (20 September 2020, to 20 September 2022). We used the Science Citation Index Expanded (SCI-EXPANDED) to retrieve only documents published in journals of science. Only review articles written in English were included in the literature review. The search was done on 20 September 2022, and retrieved 274 publication records, which were imported in plain text format into VantagePoint 11.0 data/text mining software. After reading their titles and abstracts, we selected 74 publication records for further analysis. The records were then imported into Citavi 6.1 reference management software, where we performed the literature review and managed the references. We then downloaded the complete review articles in PDF format, which were entirely read. Of the 74 review articles read, 38 were selected for the literature review and preparation of the survey questionnaire. The questionnaire considered a 10-year horizon (2022–2032) and was divided into five parts. Initially, we introduced the survey with information about the purpose of the study and aspects related to voluntary participation, absence of sensitive questions, data collection and treatment, and anonymization of results. In addition, the respondents were asked whether they consented to participate in the study—if so, they continued with the questionnaire, and if not, the questionnaire was terminated. Thus, all respondents who participated in this survey gave us their informed consent to use the data collected for research purposes. In the second part, the respondents’ level of knowledge about AI applications in cancer care was asked. Respondents who self-reported having high, good, or some knowledge were qualified for the survey and followed up, while those who reported having no knowledge were disqualified and did not answer the questionnaire. The third part asked about the likelihood of occurrence of different AI developments in cancer care: (i) whether it would be widely used, (ii) more reliable diagnostics, (iii) reduce screening cost, (iv) improve follow-up services, (v) aid the discovery of new drugs, (vi) grade and classify cancer, and (vii) improve prognostics. The third part also asked the respondents to rank different applications of AI in cancer care, considering their likelihood of success in the next ten years, as well as for them to report—considering the recent FDA approval of artificial intelligence applications in cancer care and their prospects —which specific area of interest would benefit the most from AI use in cancer care. The fourth part of the questionnaire had two questions regarding general barriers to using AI in cancer care. Respondents could select, among five options (including others), the one they considered most important (e.g., ethical and regulatory issues) and, in sequence, select the most important specific barrier from the option they selected previously (e.g., algorithmic bias). The bibliographic references for each question in the questionnaire are listed in . Finally, the fifth and sixth parts of the questionnaire were optional and were not included in the calculation of fully answered questionnaires. The fifth part consisted of an open-ended question, where the respondents were invited to leave comments, suggestions, and criticisms on the questionnaire. The last part covered five demographic questions, where the respondents could report their academic degree, professional occupation, institutional affiliation, professional experience, and region where they live. The demographics of the respondents do not influence the results of this type of study . They were used to present an overview of the study participants. 2.2. Survey Respondents The survey respondents were authors of articles or review articles on AI and cancer published between 20 September 2020 and 20 September 2022) and indexed in WoS SCI-EXPANDED. We used the same search strategy of the literature review, but with two changes: (1) instead of the tag Title, we used the tag Topic (TS), which, besides the title, searches for records in the abstract and keywords; and (2) we added the document type articles. The search was conducted on 20 September 2022, and retrieved 15,533 publication records. We imported these records into VantagePoint 11.0, where we retrieved 28,263 author emails, excluding duplicates. We then created a CSV file with author data (email, name, and article title) and used an in-house developed python code to link 84% (23,740) of these emails to their owners—which allowed us to send personalized emails to most respondents. 2.3. Survey Procedures, Ethical Aspects, and Limitations of the Study The list of respondents with linked and unlinked emails was imported into the SurveyMonkey online survey platform, where the questionnaire was designed, and the survey was conducted. After uploading, the number of emails was reduced to 25,000 due to 2726 bounced emails and 537 opted-out contacts (people who previously opted out of surveys conducted through SurveyMonkey). Before conducting the formal study, we validated the questionnaire through a pilot with a random sample of 1000 researchers (3.54% of total respondents). The questionnaire was available for eight consecutive days after the invitation email was sent, and up to three reminder emails were sent to non-responders. In both the invitation and reminder emails, and on the first page of the questionnaire, respondents were informed: (i) of the purpose of the study, (ii) that sensitive data would not be asked, (iii) that the data collected would be anonymized in the results, (iv) that participation would be voluntary, (v) and that informed consent for participation in the study would be sought. In the pilot, we evaluated the questionnaire (application routine, consistency, internal logic, completion rate, response time, etc.) and allowed the respondents to make observations, suggestions, and criticisms. The 11 respondents who answered the questionnaire did not suggest changes. Then, neither the questionnaire nor the survey procedures were changed in the formal study, and the answers collected in the pilot were included in the study’s results. The pilot was conducted between 16 and 23 October, and the formal study between 24 October and 4 November 2022. Given that the only personal data of the participants (name and email) were obtained from article records made available in a database of scientific publications (WoS), and considering voluntary participation, absence of sensitive questions, anonymization of results, and obtaining informed consent, examination of the study by an ethics committee was not necessary. In addition, this study followed the Brazilian Resolution number 510 of 7 April 2016 (Official Federal Gazette: https://www.in.gov.br/materia/-/asset_publisher/Kujrw0TZC2Mb/content/id/22917581 (accessed on 10 September 2022)), which exempts from registration and evaluation by an ethics committee public opinion research with unidentified participants. The procedures adopted in this study followed previous studies that have surveyed researchers and sought to anticipate future possibilities in science and technology . Furthermore, just like them, it shares the same limitations. One of them is the limited diversity of respondents, which is a consequence of the method of identification and selection of respondents in scientific articles. Thus, if respondents are authors of scientific articles, naturally, they will be predominantly composed of researchers and professors linked to universities and research organizations. Another limitation is related to the possibility of respondents’ optimism bias. As they are authors of articles related to AI and cancer—and thus invested in this subject—they may be more optimistic about the future of the technology than other respondent profiles (patients, managers, business people, and politicians, for example). Because they are invested in developing scientific and technological knowledge, they are naturally among the most qualified to inform about future possibilities of AI use in cancer care. As the future of AI uses in cancer care is still quite uncertain, weighing the opinions of researchers on the topic against those of other respondent profiles (with less scientific and technological knowledge) does not seem methodologically relevant to the study. A final limitation refers to the self-attribution of knowledge level by the respondents in the questionnaire. Unfortunately, it is not possible for the authors of this study to assign levels of knowledge to each respondent nor to assess whether the self-assigned level of knowledge is coherent. Thus, the self-assigned level of knowledge is a function of how the respondents assess their knowledge in the area. In any case, all participants in this study are authors of peer-reviewed scientific articles related to AI and cancer indexed in WoS, reducing the chances of including opinions from people without knowledge of the subject. 2.4. Statistical Analysis of the Sample To verify the statistical differences between the responses of different knowledge levels, we performed analyses in the software IBM-SPSS Statistics 28. To select the appropriate analysis method, we used the Kolmogorov–Smirnov and Shapiro–Wilk tests to check the distribution of the sample. These two tests are the most widely used to check the normality of the sample . The null hypothesis of both is that the data distribution is normal. Its rejection implies that the data do not have a normal distribution. The results of these two tests indicated that the distribution of responses from different levels of knowledge is not normal. Thus, we used the Kruskal–Wallis non-parametric test to check for possible differences between the groups of respondents. This is the non-parametric test recommended for testing differences between more than two group samples from the same population. This test verifies if the groups’ responses are so different that they cannot be considered to belong to the same population. Its null hypothesis is no difference between the groups’ responses . The results of the Kruskal–Wallis test indicated that the responses of the three groups of respondents were not statistically different. Therefore, the responses of the three groups were reported in aggregate. All statistics results are depicted in the .
We conducted a literature review to identify pressing issues of future AI applications in cancer care. To do so, we selected review articles recently published in WoS-indexed journals. The identification of publications was made with the following query: TI = ((“Artificial Intelligence” OR “Computational Intelligence” OR “Machine Intelligence” OR “Computer Reasoning” OR “Computer Vision System*” OR “Machine learning” OR “Transfer Learning” OR “Deep Learning” OR “Hierarchical Learning”) AND (Tumor OR Tumors OR Neoplasm OR Neoplasms OR Neoplasia OR Neoplasias OR Cancer OR Cancers)) Timespan: 20 September 2020 to 20 September 2022 (Index Date) SCI-EXPANDED Document Types: Review Article Languages: English The search strategy combined thesaurus terms related to AI and cancer collected in the Medical Subject Headings (MeSH), US National Library of Medicine ( https://www.ncbi.nlm.nih.gov/mesh , acessed on 10 September 2022). In WoS advanced search mode, we used the tag Title (TI) to search for these terms in the titles of review articles published in the last two years (20 September 2020, to 20 September 2022). We used the Science Citation Index Expanded (SCI-EXPANDED) to retrieve only documents published in journals of science. Only review articles written in English were included in the literature review. The search was done on 20 September 2022, and retrieved 274 publication records, which were imported in plain text format into VantagePoint 11.0 data/text mining software. After reading their titles and abstracts, we selected 74 publication records for further analysis. The records were then imported into Citavi 6.1 reference management software, where we performed the literature review and managed the references. We then downloaded the complete review articles in PDF format, which were entirely read. Of the 74 review articles read, 38 were selected for the literature review and preparation of the survey questionnaire. The questionnaire considered a 10-year horizon (2022–2032) and was divided into five parts. Initially, we introduced the survey with information about the purpose of the study and aspects related to voluntary participation, absence of sensitive questions, data collection and treatment, and anonymization of results. In addition, the respondents were asked whether they consented to participate in the study—if so, they continued with the questionnaire, and if not, the questionnaire was terminated. Thus, all respondents who participated in this survey gave us their informed consent to use the data collected for research purposes. In the second part, the respondents’ level of knowledge about AI applications in cancer care was asked. Respondents who self-reported having high, good, or some knowledge were qualified for the survey and followed up, while those who reported having no knowledge were disqualified and did not answer the questionnaire. The third part asked about the likelihood of occurrence of different AI developments in cancer care: (i) whether it would be widely used, (ii) more reliable diagnostics, (iii) reduce screening cost, (iv) improve follow-up services, (v) aid the discovery of new drugs, (vi) grade and classify cancer, and (vii) improve prognostics. The third part also asked the respondents to rank different applications of AI in cancer care, considering their likelihood of success in the next ten years, as well as for them to report—considering the recent FDA approval of artificial intelligence applications in cancer care and their prospects —which specific area of interest would benefit the most from AI use in cancer care. The fourth part of the questionnaire had two questions regarding general barriers to using AI in cancer care. Respondents could select, among five options (including others), the one they considered most important (e.g., ethical and regulatory issues) and, in sequence, select the most important specific barrier from the option they selected previously (e.g., algorithmic bias). The bibliographic references for each question in the questionnaire are listed in . Finally, the fifth and sixth parts of the questionnaire were optional and were not included in the calculation of fully answered questionnaires. The fifth part consisted of an open-ended question, where the respondents were invited to leave comments, suggestions, and criticisms on the questionnaire. The last part covered five demographic questions, where the respondents could report their academic degree, professional occupation, institutional affiliation, professional experience, and region where they live. The demographics of the respondents do not influence the results of this type of study . They were used to present an overview of the study participants.
The survey respondents were authors of articles or review articles on AI and cancer published between 20 September 2020 and 20 September 2022) and indexed in WoS SCI-EXPANDED. We used the same search strategy of the literature review, but with two changes: (1) instead of the tag Title, we used the tag Topic (TS), which, besides the title, searches for records in the abstract and keywords; and (2) we added the document type articles. The search was conducted on 20 September 2022, and retrieved 15,533 publication records. We imported these records into VantagePoint 11.0, where we retrieved 28,263 author emails, excluding duplicates. We then created a CSV file with author data (email, name, and article title) and used an in-house developed python code to link 84% (23,740) of these emails to their owners—which allowed us to send personalized emails to most respondents.
The list of respondents with linked and unlinked emails was imported into the SurveyMonkey online survey platform, where the questionnaire was designed, and the survey was conducted. After uploading, the number of emails was reduced to 25,000 due to 2726 bounced emails and 537 opted-out contacts (people who previously opted out of surveys conducted through SurveyMonkey). Before conducting the formal study, we validated the questionnaire through a pilot with a random sample of 1000 researchers (3.54% of total respondents). The questionnaire was available for eight consecutive days after the invitation email was sent, and up to three reminder emails were sent to non-responders. In both the invitation and reminder emails, and on the first page of the questionnaire, respondents were informed: (i) of the purpose of the study, (ii) that sensitive data would not be asked, (iii) that the data collected would be anonymized in the results, (iv) that participation would be voluntary, (v) and that informed consent for participation in the study would be sought. In the pilot, we evaluated the questionnaire (application routine, consistency, internal logic, completion rate, response time, etc.) and allowed the respondents to make observations, suggestions, and criticisms. The 11 respondents who answered the questionnaire did not suggest changes. Then, neither the questionnaire nor the survey procedures were changed in the formal study, and the answers collected in the pilot were included in the study’s results. The pilot was conducted between 16 and 23 October, and the formal study between 24 October and 4 November 2022. Given that the only personal data of the participants (name and email) were obtained from article records made available in a database of scientific publications (WoS), and considering voluntary participation, absence of sensitive questions, anonymization of results, and obtaining informed consent, examination of the study by an ethics committee was not necessary. In addition, this study followed the Brazilian Resolution number 510 of 7 April 2016 (Official Federal Gazette: https://www.in.gov.br/materia/-/asset_publisher/Kujrw0TZC2Mb/content/id/22917581 (accessed on 10 September 2022)), which exempts from registration and evaluation by an ethics committee public opinion research with unidentified participants. The procedures adopted in this study followed previous studies that have surveyed researchers and sought to anticipate future possibilities in science and technology . Furthermore, just like them, it shares the same limitations. One of them is the limited diversity of respondents, which is a consequence of the method of identification and selection of respondents in scientific articles. Thus, if respondents are authors of scientific articles, naturally, they will be predominantly composed of researchers and professors linked to universities and research organizations. Another limitation is related to the possibility of respondents’ optimism bias. As they are authors of articles related to AI and cancer—and thus invested in this subject—they may be more optimistic about the future of the technology than other respondent profiles (patients, managers, business people, and politicians, for example). Because they are invested in developing scientific and technological knowledge, they are naturally among the most qualified to inform about future possibilities of AI use in cancer care. As the future of AI uses in cancer care is still quite uncertain, weighing the opinions of researchers on the topic against those of other respondent profiles (with less scientific and technological knowledge) does not seem methodologically relevant to the study. A final limitation refers to the self-attribution of knowledge level by the respondents in the questionnaire. Unfortunately, it is not possible for the authors of this study to assign levels of knowledge to each respondent nor to assess whether the self-assigned level of knowledge is coherent. Thus, the self-assigned level of knowledge is a function of how the respondents assess their knowledge in the area. In any case, all participants in this study are authors of peer-reviewed scientific articles related to AI and cancer indexed in WoS, reducing the chances of including opinions from people without knowledge of the subject.
To verify the statistical differences between the responses of different knowledge levels, we performed analyses in the software IBM-SPSS Statistics 28. To select the appropriate analysis method, we used the Kolmogorov–Smirnov and Shapiro–Wilk tests to check the distribution of the sample. These two tests are the most widely used to check the normality of the sample . The null hypothesis of both is that the data distribution is normal. Its rejection implies that the data do not have a normal distribution. The results of these two tests indicated that the distribution of responses from different levels of knowledge is not normal. Thus, we used the Kruskal–Wallis non-parametric test to check for possible differences between the groups of respondents. This is the non-parametric test recommended for testing differences between more than two group samples from the same population. This test verifies if the groups’ responses are so different that they cannot be considered to belong to the same population. Its null hypothesis is no difference between the groups’ responses . The results of the Kruskal–Wallis test indicated that the responses of the three groups of respondents were not statistically different. Therefore, the responses of the three groups were reported in aggregate. All statistics results are depicted in the .
From 25,000 invited researchers, 1030 agreed to participate in the study—after excluding 66 respondents who did not consent to participate. This number gave us a response rate of 4.12%, which is compatible to the response rate found in other future-oriented studies that used the same survey method . Of these 1030 researchers, 26.02% reported having a high knowledge of AI in cancer care, while 42.72% and 28.93% reported having good and some knowledge, respectively. Only 2.33% self-reported having no knowledge of the subject. They were disqualified from the survey and thus did not answer the questionnaire. Of the 1006 questionnaires from high, good, and some knowledge respondents, 881 were completely filled (87.57% of total valid responses). Considering the invited researchers as the survey population (25,000), the minimum required sample size to obtain a 5% margin of error with a confidence level of 95% was 394 completely filled questionnaires. Our 881 completed filled questionnaires gave us a representative sample with a 95% confidence level and a margin of error of 3%, which was high enough to generate consistent results. The demographics of the respondents are depicted in . Most had a Doctoral degree (79.93%), while 15.85% had a Master’s degree. As for their occupation, the majority were professors or researchers (65.03%), followed by physicians/clinicians, and Doctoral and Master’s students, with approximately 14% each. Most respondents worked in universities or research organizations (75.12%) and 16.47% in hospitals or similar organizations. Concerning the length of experience, there was a similar distribution among respondents with experience between 5 and 10 years, between 10 and 20 years, and with more than 20 years—approximately 30% each. As for their location, 42.47% lived in Europe, 28.92% in Asia, and 17.19% in North America (including Central America and the Caribbean). shows the likelihood of seven different future events derived from the expected use of AI in cancer care. More than 50% of the respondents expected all seven events to occur before ten years. AI grading and classifying cancer was the event that most respondents considered likely in this timespan (73.13%), followed by providing more reliable diagnoses (69.08%). In turn, aiding in discovering new drugs and being widely used in cancer care obtained the highest likely percentages after ten years (37.96% and 32.34%, respectively). The unlikely percentages were low, with the highest also for AI aiding new drug discovery (6.77%). depicts the average ranking of AI applications considered most likely to be successful in cancer care in the next ten years. The respondents’ most preferred application was cancer diagnostics (3.79), followed closely by early cancer detection (3.77). Therapy administration was the least preferred application to be successful in this timespan (2.4). Considering AI applications in cancer care recently approved by the FDA , the respondents indicated the areas of interest that would benefit most from the use of AI in the next ten years . For about one-third of them, cancer radiology would benefit the most, followed by pathology (27.02%). Of the available options, gynecology was considered to benefit the least from using AI (1.46%). A small part of the respondents (2.58%) pointed out that other areas not listed in the questionnaire would benefit the most. A small part of them (2.58%) pointed out that other areas not listed in the questionnaire would benefit the most. The lack of standardization in cancer-related health data was considered the most likely barrier to AI use in cancer care . It was the choice of 41.95% of respondents, followed by the difficulty of incorporation into clinical practice (26.06%) and ethical or regulatory issues (22.82%). The lack of improvement in medical applications and other barriers amounted to less than 10% of total responses (894). Of the 375 respondents who selected the lack of standardization in cancer-related health data, 47.20% believed that the main reason for the lack of standardization originated from difficulties in testing, validating, certifying, and auditing AI algorithms and systems. Another 35.47% attributed the lack of standardization to difficulties accessing and sharing patient data. Among the 233 respondents who chose the difficulty of incorporating AI into clinical practice, 46.78% attributed their choice to the difficulty of aligning AI to the specific context of clinical practice. Finally, among those who chose ethical or regulatory issues (204), 66.67% believed that the use of AI in cancer care was likely to be hampered by uncertainties about legal responsibility and accountability for AI-supported clinical decisions.
The responses from 1006 authors of articles (experts) on AI and cancer provided us with an in-depth understanding of AI’s potential opportunities and challenges in cancer care. Aligning with the number of SaMD (Software as Medical Devices) recently approved by the US-FDA , a significant proportion of respondents believed that cancer radiology (34.64%) and pathology (27.02%) would be the areas that would benefit most from the use of AI in cancer care over the next decade. In turn, only 1.46% chose gynecology. Some possible reasons could be the complexity of the field, which involves not only diagnosis and treatment, but also a wide range of mother and child conditions and responses to the treatment. It requires intuitive decision-making from the consultant. Another reason could be the lack of data to train AI models. Gynecologists may be skeptical about AI’s reliability and effectiveness in real-time medical practice, which could limit its adoption. Interestingly, this study’s results align with reports that gynecology is the least SaMD-approved field by the FDA . On the other hand, an increasing number of studies report using AI to evaluate images such as MRI, colposcopy, fetal ultrasound , and hysteroscopy . If we had asked for gynecology–radiology, perhaps we would have had more respondents choosing it. One of the major areas where AI could have a significant impact is early cancer detection . This was one of the most preferred applications of AI in cancer care by the respondents of this survey. Currently, many cancer cases are not diagnosed until the disease has advanced, making treatment more difficult and less successful. Several large-scale studies have reported that using AI to analyze lung CT images for lung cancer screening confirms survival benefits and that it helps with the precise diagnosis and treatment of liver and brain cancers . By analyzing medical images and other data, AI algorithms can help identify signs of cancer at an early stage, increasing the chances of successful treatment. For example, AI can analyze mammograms to identify breast cancer, CT scans to identify lung cancer, and polyps indicative of colorectal cancer in real-time . AI can also help with early detection by analyzing a patient’s medical history and test results to identify patterns that may indicate the presence of cancer . For example, AI algorithms can analyze genetic data to identify mutations associated with increased cancer risk. By analyzing this data, doctors can determine the most appropriate course of treatment for each patient . When asked about the likelihood of future AI applications in cancer care, 73.13% of respondents selected grading and classifying the cancer stages, which means image analysis, followed by 69.08% of respondents who thought AI would be useful to provide a more reliable diagnosis within the next ten years. For example, Watson analyzed a patient’s medical record and generated recommendations for treatment options by selecting from a list of possibilities, scoring their appropriateness for the patient on a percentage scale, and presenting them to the clinician for consideration . Another area where AI could majorly impact is the discovery of new drugs. This expectation was shared by most of the respondents in this study. By analyzing data from clinical trials and other sources, AI algorithms can identify patterns that may indicate the potential effectiveness of a new drug . This could help speed up the drug development process, potentially leading to new treatments that are more effective and have fewer side effects. When we surveyed the factors hampering the use of AI in cancer care, uncertainty about legal responsibility and accountability for AI-supported clinical decisions was the choice of most (66.67%) of the 22.82% of respondents who selected ethical or regulatory issues. To address the uncertainty, it may be necessary to establish clear guidelines and regulations around the use of AI in clinical practice, including standards for data collection, storage, and use, as well as guidelines for transparency and accountability in decision-making processes . The explainability of AI models is gaining importance in clinical practices. Transparent algorithms or explanatory approaches create trust and can make adopting AI systems less risky for clinical practitioners . Another important factor that must be considered is AI’s ethical implications in cancer care, including issues of bias. As it is known, one potential issue with the use of AI in cancer care is the risk of bias in the algorithms used . About 17% of the 22.82% of the respondents who chose ethical or regulatory issues believed algorithmic bias caused by the underrepresentation of minorities and underrated groups was the most likely factor to hamper Ai use in cancer care. If the data used to train the algorithms were not representative of the treated population, the AI may not be able to diagnose or treat certain groups of patients accurately. For example, if the data used to train an AI algorithm to detect breast cancer are predominantly from white women, the algorithm may not be as effective at detecting breast cancer in women of other races . Thus, one of the biggest challenges is the generalizability of AI algorithms. To address this issue, ensuring that the data used to train AI algorithms are diverse and representative of the population being treated is important. To create more reliable, accurate, and generalizable AI models, it is necessary to have a deeper understanding of the ethical considerations surrounding the use of AI, including how to interpret its performance, standardize techniques, and identify and address bias . To ensure that the AI is trained accurately and effectively, it is necessary to include a diverse range of individuals in terms of ethnicity, age, and sex, as well as examples of benign and malignant tumors. Additionally, when implementing precision medicine and AI in real-world clinical settings, it is important to consider environmental factors, challenges related to providing care in resource-poor areas, and multiple concurrent medical conditions . The difficulty of incorporating AI into clinical practices was another important hampering factor, as 26.06% of the respondents reported it could hamper AI use in cancer care in the Future. This would be mainly due to issues regarding the alignment of AI to the specific context of clinical practice, according to 46.78% of those respondents. One challenge in integrating AI with clinical practice is the generalizability and reproducibility of AI algorithms. This is because, in clinical practice, machine learning models may encounter real-world data that are incomplete or contains errors, despite being trained on datasets that have been carefully cleaned to eliminate poor-quality information . Many respondents who reported that the lack of standardization in cancer-related health data was the most likely factor to hamper AI use in cancer care believed this would be due to difficulties in testing, validating, certifying, and auditing AI algorithms and systems (47.20%). Still, regarding the lack of standardization, the lack or misuse of electronic health records was considered by fewer respondents (13.87%) as likely to hamper the use of AI in cancer care. With the increasing use of electronic health records, there was a risk that patient data could be accessed or shared without the patient’s knowledge or consent . Therefore, it is important to ensure appropriate safeguards to protect patient privacy and prevent data misuse. Additionally, challenges such as data breaches, ransomware attacks, and hackers have hampered the adoption process among healthcare providers .
This study presents a comprehensive evaluation of the views and expectations of 1030 experts in the field of AI applications in cancer. Through a cross-sectional, global, web-based survey, the researchers were asked about their views on the future of AI in cancer care. The results indicated that most respondents believed AI would play a critical role in cancer prediction, early detection, grading, and classification, thus improving follow-up services and providing more reliable diagnostics. Despite these benefits, incorporating AI into clinical practice may be challenging due to the lack of standardization in cancer-related health data. Specifically, these limitations may hinder the testing, validation, certification, and auditing of AI algorithms and systems. The results of this study provide valuable insights into the future trends and potential of AI in cancer care and can inform the research and development decisions of various stakeholders, including individual researchers and research funding agencies, both public and private. In conclusion, this study highlights the importance of addressing the barriers to the widespread adoption of AI in cancer care to fully realize the potential of these technologies in improving patient outcomes and reducing the burden of cancer worldwide.
|
Development of Improved DNA Collection and Extraction Methods for Handled Documents | bc60c03a-76f0-4435-8e4e-5bf83e3535eb | 10048427 | Forensic Medicine[mh] | Paper evidence may be recovered from a crime scene and is frequently encountered in forgery, aggravated harassment, and robbery investigations. Such evidence may be handwritten or typed, but likely contains DNA deposits due to handling. Individuals vary in the amount of DNA they deposit on a surface . Time since handwashing and touching period are all factors in the amount of DNA deposited; however, an individual’s shedder status may also play a role . Which finger(s) or area of the hand contact the paper may impact the quantity of DNA recovered . Regardless of which fingers touch the paper surface, brief contact with paper may transfer enough DNA to yield an interpretable DNA profile . The variable composition of touch DNA deposits, which may include nucleated cells and cell-free DNAs from facial touching, sweat, and sebaceous secretions , can be attributed to activities prior to touch deposition. Larger quantities of DNA have been recovered from touched surfaces when tested individuals were allowed to perform activities after handwashing . It has also been shown that more DNA accumulates with increasing time intervals after handwashing . Optimization of sampling techniques is critical due to the variable nature of touch DNA deposition from fingermarks. Additionally, paper evidence may require analysis by multiple disciplines including DNA, questioned documents, and latent fingerprints. For this reason, evaluating a non-destructive sample collection method and extraction method may help in preserving evidence integrity and improving DNA recovery for this difficult material. Few methods exist for non-destructive processing of paper evidence for trace DNA deposits from fingermarks. Traditional methods, including direct cutting and wet/dry swabbing of paper , destroy the paper surface and prevent additional analyses. The porous and thin nature of paper requires imaginative sampling techniques to collect DNA from the paper surface, such as dry/dry swabbing or vacuuming . While the method selection is important, equally important is the swab itself, if that is the chosen collection method. Previous work indicates that flocked swabs may improve DNA recovery from porous substrates when compared to traditional cotton swabs . The highest extraction and recovery efficiencies have been observed for polyester Alpha ® swabs and nylon flocked swabs . Fiber type and swab morphology may play a role in how DNA is collected from the surface and subsequently released in solution during purification. Additionally, the analyst must address the PCR inhibitors present in the paper, if direct cutting, or on the paper surface, if swabbing. Use of an extraction kit tailored for plants may aid in lysis of cellular components and purification of DNA from plant-based samples, like paper . Further purification may be necessary to combat bleaching agents and other components of paper that may inhibit PCR downstream. While the method of DNA recovery is important, ensuring DNA degradation does not occur prior to sample collection is of equal importance. Evidence sampling may not occur within days or months following the original evidence collection and submission to the property and evidence unit. This study compared several sampling methods, specifically, using flocked swabs to improve DNA recovery from handled documents, and tested a dry vacuum approach for single fingermarks. Traditional sampling by directly cutting from the paper was included as a benchmark. Furthermore, a stability study evaluated the stability of DNA fingermark evidence deposited on paper over 11 months.
Volunteers were recruited, and coded samples were collected in accordance with CUNY IRB#2017-0306 using an informed consent form. Prior to sample collection from volunteers, half sheets of 8.5 in × 11 in 20 lb copy paper (Staples, Framingham, MA, USA) were irradiated with ultra-violet light for 15 min on each side and stored in a clean plastic container. An irradiated half sheet of copy paper was placed on a large lint-free wipe that had been placed on a clean laboratory table covered with new bench paper. Each volunteer was instructed to wash and dry their hands, touch their face and neck for 15 s, and rub their hands together for 15 s. This preparation was modeled after the Plaza et al. “charging” protocol mimicking involuntary face touching. Once the volunteer had “charged” their hands, they were advised to press four fingers onto the half sheet of copy paper for five seconds. After deposition, the sheet was labeled with the individual’s code number and “L” or “R” with a pencil to designate handedness. The sample was then placed in a manila envelope labeled with the volunteer’s code number, date of collection, and sequence number. Only four fingers were targeted to eliminate donor hand size variability. While some volunteers would have been able to fit all five fingers on the paper half sheet, this did not work for others, which is why all volunteers were advised to press only four fingers on the paper during sample collection. For the swab/direct cutting study, six pairs of samples were collected in sequence from three volunteers. For the vacuum study, again three volunteers were recruited, and three pairs of samples were collected from each individual. For the stability study, three pairs of samples from three volunteers were collected on two separate dates, for a total of six pairs of samples for each donor. Prior to the cutting, swabbing, or vacuuming recovery samples were stored at room temperature. After processing with one of the three methods, sample tubes were stored at room temperature until extraction. 2.1. Dry/Dry Flocked Swab Technique A PurFlock Ultra polyester dry flocked swab (Puritan, Guilford, ME, USA) was rubbed on the entire surface of the half sheet of copy paper. The swab head was cut and placed into a 1.5 mL microcentrifuge tube (Promega, Madison, WI, USA). The swabbing procedure was repeated with a second polyester dry flocked swab. The swab head was added to the 1.5 mL microcentrifuge tube containing the first swab head. 2.2. Wet/Dry Flocked Swab Technique A polyester flocked swab pre-moistened with 10 µL 0.1% Triton X-100 (Sigma-Aldrich, Allentown, PA, USA) was rubbed on the entire surface of the half sheet of copy paper. The swab head was cut into a 1.5 mL microcentrifuge tube. The swabbing procedure was repeated with a second dry polyester flocked swab and placed into the same 1.5 mL microcentrifuge tube containing the pre-moistened swab head. 2.3. Direct Cutting Technique One-inch by one-inch cuttings were made from the half sheet of copy paper and placed into a 15 mL conical tube. Immediately prior to extraction, cutting samples were soaked with 800 µL of deionized water for 30 min at room temperature. The liquid portion was transferred to a QIAGEN Lyse and Spin (QIAGEN, Hilden, Germany) column-tube assembly and centrifuged for one minute at maximum speed (20,000× g ). Once all liquid passed through the column, the column was discarded and tube containing the liquid was retained for extraction. 2.4. Vacuum Swab Technique Using McLaughlin et al. method as a guide, the vacuum swab technique was tested for wet or dry cotton swabs and glass or plastic pipettes. Prior to vacuum swabbing ( ), the touched half sheet of copy paper was suspended on magnetic clips attached to a metal and wood base to prevent suctioning of the table surface while vacuuming the paper surface. The metal clips and metal sheet attached to the wood block were cleaned with DNA Away (ThermoFisher Scientific, Waltham, MA, USA), deionized water, and ethanol in between each sample. The laboratory table was cleaned and covered with bench paper. 2.4.1. Glass Pasteur Pipette A new clean Pasteur pipette was obtained, and the thin tip of the pipette was snapped off immediately prior to processing with the vacuum swab apparatus. The wood tip of a dry cotton swab (Puritan, Guilford, ME, USA) was snapped to shorten and inserted into the top of the Pasteur pipette. The cotton tip was situated just below the top. The tapered end of the altered pipette was attached to vacuum tubing connected to the vacuum of the M-Vac apparatus (M-Vac Systems, Inc., Sandy, UT, USA). The vacuum was turned on and the pipette/swab apparatus was pulled across the sample sheet from left to right. This movement was repeated until the entire surface was vacuumed, then the vacuum was turned off. The cotton tip was cut into a microcentrifuge tube and stored at room temperature until DNA extraction. Samples from three volunteers were processed in this manner. This procedure was repeated with additional samples from three volunteers using cotton swabs moistened with 10 µL of 0.1% Triton X-100 applied to the swab head. 2.4.2. Plastic Pipette Tip Prior to sample collection, 1000 µL plastic pipette tips were autoclaved. A cotton swab was shortened and inserted into the wide end of the pipette tip. The tapered end of the pipette tip was attached to vacuum tubing connected to a vacuum. Sample collection proceeded in the same manner as the glass Pasteur pipette. Samples from three volunteers were collected with dry swabs. An additional set of samples from three volunteers was collected using a swab moistened with 0.1% Triton X-100 as described in . 2.5. Extraction, Quantitation, Amplification Half of the swabbing and direct cutting samples and all of the vacuum swab samples were extracted using a protocol based on the one described by Forsberg et al. . To sample tubes, 200 µL of 5% Chelex 100 sodium form (Sigma-Aldrich, Allentown, PA, USA), 5 µL of 13.5 mg/mL Proteinase K (Promega, Madison, WI, USA) 2 µL of 10% Tween 20 (Sigma-Aldrich, Allentown, PA, USA), and 300 µL of deionized water were added. Additional deionized water was added to each tube to bring the volume up to 800 µL. Samples were incubated for 30 min at room temperature at the bench with vortexing or agitation occasionally. After 30 min, samples were incubated with a Multitherm shaker (Benchmark Scientific, Sayreville, NJ, USA) for 45 min at 56 °C shaking at 400 RPM. Next, samples were incubated at 98 °C with a Multitherm shaker set to no agitation for 10 min. After incubation, each swab substrate was transferred to a spin basket (Promega, Madison, WI, USA) placed in a Dolphin tube, and centrifuged for five minutes at 1500× g . The collected liquid was transferred to the original sample tube. The substrate and spin basket-tube assembly was discarded. The DNA extract was concentrated using Microcon DNA Fastflow (MW100) filters (Millipore Sigma Burlington, MA, USA). Three hundred microliters of the liquid were transferred to the Microcon filter unit -tube assembly and centrifuged for 30 min at 500× g , the step was repeated for any remaining liquid. After the liquid on the Microcon filter was reduced to 5 µL of liquid, 30 µL of TE −4 buffer were added and the Microcon filter was inverted into a new collection tube and centrifuged for three minutes at 1000× g . The liquid collected in the tube was transferred to a new 1.5 mL tube and stored frozen until quantitation. The remaining samples were extracted using the QIAamp DNA Investigator kit (QIAGEN, Hilden, Germany). Two and a half milliliters of TE −4 buffer were added to the direct cutting samples and incubated at room temperature for five minutes. The liquid was split into two microcentrifuge tubes and processed as described below. Twenty microliters of 20 mg/mL Proteinase K (QIAGEN, Hilden, Germany) and 400 µL Buffer ATL were added to each sample and samples were vortexed briefly. Samples were incubated at 56 °C for one hour with shaking with a Multitherm shaker at 750 RPM. Following incubation, samples were briefly centrifuged. For samples containing swab heads, the material was transferred to a spin basket-tube assembly and centrifuged at 20,000× g for two minutes. The spin basket and material were discarded and 400 µL of Buffer AL was added to each tube. Samples were briefly vortexed and incubated for 10 min at 70 °C with shaking with a Multitherm shaker at 750 RPM. Samples were briefly centrifuged, then 200 µL of 96–100% ethanol was added to each sample. Samples were briefly vortexed and centrifuged. The entire liquid for each sample was transferred to a QIAamp MinElute column and centrifuged at 6000× g for one minute. The columns were transferred to new tubes. To each sample, 500 µL of Buffer AW1 was added. Samples were centrifuged at 6000× g for one minute and then columns were transferred to new collection tubes. Seven hundred microliters of Buffer AW2 was added to each sample and centrifuged for one minute at 6000× g . Columns were transferred to new collection tubes and 700 µL of ethanol was added. Samples were centrifuged at 6000× g for one minute. All columns were transferred to new collection tubes and centrifuged for three minutes at 20,000× g to dry column membranes. Columns were transferred to new collection tubes and the lids were opened. All samples were incubated at room temperature for 10 min at the bench. After incubation, 35 µL of TE −4 buffer was added to each column, tubes were recapped, and samples were incubated for five minutes at room temperature at the bench. Samples were centrifuged at 20,000× g for one minute and the liquid collected at the bottom of the tube was stored at −20 °C until quantitation. DNA concentrations were determined by real-time PCR quantitation with the Quantifiler Trio kit on a 7500 Real-Time PCR instrument (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s specifications . Swabbing, direct cutting, and stability study samples were amplified with the Globalfiler kit (Thermo Fisher Scientific, Waltham, MA, USA) targeting an input of 0.75 ng. Amplification reactions were prepared per the manufacturer’s specifications. The appropriate amounts of sample DNA and TE −4 buffer were added to each tube to achieve 0.75 ng of DNA in the maximum input volume of 15 µL. For samples with low quantities of DNA, 15 µL of sample DNA was added to each tube. Samples were amplified with 28 cycles using validated conditions . Following amplification, samples were separated on the 3500 Genetic Analyzer instrument (Thermo Fisher Scientific, Waltham, MA, USA). DNA profiles were evaluated with Genemapper ID-X analysis software v1.6 (Thermo Fisher Scientific, Waltham, MA, USA) using a lower analytical threshold (LAT) of 30 relative fluorescence units (RFU), analytical threshold of 75 RFU, and a stochastic threshold of 350 RFU. 2.6. Stability Study Samples were collected from three volunteers as described above using half sheets of copy paper irradiated with ultraviolet light for 15 min per side. Samples were stored at room temperature. Month 0 samples were extracted within a week of initial sample collection to set a baseline for each donor. A sample from each volunteer was swabbed and extracted once a month over 11 months. A half sheet from each volunteer was dry swabbed with a flocked swab and extracted using the Chelex-Tween method described above. Samples were quantified on the 7500 instrument using the Quantifiler Trio kit. Samples were amplified with Globalfiler and run on the 3500 Genetic Analyzer instrument as described above. Statistical analyses were performed using R version 3.6.2 and R studio version 2022.12.0+353. See Morgan for more details.
A PurFlock Ultra polyester dry flocked swab (Puritan, Guilford, ME, USA) was rubbed on the entire surface of the half sheet of copy paper. The swab head was cut and placed into a 1.5 mL microcentrifuge tube (Promega, Madison, WI, USA). The swabbing procedure was repeated with a second polyester dry flocked swab. The swab head was added to the 1.5 mL microcentrifuge tube containing the first swab head.
A polyester flocked swab pre-moistened with 10 µL 0.1% Triton X-100 (Sigma-Aldrich, Allentown, PA, USA) was rubbed on the entire surface of the half sheet of copy paper. The swab head was cut into a 1.5 mL microcentrifuge tube. The swabbing procedure was repeated with a second dry polyester flocked swab and placed into the same 1.5 mL microcentrifuge tube containing the pre-moistened swab head.
One-inch by one-inch cuttings were made from the half sheet of copy paper and placed into a 15 mL conical tube. Immediately prior to extraction, cutting samples were soaked with 800 µL of deionized water for 30 min at room temperature. The liquid portion was transferred to a QIAGEN Lyse and Spin (QIAGEN, Hilden, Germany) column-tube assembly and centrifuged for one minute at maximum speed (20,000× g ). Once all liquid passed through the column, the column was discarded and tube containing the liquid was retained for extraction.
Using McLaughlin et al. method as a guide, the vacuum swab technique was tested for wet or dry cotton swabs and glass or plastic pipettes. Prior to vacuum swabbing ( ), the touched half sheet of copy paper was suspended on magnetic clips attached to a metal and wood base to prevent suctioning of the table surface while vacuuming the paper surface. The metal clips and metal sheet attached to the wood block were cleaned with DNA Away (ThermoFisher Scientific, Waltham, MA, USA), deionized water, and ethanol in between each sample. The laboratory table was cleaned and covered with bench paper. 2.4.1. Glass Pasteur Pipette A new clean Pasteur pipette was obtained, and the thin tip of the pipette was snapped off immediately prior to processing with the vacuum swab apparatus. The wood tip of a dry cotton swab (Puritan, Guilford, ME, USA) was snapped to shorten and inserted into the top of the Pasteur pipette. The cotton tip was situated just below the top. The tapered end of the altered pipette was attached to vacuum tubing connected to the vacuum of the M-Vac apparatus (M-Vac Systems, Inc., Sandy, UT, USA). The vacuum was turned on and the pipette/swab apparatus was pulled across the sample sheet from left to right. This movement was repeated until the entire surface was vacuumed, then the vacuum was turned off. The cotton tip was cut into a microcentrifuge tube and stored at room temperature until DNA extraction. Samples from three volunteers were processed in this manner. This procedure was repeated with additional samples from three volunteers using cotton swabs moistened with 10 µL of 0.1% Triton X-100 applied to the swab head. 2.4.2. Plastic Pipette Tip Prior to sample collection, 1000 µL plastic pipette tips were autoclaved. A cotton swab was shortened and inserted into the wide end of the pipette tip. The tapered end of the pipette tip was attached to vacuum tubing connected to a vacuum. Sample collection proceeded in the same manner as the glass Pasteur pipette. Samples from three volunteers were collected with dry swabs. An additional set of samples from three volunteers was collected using a swab moistened with 0.1% Triton X-100 as described in .
A new clean Pasteur pipette was obtained, and the thin tip of the pipette was snapped off immediately prior to processing with the vacuum swab apparatus. The wood tip of a dry cotton swab (Puritan, Guilford, ME, USA) was snapped to shorten and inserted into the top of the Pasteur pipette. The cotton tip was situated just below the top. The tapered end of the altered pipette was attached to vacuum tubing connected to the vacuum of the M-Vac apparatus (M-Vac Systems, Inc., Sandy, UT, USA). The vacuum was turned on and the pipette/swab apparatus was pulled across the sample sheet from left to right. This movement was repeated until the entire surface was vacuumed, then the vacuum was turned off. The cotton tip was cut into a microcentrifuge tube and stored at room temperature until DNA extraction. Samples from three volunteers were processed in this manner. This procedure was repeated with additional samples from three volunteers using cotton swabs moistened with 10 µL of 0.1% Triton X-100 applied to the swab head.
Prior to sample collection, 1000 µL plastic pipette tips were autoclaved. A cotton swab was shortened and inserted into the wide end of the pipette tip. The tapered end of the pipette tip was attached to vacuum tubing connected to a vacuum. Sample collection proceeded in the same manner as the glass Pasteur pipette. Samples from three volunteers were collected with dry swabs. An additional set of samples from three volunteers was collected using a swab moistened with 0.1% Triton X-100 as described in .
Half of the swabbing and direct cutting samples and all of the vacuum swab samples were extracted using a protocol based on the one described by Forsberg et al. . To sample tubes, 200 µL of 5% Chelex 100 sodium form (Sigma-Aldrich, Allentown, PA, USA), 5 µL of 13.5 mg/mL Proteinase K (Promega, Madison, WI, USA) 2 µL of 10% Tween 20 (Sigma-Aldrich, Allentown, PA, USA), and 300 µL of deionized water were added. Additional deionized water was added to each tube to bring the volume up to 800 µL. Samples were incubated for 30 min at room temperature at the bench with vortexing or agitation occasionally. After 30 min, samples were incubated with a Multitherm shaker (Benchmark Scientific, Sayreville, NJ, USA) for 45 min at 56 °C shaking at 400 RPM. Next, samples were incubated at 98 °C with a Multitherm shaker set to no agitation for 10 min. After incubation, each swab substrate was transferred to a spin basket (Promega, Madison, WI, USA) placed in a Dolphin tube, and centrifuged for five minutes at 1500× g . The collected liquid was transferred to the original sample tube. The substrate and spin basket-tube assembly was discarded. The DNA extract was concentrated using Microcon DNA Fastflow (MW100) filters (Millipore Sigma Burlington, MA, USA). Three hundred microliters of the liquid were transferred to the Microcon filter unit -tube assembly and centrifuged for 30 min at 500× g , the step was repeated for any remaining liquid. After the liquid on the Microcon filter was reduced to 5 µL of liquid, 30 µL of TE −4 buffer were added and the Microcon filter was inverted into a new collection tube and centrifuged for three minutes at 1000× g . The liquid collected in the tube was transferred to a new 1.5 mL tube and stored frozen until quantitation. The remaining samples were extracted using the QIAamp DNA Investigator kit (QIAGEN, Hilden, Germany). Two and a half milliliters of TE −4 buffer were added to the direct cutting samples and incubated at room temperature for five minutes. The liquid was split into two microcentrifuge tubes and processed as described below. Twenty microliters of 20 mg/mL Proteinase K (QIAGEN, Hilden, Germany) and 400 µL Buffer ATL were added to each sample and samples were vortexed briefly. Samples were incubated at 56 °C for one hour with shaking with a Multitherm shaker at 750 RPM. Following incubation, samples were briefly centrifuged. For samples containing swab heads, the material was transferred to a spin basket-tube assembly and centrifuged at 20,000× g for two minutes. The spin basket and material were discarded and 400 µL of Buffer AL was added to each tube. Samples were briefly vortexed and incubated for 10 min at 70 °C with shaking with a Multitherm shaker at 750 RPM. Samples were briefly centrifuged, then 200 µL of 96–100% ethanol was added to each sample. Samples were briefly vortexed and centrifuged. The entire liquid for each sample was transferred to a QIAamp MinElute column and centrifuged at 6000× g for one minute. The columns were transferred to new tubes. To each sample, 500 µL of Buffer AW1 was added. Samples were centrifuged at 6000× g for one minute and then columns were transferred to new collection tubes. Seven hundred microliters of Buffer AW2 was added to each sample and centrifuged for one minute at 6000× g . Columns were transferred to new collection tubes and 700 µL of ethanol was added. Samples were centrifuged at 6000× g for one minute. All columns were transferred to new collection tubes and centrifuged for three minutes at 20,000× g to dry column membranes. Columns were transferred to new collection tubes and the lids were opened. All samples were incubated at room temperature for 10 min at the bench. After incubation, 35 µL of TE −4 buffer was added to each column, tubes were recapped, and samples were incubated for five minutes at room temperature at the bench. Samples were centrifuged at 20,000× g for one minute and the liquid collected at the bottom of the tube was stored at −20 °C until quantitation. DNA concentrations were determined by real-time PCR quantitation with the Quantifiler Trio kit on a 7500 Real-Time PCR instrument (Thermo Fisher Scientific, Waltham, MA, USA) according to the manufacturer’s specifications . Swabbing, direct cutting, and stability study samples were amplified with the Globalfiler kit (Thermo Fisher Scientific, Waltham, MA, USA) targeting an input of 0.75 ng. Amplification reactions were prepared per the manufacturer’s specifications. The appropriate amounts of sample DNA and TE −4 buffer were added to each tube to achieve 0.75 ng of DNA in the maximum input volume of 15 µL. For samples with low quantities of DNA, 15 µL of sample DNA was added to each tube. Samples were amplified with 28 cycles using validated conditions . Following amplification, samples were separated on the 3500 Genetic Analyzer instrument (Thermo Fisher Scientific, Waltham, MA, USA). DNA profiles were evaluated with Genemapper ID-X analysis software v1.6 (Thermo Fisher Scientific, Waltham, MA, USA) using a lower analytical threshold (LAT) of 30 relative fluorescence units (RFU), analytical threshold of 75 RFU, and a stochastic threshold of 350 RFU.
Samples were collected from three volunteers as described above using half sheets of copy paper irradiated with ultraviolet light for 15 min per side. Samples were stored at room temperature. Month 0 samples were extracted within a week of initial sample collection to set a baseline for each donor. A sample from each volunteer was swabbed and extracted once a month over 11 months. A half sheet from each volunteer was dry swabbed with a flocked swab and extracted using the Chelex-Tween method described above. Samples were quantified on the 7500 instrument using the Quantifiler Trio kit. Samples were amplified with Globalfiler and run on the 3500 Genetic Analyzer instrument as described above. Statistical analyses were performed using R version 3.6.2 and R studio version 2022.12.0+353. See Morgan for more details.
3.1. Comparison of Sampling Technique No DNA was detected for the direct cutting sampling method for either extraction method ( ). Higher mean quantitation values were observed for samples extracted with the Chelex-Tween method compared to the QIAamp method. A multiple regression model evaluated the predictive value of extraction method and sampling method on quantitation value. No statistically significant effect (F(3,14) = 2.74, p = 0.082, R 2 = 0.370) was observed for the model. None of the individual factors were statistically significant predictors for the quantitation value. (See for multiple regression summary). All samples were amplified with the Globalfiler kit and DNA profiles were evaluated for number of peaks detected above the lower analytical threshold and analytical threshold. As expected based on the low DNA quantitation values, none of the samples yielded a full DNA profile. The number of peaks observed for all sampling conditions were compared for the Chelex-Tween extracted samples ( ). The dry/dry swabbing condition showed the largest spread in number of peaks observed and the largest mean. Only a few peaks were observed in the wet/dry swabbing condition; the mean value was less than for the dry/dry condition. Direct cutting samples did not yield any peak information. Sampling methods for QIAamp extracted samples were also compared ( ). Wet/dry swabbing performed the best of the QIAamp extracted samples. No peaks were observed above the lower analytical threshold for both the cutting and dry/dry swab conditions. A statistically significant effect (F(4,13) = 12.19, p < 0.001, R 2 = 0.789)) was observed for the multiple regression model evaluating the predictive value of sampling method, extraction method and quantitation value on the number of peaks observed above the lower analytical threshold. Quantitation value was a statistically significant predictor for the model (t = 5.38, p < 0.001). (See for multiple regression summary). 3.2. Vacuum Swab Technique No statistically significant effect (F(2,19) = 2.01, p = 0.189, R 2 = 0.309)) was observed in the multiple regression model evaluating the predictive value of pipette used for the vacuum swabbing and the swab being wet or dry on the quantitation value. (See for multiple regression summary). Low DNA recovery was observed for a small subset of glass Pasteur pipette samples that had been extracted with ForensicGEM SexCrime kit (data not included). A spread in quantitation values was observed for all conditions evaluated ( ). Vacuum swab methods with the plastic pipette tip demonstrated the largest spread in quantitation values for both the dry and wet swab conditions. The glass pipette wet and dry swab conditions exhibited lower means and smaller spreads in DNA quantitation values. Quantitation values were similar for both swab conditions using the glass pipette, with means that were close to zero. Evaluation by one-way analysis of variance (ANOVA) showed no statistically significant difference among the means (F(3) = 1.284, p = 0.344). (See for one-way ANOVA summary). 3.3. Stability Study The stability of DNA from fingermarks deposited on paper was evaluated over an 11-month period. Samples collected from three donors were extracted, quantified, and amplified once a month for 11 months. As seen in below, the amount of DNA recovered remained consistent for each donor over 11 months. The quantitation value for month 2 for donor 002C demonstrated a quantitation value of 3.2568 ng/µL, which was much larger than all other values observed for this individual. A Grubbs test was performed to determine if there was a possible outlier in the data. The resultant G statistic was G = 5.82 and p < 0.001, which indicated the large value observed for month 2, donor 002C was an outlier. Once identified, the outlier was not included in further analysis. (See for Grubbs Test summary). DNA averages varied by donor and exhibited large standard deviations, commonly seen in fingermark DNA studies. Donor 002C had larger DNA yields compared to donors 001C and 005C. A large spread in quantitation values among the three donors was observed for months five, six, and seven ( ). Months two and ten showed the smallest spread in quantitation values. Median values across all months are consistent, while the means also show some variability month to month. A one-way analysis of variance (ANOVA) was performed to determine if there was a statistically significant difference among the means. No statistically significant difference was observed F(11) = 0.34 p = 0.97 among the monthly means. (See for one-way ANOVA summary). While quantitation value may indicate success of a DNA collection and extraction method, profile quality and completeness are additional indicators of long-term stability of DNA from fingermarks on paper. The number of alleles detected, quantitation value, and the month were plotted in three dimensions to visualize the relationship among the factors ( ). There was mixed success in the amount of DNA detected and the number of alleles detected for samples each month. In general, low quantity samples yielded a low number of alleles observed. As DNA quantity increased, the number of alleles observed varied. A statistically significant effect (F(12,22) = 2.92, p = 0.014, R 2 = 0.614) was observed for a multiple regression model evaluating the predictive value of the quantitation value and month of extraction on the number of alleles observed. The quantitation value was a statistically significant individual predictor for the model (t = 5.56, p < 0.001). Month of extraction was not a statistically significant predictor for the model. (See for multiple regression summary).
No DNA was detected for the direct cutting sampling method for either extraction method ( ). Higher mean quantitation values were observed for samples extracted with the Chelex-Tween method compared to the QIAamp method. A multiple regression model evaluated the predictive value of extraction method and sampling method on quantitation value. No statistically significant effect (F(3,14) = 2.74, p = 0.082, R 2 = 0.370) was observed for the model. None of the individual factors were statistically significant predictors for the quantitation value. (See for multiple regression summary). All samples were amplified with the Globalfiler kit and DNA profiles were evaluated for number of peaks detected above the lower analytical threshold and analytical threshold. As expected based on the low DNA quantitation values, none of the samples yielded a full DNA profile. The number of peaks observed for all sampling conditions were compared for the Chelex-Tween extracted samples ( ). The dry/dry swabbing condition showed the largest spread in number of peaks observed and the largest mean. Only a few peaks were observed in the wet/dry swabbing condition; the mean value was less than for the dry/dry condition. Direct cutting samples did not yield any peak information. Sampling methods for QIAamp extracted samples were also compared ( ). Wet/dry swabbing performed the best of the QIAamp extracted samples. No peaks were observed above the lower analytical threshold for both the cutting and dry/dry swab conditions. A statistically significant effect (F(4,13) = 12.19, p < 0.001, R 2 = 0.789)) was observed for the multiple regression model evaluating the predictive value of sampling method, extraction method and quantitation value on the number of peaks observed above the lower analytical threshold. Quantitation value was a statistically significant predictor for the model (t = 5.38, p < 0.001). (See for multiple regression summary).
No statistically significant effect (F(2,19) = 2.01, p = 0.189, R 2 = 0.309)) was observed in the multiple regression model evaluating the predictive value of pipette used for the vacuum swabbing and the swab being wet or dry on the quantitation value. (See for multiple regression summary). Low DNA recovery was observed for a small subset of glass Pasteur pipette samples that had been extracted with ForensicGEM SexCrime kit (data not included). A spread in quantitation values was observed for all conditions evaluated ( ). Vacuum swab methods with the plastic pipette tip demonstrated the largest spread in quantitation values for both the dry and wet swab conditions. The glass pipette wet and dry swab conditions exhibited lower means and smaller spreads in DNA quantitation values. Quantitation values were similar for both swab conditions using the glass pipette, with means that were close to zero. Evaluation by one-way analysis of variance (ANOVA) showed no statistically significant difference among the means (F(3) = 1.284, p = 0.344). (See for one-way ANOVA summary).
The stability of DNA from fingermarks deposited on paper was evaluated over an 11-month period. Samples collected from three donors were extracted, quantified, and amplified once a month for 11 months. As seen in below, the amount of DNA recovered remained consistent for each donor over 11 months. The quantitation value for month 2 for donor 002C demonstrated a quantitation value of 3.2568 ng/µL, which was much larger than all other values observed for this individual. A Grubbs test was performed to determine if there was a possible outlier in the data. The resultant G statistic was G = 5.82 and p < 0.001, which indicated the large value observed for month 2, donor 002C was an outlier. Once identified, the outlier was not included in further analysis. (See for Grubbs Test summary). DNA averages varied by donor and exhibited large standard deviations, commonly seen in fingermark DNA studies. Donor 002C had larger DNA yields compared to donors 001C and 005C. A large spread in quantitation values among the three donors was observed for months five, six, and seven ( ). Months two and ten showed the smallest spread in quantitation values. Median values across all months are consistent, while the means also show some variability month to month. A one-way analysis of variance (ANOVA) was performed to determine if there was a statistically significant difference among the means. No statistically significant difference was observed F(11) = 0.34 p = 0.97 among the monthly means. (See for one-way ANOVA summary). While quantitation value may indicate success of a DNA collection and extraction method, profile quality and completeness are additional indicators of long-term stability of DNA from fingermarks on paper. The number of alleles detected, quantitation value, and the month were plotted in three dimensions to visualize the relationship among the factors ( ). There was mixed success in the amount of DNA detected and the number of alleles detected for samples each month. In general, low quantity samples yielded a low number of alleles observed. As DNA quantity increased, the number of alleles observed varied. A statistically significant effect (F(12,22) = 2.92, p = 0.014, R 2 = 0.614) was observed for a multiple regression model evaluating the predictive value of the quantitation value and month of extraction on the number of alleles observed. The quantitation value was a statistically significant individual predictor for the model (t = 5.56, p < 0.001). Month of extraction was not a statistically significant predictor for the model. (See for multiple regression summary).
Due to the limited sample size, handedness was not evaluated as a variable in this work. Additional consideration may be needed in a replicate study with a larger number of participants; however, no significant difference in DNA quantity was observed for right or left hands in work completed by Heftez et al. . Other factors to be considered in additional studies are how to control finger “charging”, e.g., collecting the time since last face washing , and measuring the exact pressure of DNA deposition while placing fingers. Higher deposition pressure has been shown to transfer higher quantities of DNA to the surface of the substrate . 4.1. Direct Cutting and Swabbing Study No statistically significant effect was observed for the extraction method and sampling method on the quantitation value obtained for the cutting/swabbing study. Similar results were observed by Parsons et al. , in which no significant difference in DNA quantity was observed for the wet/dry and dry/dry swabbing conditions. This result supports implementation of either sampling method; however, in a multi-discipline laboratory where other sections may analyze the evidence, the dry/dry sampling technique may be preferred. Unlike Balogh et al. , no DNA profiles were obtained from paper cutting samples; however, the sample preparation method differed from this study and may partially explain the discrepancy. Sewell et al. did observe low quantities of DNA from office paper cuttings which is consistent with the results observed in this study. However, Sewell et al. observed a statistically significant increase in DNA recovery from paper cuttings extracted with the DNeasy plant mini kit compared to the QIAamp mini kit. Their result suggests the kit chemistry may improve sample purification and removal of PCR inhibitors from direct cuttings. In the future, comparison of the Chelex-Tween method with the DNeasy method may be worthwhile for the direct cutting sampling method; however, other commercially available extraction kits could improve DNA yields in swab samples from paper surfaces. The quantitation value was a statistically significant predictor for the number of peaks observed above the lower analytical threshold. As expected, a higher quantitation value predicts an increased number of peaks observed in a DNA profile. A similar result was also observed in the stability study, the quantitation value was a statistically significant predictor for the number of alleles detected in the DNA profile. 4.2. Vacuum Swab Study DNA recovery for both pipette materials was low and the multiple linear regression model exhibited no statistically significant effect for pipette material or swabbing method on the quantitation value when evaluated together. Similar to the direct cutting and swabbing study, none of the individual variables were statistically significant predictors. Overall, the vacuum swab samples had slightly higher DNA yields than the cutting and swabbing samples. This could mean the method is more efficient but may also be due to the difference in DNA donors for the samples. Additional comparison work with an identical pool of donors is recommended to confirm. A large spread in quantitation values for the 1000 µL plastic pipette tip vacuum swab samples may be due to the experimenter observed difference in suction strength during sample collection. The plastic pipette tip had stronger suction while pulling across the sheet of copy paper and may have collected more DNA from the surface compared to the glass pipette. A disadvantage of the plastic pipette tip is that the stronger suction made it harder to manipulate the swab-pipette tip apparatus while vacuuming the paper surface. To combat the strong suction, it may be worth testing the vacuum collection method with a filtered pipette tip and removing the filter for lysis and purification as recently described for fabric . Use of the filter rather than a swab head may decrease the strength of suction as it is dragged across the document. The vacuum DNA recovery here was also low compared to the work by McLaughlin et al. , but both studies demonstrated the utility of the vacuum swab technique. The lower yield can be explained by a difference in study design with only half sheets of paper and four individual prints here and a full letter size sheet with handwritten text in the other study. A study on mock casework handwritten documents showed a 67% success rate of interpretable STR profiles in copy paper and a 92% success rate for notepad paper . Further modifications to the method should be explored. Bruijns et al. noted a maximum extraction efficiency with nylon flocked swabs at approximately 48%, while the cotton swab performance was approximately 21%. Substitution of the cotton swab for a nylon flocked swab for the vacuum method may improve DNA recovery in subsequent studies. An alternative may be using inverted filter tips, but not all filter tips may show the same DNA recovery . 4.3. Stability Study One of the three donors for the stability study demonstrated consistently large quantities of DNA from the paper samples, which is consistent with the literature regarding shedder status and variability in DNA deposition . While differences in donors were observed, there was no statistically significant difference among the monthly means, which suggests that DNA from fingermarks remains stable through 11 months of storage at room temperature. This result is in line with previous work by Wurmbach and Ostojic evaluating the effect of storage time on STR typing success. No significant difference in STR success was observed.
No statistically significant effect was observed for the extraction method and sampling method on the quantitation value obtained for the cutting/swabbing study. Similar results were observed by Parsons et al. , in which no significant difference in DNA quantity was observed for the wet/dry and dry/dry swabbing conditions. This result supports implementation of either sampling method; however, in a multi-discipline laboratory where other sections may analyze the evidence, the dry/dry sampling technique may be preferred. Unlike Balogh et al. , no DNA profiles were obtained from paper cutting samples; however, the sample preparation method differed from this study and may partially explain the discrepancy. Sewell et al. did observe low quantities of DNA from office paper cuttings which is consistent with the results observed in this study. However, Sewell et al. observed a statistically significant increase in DNA recovery from paper cuttings extracted with the DNeasy plant mini kit compared to the QIAamp mini kit. Their result suggests the kit chemistry may improve sample purification and removal of PCR inhibitors from direct cuttings. In the future, comparison of the Chelex-Tween method with the DNeasy method may be worthwhile for the direct cutting sampling method; however, other commercially available extraction kits could improve DNA yields in swab samples from paper surfaces. The quantitation value was a statistically significant predictor for the number of peaks observed above the lower analytical threshold. As expected, a higher quantitation value predicts an increased number of peaks observed in a DNA profile. A similar result was also observed in the stability study, the quantitation value was a statistically significant predictor for the number of alleles detected in the DNA profile.
DNA recovery for both pipette materials was low and the multiple linear regression model exhibited no statistically significant effect for pipette material or swabbing method on the quantitation value when evaluated together. Similar to the direct cutting and swabbing study, none of the individual variables were statistically significant predictors. Overall, the vacuum swab samples had slightly higher DNA yields than the cutting and swabbing samples. This could mean the method is more efficient but may also be due to the difference in DNA donors for the samples. Additional comparison work with an identical pool of donors is recommended to confirm. A large spread in quantitation values for the 1000 µL plastic pipette tip vacuum swab samples may be due to the experimenter observed difference in suction strength during sample collection. The plastic pipette tip had stronger suction while pulling across the sheet of copy paper and may have collected more DNA from the surface compared to the glass pipette. A disadvantage of the plastic pipette tip is that the stronger suction made it harder to manipulate the swab-pipette tip apparatus while vacuuming the paper surface. To combat the strong suction, it may be worth testing the vacuum collection method with a filtered pipette tip and removing the filter for lysis and purification as recently described for fabric . Use of the filter rather than a swab head may decrease the strength of suction as it is dragged across the document. The vacuum DNA recovery here was also low compared to the work by McLaughlin et al. , but both studies demonstrated the utility of the vacuum swab technique. The lower yield can be explained by a difference in study design with only half sheets of paper and four individual prints here and a full letter size sheet with handwritten text in the other study. A study on mock casework handwritten documents showed a 67% success rate of interpretable STR profiles in copy paper and a 92% success rate for notepad paper . Further modifications to the method should be explored. Bruijns et al. noted a maximum extraction efficiency with nylon flocked swabs at approximately 48%, while the cotton swab performance was approximately 21%. Substitution of the cotton swab for a nylon flocked swab for the vacuum method may improve DNA recovery in subsequent studies. An alternative may be using inverted filter tips, but not all filter tips may show the same DNA recovery .
One of the three donors for the stability study demonstrated consistently large quantities of DNA from the paper samples, which is consistent with the literature regarding shedder status and variability in DNA deposition . While differences in donors were observed, there was no statistically significant difference among the monthly means, which suggests that DNA from fingermarks remains stable through 11 months of storage at room temperature. This result is in line with previous work by Wurmbach and Ostojic evaluating the effect of storage time on STR typing success. No significant difference in STR success was observed.
This work demonstrates the potential for paper evidence processing using the dry vacuum technique in conjunction with the Chelex-Tween extraction method . The materials for both the dry vacuum technique and Chelex-Tween extraction are inexpensive and the methods are simple to perform. For crime laboratories that do not currently process paper evidence, implementation of this technique requires minimal low-cost equipment and materials for internal validation and training of personnel. From a workflow point of view the dry vacuum method has the advantage that it preserves other evidence types such as indented writing and fingermarks . As this study has shown, vacuuming is more effective than other DNA recovery methods, which means it can be used to collect DNA prior to any other examinations thus avoiding DNA loss and DNA contamination. The authors acknowledge the small sample sizes for each study and suggest additional work with a larger sample size to confirm findings presented in this paper. Replicate studies for the sampling method comparison and the stability study should include improved sample preparation, a larger pool of volunteers, standardized pressure during fingermark deposition, and use of “realistic” mock evidence samples to provide additional insight.
|
Validation of Probabilistic Genotyping Software for Single Cell STR Analysis | 6b2b376e-90d6-4832-a686-e85797f51890 | 10048617 | Forensic Medicine[mh] | The application of single/few cell analysis to mixture deconvolution is a growing field in forensic DNA analysis. By analyzing many individual or few cell subsamples collected from a bulk complex DNA mixture, an increase in probative DNA information is often achieved compared to standard DNA mixture approaches (i.e., the bulk homogenization and extraction of a cellular stain followed by PCR amplification, capillary electrophoresis (CE), and probabilistic genotyping (PG)) [ , , , , , , , , , , , , ]. In the event of single cell analysis, high quality single source DNA profiles are often obtainable allowing for a significant decrease in the complexity of mixture analysis, since a single recovered cell originates from only one of the several individuals comprising the mixture. To further improve the DNA quantity and, more often than not, the profile quality, multiple cells can be collected within an individual subsample. This improves the chances of recovering a full DNA profile if all of the collected cells originate from the same donor, or a simplified, reduced complexity mixed DNA profile (referred to as a “mini-mixture”) if the cells originated from multiple donors. Even in the instances of these mini-mixtures, profile complexity is often decreased by artificially altering the number of contributors (NOC) or the donor weight ratios compared to the standard bulk mixture. PG analysis has primarily been applied to bulk DNA extracts. However, these applications do not always result in probative information about the identity of all donors to said mixtures/extracts . This is primarily due to the mixture complexity caused by overlapping alleles, the presence of artifacts, and minor donors [ , , , , , , ]. One way of reducing mixture complexity is to perform direct single cell subsampling of the bulk mixture prior to genotyping and interpretation. The analysis of low template DNA samples, including from single or few cells, has also benefited from the application of PG methods. As single cell methodologies have continued to develop, PG applications have now been applied to single or few cell subsamples [ , , ]. With the application of PG, multiple cell subsamples originating from the same donor can be combined into a single analysis using the software replicate analysis function, often resulting in full profile donor information (i.e., likelihood ratios (LRs) equaling the inverse of the random match probability (RMP) of the donor reference profile) . In the present work, two PG systems that use different models, STRmix TM and EuroForMix (EFM), were validated for single or few cell applications. The validation process was comprised of two phases, the empirical determination of analytically derived parameters that often confound forensic DNA mixture analysis, and then the testing of said parameters to analyze known samples.
The direct single cell subsampling (DSCS) methodology used in this study has been previously reported in various publications [ , , , , , , ] with a detailed on-line video demonstrating the collection procedure provided in reference . 2.1. Cell Suspensions/Mixture Samples/Gel-Film Slide Creation Donor buccal swabs (freshly collected according to procedures approved by the Institutional Review Board for the University of Central Florida or dried and frozen) were used to create cell suspensions by agitating individual swabs in 300 μL of TE−4 (or water/1× PBS depending upon the experiment). The resulting solutions were centrifuged at 300 RCF for 7 min resulting in an epithelial cell pellet. The supernatant was discarded and 300 μL of TE−4 (or water/1× PBS) was used to resuspend each pellet. The Countess™ II FL (ThermoFisher Scientific, Carlsbad, CA, USA) automated cell counter was then used to determine the cell concentration of each cell suspension according to the manufacturer’s recommended protocols. Appropriate volumes of individual donor cell suspensions were mixed to create various mixture samples with the desired donor ratios (i.e., 1:1 2-person mixture, 1:1:1 3-person mixture, 1:1:1:1 4-person mixture, 1:1:1:1:1 5-person mixture, and 1:1:1:1:1:1 6-person mixtures). Samples were stored frozen (4 °C). Sixty microliters of each mixture was then pipetted onto Gel Film ® microscope slides and spread out with a sterile swab. Each slide was stained 1–2 min with Trypan Blue, gently rinsed with nuclease-free water, and air-dried overnight. The Gel Film ® microscope slides were created prior to mixture deposition by affixing Gel-Pak ® Gel-Film ® (WF, ×8 retention level, Hayward, CA, USA) to clean glass microscope slides via the adhesive backing and the clear protective covering was removed. 2.2. 3M TM Adhesive Slide Creation Prior to cell collection, an adhesive slide reservoir was created by adhering 3M™ adhesive (Allied Electronics, Fort Worth, TX, USA) to a clean glass microscope slide with double-sided tape. The adhesive back was then removed, and the slide stored in a desiccator until needed . 2.3. Cell Recovery Sample slides were visualized using a Leica M205C stereomicroscope (190–240× magnification) and cells collected directly into 1 μL Prep-n-Go Buffer (ThermoFisher Scientific, Carlsbad, CA, USA) in sterile 0.2 mL PCR flat-cap tubes via a tungsten needle and 3M TM water soluble adhesive. More specifically, the tungsten needle was utilized to obtain a small ball of 3M™ adhesive from the created adhesive slide reservoir. The adhesive-tipped needle was then used to adhere selected cells from the sample slide . The needle was then inserted into the lysis containing amplification tube until the 3M™ adhesive was observed to solubilize . 2.4. Direct Lysis/Autosomal Short Tandem Repeat (STR) Amplification of Cells One to five-cell subsamples were collected directly in 1 μL of Prep-n-Go Buffer (ThermoFisher Scientific, Carlsbad, CA, USA) and incubated at 90 °C @ 20 min; 25 °C @ 15 min. The GlobalFiler TM Express amplification kit (ThermoFisher Scientific, Carlsbad, CA, USA) was then used to amplify the cell subsamples after lysis. The amplification reaction mix consisted of 2 μL master mix and 2 μL primer mix added directly to the 0.2 mL flat cap amp tubes containing the lysed cells. A protocol of 95 °C @ 1 min; 32 cycles: 94 °C @ 3 s, 60 °C @ 30 s; 60 °C @ 8 min; 4 °C @ hold was used. This change to 32 cycles is an increase compared to the 29 cycles used with standard analysis. For positive control samples, 1 μL of a 31.25 pg/μL dilution of 007 control DNA was used. Positive control samples (typically 3–4) and negative control samples (0-cell collections and amplification blanks) were also included in every batch. 2.5. Reference Samples/Standard Bulk Mixtures According to manufacturer recommended protocols, reference profiles and standard bulk mixture samples were extracted using the QIAamp DNA Ivestigator Kit (QIAGEN, Germantown, MD, USA) and quantified using Quantifiler ® Duo DNA Quantification kit (ThermoFisher Scientific, Carlsbad, CA, USA) on the Applied Biosystems’ 7500 real-time PCR instrument (ThermoFisher Scientific, Carlsbad, CA, USA). One nanogram of DNA extracts were amplified using the GlobalFiler TM amplification kit (ThermoFisher Scientific, Carlsbad, CA, USA) at 29 PCR cycles. Probabilistic genotyping of donor references profiles and complex equimolar 2-6 person mixtures was conducted using STRmix™ v2.8 and EuroForMix v3.1.0 (Quantitative LR MLE based) according to the known NOC. 2.6. PCR Product Detection One microliter of amplified product was added to a master mix consisting of 9.5 μL Hi-Di formamide (ThermoFisher Scientific, Carlsbad, CA, USA) and 0.5 μL GeneScan™ 600 LIZ™ size standard (ThermoFisher Scientific, Carlsbad, CA, USA). Module J6 (15 s injection, 1.2 kV, 60 °C) was used to inject samples on the 3500 Genetic Analyzer (ThermoFisher Scientific, Carlsbad, CA, USA) using POP-4 TM polymer (ThermoFisher Scientific, Carlsbad, CA, USA). GeneMapper TM ID-X v1.6 (ThermoFisher Scientific, Carlsbad, CA, USA) software was used to analyze samples. Analytical thresholds of Blue: 53, Green: 86, Yellow: 46, Red: 63, and Purple: 63 were determined using Equation (1), analyzing 30 negative control subsamples (i.e., thirty 0-cell subsamples as well as amplification blanks). No stutter filtering was applied within the GeneMapper TM software. However, half-back and double back stutter was manually removed from samples prior to analysis with EuroForMix due to the software only modeling forward and reverse stutter. (1) A T = 2 ( h i g h e s t p e a k − l o w e s t t r o u g h ) 2.7. Optimal Cell Suspension Medium Single source cell suspensions were created according to using three different suspension mediums 1× PBS, nuclease-free water, and TE −4 . One-cell and two-cell subsamples (5×) were collected from each cell suspension type. Box plots were created comparing the allele recovery obtained to each cell number and suspension medium. 2.8. Stutter and Drop-In Parameters STRmix TM specific stutter files were created utilizing 75 subsamples (ranging from 1–5 cells) from 5 donors. For STRmix™, stutter regression files were created according to the implementation and validation guide in which the stutter ratio (SR) was calculated per allele and stutter type using Equation (2). (2) S R = s t u t t e r h e i g h t a l l e l e h e i g h t The drop-in rate utilized with both systems (0.0164) was determined utilizing thirty-five negative controls (i.e., 0-cell subsamples or amplification blanks). 2.9. DSCS Probabilistic Genotyping (PG) Probabilistic Genotyping Software STRmix™ v2.8 and EuroForMix v3.1.0 (Quantitative LR MLE based) was validated for use with 1–5 cell subsamples. One-cell samples were analyzed as single source and the log(LR) was reported. Two-cell subsamples were analyzed as either single source or as a two-person mixture depending upon the NOC of the sample, except for instances in which the impact of over/underestimating NOC were examined. When EuroForMix was utilized, all subsamples were modeled with degradation, forward, and reverse stutter unless the model failed, in which case the sample was modeled without degradation. The degradation max parameter in STRmix TM was set to 0.1 (as suggested by STRmix TM support), an increase from the default 0.01 setting used for standard DNA mixture parameters. Population frequencies from the FBI extended Caucasian database and a theta value of 0.01 were used with both systems. With EuroForMix, a drop-in lambda value of 0.01 was utilized as well as 3 optimizations. STRmix TM tests were conducted to examine the impact of various burn-in accepts (i.e., 5000, 50,000, and 500,000) on LR precision of a single source 3-cell subsample. The STRmix TM Model Maker function was utilized to determine various stutter, amplification, and allele distributions using 161 × −1, −2, −3, −4, or 5-cell subsamples. Subsequent testing was conducted with both systems to determine adequate parameter optimization including tests comparing subsample LRs calculated by hand, and each PG system, and the LR impact of drop-in, inhibition, and degradation. Sensitivity and specificity were tested utilizing 110 × 1 cell subsamples, 81 × 2-cell single source subsamples, 50 × 2-cell mini mixture known donor profiles with a N = 1000 person known non-contributor database provided by STRmix™ support (from their training courses). Additionally, the impact of under (N − 1) and overestimating (N + 1) subsample NOC was examined (n = 35 single source subsamples).
Donor buccal swabs (freshly collected according to procedures approved by the Institutional Review Board for the University of Central Florida or dried and frozen) were used to create cell suspensions by agitating individual swabs in 300 μL of TE−4 (or water/1× PBS depending upon the experiment). The resulting solutions were centrifuged at 300 RCF for 7 min resulting in an epithelial cell pellet. The supernatant was discarded and 300 μL of TE−4 (or water/1× PBS) was used to resuspend each pellet. The Countess™ II FL (ThermoFisher Scientific, Carlsbad, CA, USA) automated cell counter was then used to determine the cell concentration of each cell suspension according to the manufacturer’s recommended protocols. Appropriate volumes of individual donor cell suspensions were mixed to create various mixture samples with the desired donor ratios (i.e., 1:1 2-person mixture, 1:1:1 3-person mixture, 1:1:1:1 4-person mixture, 1:1:1:1:1 5-person mixture, and 1:1:1:1:1:1 6-person mixtures). Samples were stored frozen (4 °C). Sixty microliters of each mixture was then pipetted onto Gel Film ® microscope slides and spread out with a sterile swab. Each slide was stained 1–2 min with Trypan Blue, gently rinsed with nuclease-free water, and air-dried overnight. The Gel Film ® microscope slides were created prior to mixture deposition by affixing Gel-Pak ® Gel-Film ® (WF, ×8 retention level, Hayward, CA, USA) to clean glass microscope slides via the adhesive backing and the clear protective covering was removed.
TM Adhesive Slide Creation Prior to cell collection, an adhesive slide reservoir was created by adhering 3M™ adhesive (Allied Electronics, Fort Worth, TX, USA) to a clean glass microscope slide with double-sided tape. The adhesive back was then removed, and the slide stored in a desiccator until needed .
Sample slides were visualized using a Leica M205C stereomicroscope (190–240× magnification) and cells collected directly into 1 μL Prep-n-Go Buffer (ThermoFisher Scientific, Carlsbad, CA, USA) in sterile 0.2 mL PCR flat-cap tubes via a tungsten needle and 3M TM water soluble adhesive. More specifically, the tungsten needle was utilized to obtain a small ball of 3M™ adhesive from the created adhesive slide reservoir. The adhesive-tipped needle was then used to adhere selected cells from the sample slide . The needle was then inserted into the lysis containing amplification tube until the 3M™ adhesive was observed to solubilize .
One to five-cell subsamples were collected directly in 1 μL of Prep-n-Go Buffer (ThermoFisher Scientific, Carlsbad, CA, USA) and incubated at 90 °C @ 20 min; 25 °C @ 15 min. The GlobalFiler TM Express amplification kit (ThermoFisher Scientific, Carlsbad, CA, USA) was then used to amplify the cell subsamples after lysis. The amplification reaction mix consisted of 2 μL master mix and 2 μL primer mix added directly to the 0.2 mL flat cap amp tubes containing the lysed cells. A protocol of 95 °C @ 1 min; 32 cycles: 94 °C @ 3 s, 60 °C @ 30 s; 60 °C @ 8 min; 4 °C @ hold was used. This change to 32 cycles is an increase compared to the 29 cycles used with standard analysis. For positive control samples, 1 μL of a 31.25 pg/μL dilution of 007 control DNA was used. Positive control samples (typically 3–4) and negative control samples (0-cell collections and amplification blanks) were also included in every batch.
According to manufacturer recommended protocols, reference profiles and standard bulk mixture samples were extracted using the QIAamp DNA Ivestigator Kit (QIAGEN, Germantown, MD, USA) and quantified using Quantifiler ® Duo DNA Quantification kit (ThermoFisher Scientific, Carlsbad, CA, USA) on the Applied Biosystems’ 7500 real-time PCR instrument (ThermoFisher Scientific, Carlsbad, CA, USA). One nanogram of DNA extracts were amplified using the GlobalFiler TM amplification kit (ThermoFisher Scientific, Carlsbad, CA, USA) at 29 PCR cycles. Probabilistic genotyping of donor references profiles and complex equimolar 2-6 person mixtures was conducted using STRmix™ v2.8 and EuroForMix v3.1.0 (Quantitative LR MLE based) according to the known NOC.
One microliter of amplified product was added to a master mix consisting of 9.5 μL Hi-Di formamide (ThermoFisher Scientific, Carlsbad, CA, USA) and 0.5 μL GeneScan™ 600 LIZ™ size standard (ThermoFisher Scientific, Carlsbad, CA, USA). Module J6 (15 s injection, 1.2 kV, 60 °C) was used to inject samples on the 3500 Genetic Analyzer (ThermoFisher Scientific, Carlsbad, CA, USA) using POP-4 TM polymer (ThermoFisher Scientific, Carlsbad, CA, USA). GeneMapper TM ID-X v1.6 (ThermoFisher Scientific, Carlsbad, CA, USA) software was used to analyze samples. Analytical thresholds of Blue: 53, Green: 86, Yellow: 46, Red: 63, and Purple: 63 were determined using Equation (1), analyzing 30 negative control subsamples (i.e., thirty 0-cell subsamples as well as amplification blanks). No stutter filtering was applied within the GeneMapper TM software. However, half-back and double back stutter was manually removed from samples prior to analysis with EuroForMix due to the software only modeling forward and reverse stutter. (1) A T = 2 ( h i g h e s t p e a k − l o w e s t t r o u g h )
Single source cell suspensions were created according to using three different suspension mediums 1× PBS, nuclease-free water, and TE −4 . One-cell and two-cell subsamples (5×) were collected from each cell suspension type. Box plots were created comparing the allele recovery obtained to each cell number and suspension medium.
STRmix TM specific stutter files were created utilizing 75 subsamples (ranging from 1–5 cells) from 5 donors. For STRmix™, stutter regression files were created according to the implementation and validation guide in which the stutter ratio (SR) was calculated per allele and stutter type using Equation (2). (2) S R = s t u t t e r h e i g h t a l l e l e h e i g h t The drop-in rate utilized with both systems (0.0164) was determined utilizing thirty-five negative controls (i.e., 0-cell subsamples or amplification blanks).
Probabilistic Genotyping Software STRmix™ v2.8 and EuroForMix v3.1.0 (Quantitative LR MLE based) was validated for use with 1–5 cell subsamples. One-cell samples were analyzed as single source and the log(LR) was reported. Two-cell subsamples were analyzed as either single source or as a two-person mixture depending upon the NOC of the sample, except for instances in which the impact of over/underestimating NOC were examined. When EuroForMix was utilized, all subsamples were modeled with degradation, forward, and reverse stutter unless the model failed, in which case the sample was modeled without degradation. The degradation max parameter in STRmix TM was set to 0.1 (as suggested by STRmix TM support), an increase from the default 0.01 setting used for standard DNA mixture parameters. Population frequencies from the FBI extended Caucasian database and a theta value of 0.01 were used with both systems. With EuroForMix, a drop-in lambda value of 0.01 was utilized as well as 3 optimizations. STRmix TM tests were conducted to examine the impact of various burn-in accepts (i.e., 5000, 50,000, and 500,000) on LR precision of a single source 3-cell subsample. The STRmix TM Model Maker function was utilized to determine various stutter, amplification, and allele distributions using 161 × −1, −2, −3, −4, or 5-cell subsamples. Subsequent testing was conducted with both systems to determine adequate parameter optimization including tests comparing subsample LRs calculated by hand, and each PG system, and the LR impact of drop-in, inhibition, and degradation. Sensitivity and specificity were tested utilizing 110 × 1 cell subsamples, 81 × 2-cell single source subsamples, 50 × 2-cell mini mixture known donor profiles with a N = 1000 person known non-contributor database provided by STRmix™ support (from their training courses). Additionally, the impact of under (N − 1) and overestimating (N + 1) subsample NOC was examined (n = 35 single source subsamples).
3.1. Creating Cell-Suspensions/Mixtures Analysis of single cells was posited as an approach to deconvolute complex DNA mixtures by obtaining single source DNA profiles from all donors [ , , , , , , , , , , , , ]. To test the applicability of such an approach which we term DSCS (direct single cell subsampling) , the creation of known composition complex DNA mixtures was required. One such way to accomplish this is to create individual donor cell suspensions; then, the appropriate concentration of each of the donor’s cells can be combined to create defined mixture ratios. Because the structural integrity of individual cells will depend upon their surrounding environment, different suspension mediums were tested including 1× PBS, water, and TE −4 . PBS buffer, which although routinely emphasized as a requirement for single cell analysis had a detrimental effect on the resulting DNA profiles. This may be due to the fact that cells analyzed in forensic applications are already dehydrated/lysed and therefore PBS buffer, normally required to maintain a balanced isotonic state between the inside and outside of living, viable cells, may be redundant as well as lacking nuclease protection components. TE −4 was therefore chosen as the cell suspension medium of choice due to improved allele recovery compared to 1× PBS and nuclease-free water ( ). This finding is not unsurprising as TE −4 does contain nuclease-inhibiting activity. Additionally, aged samples in the form of dried saliva stains and dried stains originating from TE −4 cell suspensions were compared in which relatively comparable allele recovery was obtained from both sample types. Although a slight decrease in allele recovery was seen from the cell suspension stains, cell suspensions were concluded to be a viable convenient alternative to saliva to use as a means of preparing defined composition mixtures ( ). 3.2. Stutter Files STRmix TM requires optimized stutter parameters for individual loci and allows for modeling of any stutter type. EuroForMix, on the other hand, only models forward and reverse stutter and models all stutter with the same distribution as opposed to per loci and does not require optimization on the part of the user. A graph of SR vs. allele number per loci and stutter type was created ( ). provide the slope and intercept (linear stutter) or the average (non-linear stutter) for each locus and stutter type. Due to the presence of elevated stutter, R 2 values obtained for single cell subsamples are lower than typically observed with standard analysis and high maximum allowable stutter percentages were utilized (back stutter 0.7, forward stutter 0.7, double back stutter 0.3, half-back stutter 0.5, half-forward stutter 0.15, −1.5 repeat stutter 0.15). The stutter distribution parameters were further determined utilizing the STRmix TM Model Maker function. 3.3. Saturation Threshold The saturation threshold for the 3500 Genetic Analyzer was determined (30,000 RFUs) by comparing the expected allele height (Ea) and the observed allele height (Oa) from single source samples. The point at which (Ea) and (Oa) begin to consistently diverge is the saturation limit of the instrument. Therefore, any profiles with allele heights > 30,000 RFUs should be diluted and capillary electrophoresis reconducted to prevent the software from incorrectly estimating the Ea in DNA profiles. Because cell subsamples are low template in nature, stochastic effects such as elevated stutter are commonly seen. Therefore, the instrument saturation limit was determined using standard DNA samples (i.e., 25 μL rxn volume amplification of 62.5 pg to 1 ng DNA extracts n = 38) ( A). Loci with simple repeat structure were utilized (D16S539, CSF1PO, TPOX, D5S818, D13S317, D7S820, and D10S1248). The expected allele heights are determined using the back stutter regression lines determined when creating stutter files for each loci and Equation (3). (3) E a = s t u t t e r p e a k h e i g h t s t u t t e r r a t i o For example, by using the DSCS stutter regression line for D16S539 (𝑆𝑅 = (0.0177 × 𝐴𝑙𝑙𝑒𝑙𝑒 𝑛𝑢𝑚𝑏𝑒𝑟) − 0.1249) ( ), the stutter ratio expected for allele 11 at D16S539 is determined to be SR= 0.0698. However, due to the aforementioned elevated stutter exhibited with DSCS subsamples (n = 189), divergence in the Ea and Oa for our DSCS subsamples is occasionally still seen even below the 30,000 RFUs threshold ( B). 3.4. Drop-In Drop-in can be described as 1–2 low-level non-reproducible peaks present within a DNA profile . However, because of the low template nature of single cell analysis, drop-in peaks with DSCS samples may have higher peak heights than seen with standard analysis methods. Therefore, a drop-in cap of 30,000 RFUs was used with STRmix TM to allow for any height allele to be considered as possible drop-in. However, re-evaluation of all single-cell subsamples analyzed within this DSCS PG validation as well as subsequent experiments [ , , ] indicated the highest drop-in peak observed was around 4800 RFU, indicating a drop-in cap of 5000 RFU may be more appropriate and may prevent the software from falsely identifying true allele peaks as potential drop-in. To determine the rate of drop-in occurrences, thirty-five negative controls (i.e., zero-cell subsamples or amplification blanks) were analyzed for the presence of drop-in alleles using Equation (4). (4) D r o p i n r a t e = n u m b e r o f d r o p i n e v e n t s n u m b e r o f l o c i s c o r e d × n u m b e r o f s a m p l e s The drop-in data did not fit a gamma distribution and limited drop-in data were available. Therefore, a uniform distribution was utilized with STRmix TM . The drop-in rate was determined to be 0.0164 ( ). The same drop-in rate was used for EuroForMix analysis with the default lambda of 0.01. 3.5. Model Maker Results Utilization of the STRmix TM software for sample analysis requires creation of a specific protocol-based statistical model that includes the determination of various empirically determined parameter variance probability distribution plots (i.e., stutter, allele, and locus specific amplification efficiency (LSAE)). This requires utilization of the Model Maker function in which single source subsamples are used. It is important for samples utilized within Model Maker to be consistent with the type and quality of samples expected to be encountered in case work. If pristine quality samples were used, then the system may not be able to account for the extreme heterozygous peak height imbalance that is commonly seen with single cell analysis due to the low template nature of the samples ( ). Because of this, 161 × −1, −2, −3, −4, or 5-cell subsamples comprised of fresh buccal cells as well as cells collected from cell suspension mixtures were utilized. The EFM software does not utilize a Model Maker type function. Model Maker performance DSCS versus standard bulk analysis was compared by plotting the log(Hb) vs. Average Peak Height of single source bulk or single cell samples ( ) with the dashed red lines indicating the 95% bounds (+/− 2 standard deviation of the mean) calculated using Equation (5). (5) ± 2 × 1.96 C 2 A P H C 2 = the 50th percentile of the allelic peak height variance Utilizing a normal distribution, 87.6% of the single source cell data fell within the 95% bounds. Ideally ≥95% of the data would have been encapsulated within the 95% bounds. A comparison of A and B shows that the 95% bounds for the DSCS samples are much broader than standard analysis parameters to account for the low template and stochastic effects seen with single cells (i.e., the extreme peak height imbalance). 3.6. Post Burn-In Accepts With STRmix TM , various post burn-in accepts (5000, 50,000, and 500,000) were tested 5× with a single source three-cell subsample ( ). As the number of burn-in accepts increases, so does the time required for analysis as well as the precision of the LR results. STRmix TM default setting is 50,000 burn-in accepts. However, when analyzing single cell subsamples, the most consistent log (LR) results occurred at 500,000 post burn-in accepts. 3.7. PG Test of Optimized Parameters To test the accuracy of the optimized PG parameters, a three-cell subsample (“S5-3C-1”) was analyzed with both PG software systems and the log(LR) was reported. The sample was then artificially degraded by decreasing the peak heights of high molecular weight alleles by 80% and low molecular weight alleles by 5%. An inhibited version of the sample in which the alleles at loci D22S1045, D21S11, D13S317, and D2S1338 were artificially inhibited by decreasing their peak heights by 40% was analyzed. The sample was also tested in which a drop-in eight allele with 15,000 RFUs was artificially added at loci D5S818. With EFM, the same log(LR) results were obtained for all four profiles (i.e., original, with drop in, inhibited, degraded) (EFM ). A figure showing a single dye channel of these artificially created samples was previously published in along with the STRmix TM results demonstrating the sample log(LR) for each of the four sample state scenarios (i.e., log(LR) = 27). With STRmix TM , the LSAE and APH for the subsample S5-3C-1 are shown in according to increasing molecular weight of the STR loci, indicating that similar trends are seen with the two different parameters. EFM, on the other hand, does not allow for the possibility of different loci having different amplification efficiencies. A comparison of the LSAE obtained for the regular sample and the inhibited sample is then shown in where a decrease in the amplification efficiency trend is obtained for the artificially inhibited loci. The degradation curves obtained from STRmix TM , are provided in for the regular sample, degraded sample, and inhibited sample, indicating that only the artificially degraded sample exhibited a characteristic degradation curve that indicated the presence of degradation. 3.8. Comparison of PG LR to Hand Calculated LR The log (LR) for a full single source three-cell profile was calculated by hand and with the PG systems STRmix TM and EuroForMix demonstrating concordant results (i.e., Hand: 27.45, STRmix TM 27.54, and EFM 27.55). The equations utilized to calculate the loci LRs are listed in Equations (6) and (7) . The total log(LR) for each method indicates similar results. (6) H e t e r o z y g o u s l o c i = 2 [ θ + ( 1 − θ ) p i ] [ θ + ( 1 − θ ) p j ] ( 1 + θ ) ( 1 + 2 θ ) (7) H o m o z y g o u s l o c i = [ 3 θ + ( 1 − θ ) p i ] [ 2 θ + ( 1 − θ ) p i ] ( 1 + θ ) ( 1 + 2 θ ) Fst = θ = 0.01 p i = allele frequency for allele i p j = allele frequency for allele j 3.9. Sensitivity/Specificity The sensitivity and specificity of the PG systems for DSCS was tested by analyzing 110 × 1 cell subsamples, 81 × 2-cell single source subsamples, 50 × 2-cell mini mixtures samples. With regards to sensitivity (i.e., true positive rate), each subsample was compared (within the PG software) to the known reference DNA profile and a likelihood ratio was reported indicating a strength of inclusion for that known donor. Any instance in which a log (LR) > 0 was obtained was considered inclusionary. A second threshold of log (LR) = 6 was also examined as advantageous matches of unrelated non-donors were not seen when log (LR)s ≥ 6 were achieved. A previous publication has reported the STRmix TM DSCS specificity obtained using these two thresholds. The EFM specificity is reported in . Replicate refers to ≤6 cell subsamples originating from the same donor probabilistically combined (via a joint probability function) to produce a single LR. As seen in the table and previous publication, utilization of the replicate analysis function resulted in 100% of replicate samples returning inclusionary LRs with both PG systems, while replicate log(LR)s ≥ 6 occurred in 94%, 93%, and 86% of STRmix TM 1-cell, single source 2-cell, and mixed 2-cell subsamples, respectively . A more detailed discussion of replicate analysis is provided in . In order to test the specificity (i.e., true negative rate) of the cell subsamples, a 1000 person known non-contributor database was utilized to test for false positives. False positives occurred any time an inclusionary LR (i.e., log(LR) > 0) was obtained for a known non-donor. So, for every subsample analyzed, 1000 known non-contributor profiles were substituted in the H 1 position of the likelihood ratio. The same previous publication has reported the results for STRmix TM in which the log (LR) values for known donor subsamples are plotted with respect to the number of alleles detected. The EFM results are provided in where each single source subsample returns a single LR for the known contributor (green or yellow) and 1000 LRs for the non-contributors (orange). Likelihood ratios of 0 were plotted as a log(LR) = −350. As allele count increased, the log (LR) attained for known donors increased while the log(LR)s achieved for non-donors decreased. The majority of false positives occurred at log (LR)s = 1–3, indicative of ‘uninformative’ or ‘limited support’ . However, when a 10 6 LR threshold (dashed line) was utilized, non-contributor false positives were not seen above this measure, and when allele counts exceeded 15, there was generally good separation between known and non-donors. The (LR)s achieved from 2-cell mini mixture subsamples were also analyzed with no false positives being seen at or above the log(LR) = 6 threshold ( ). 3.10. Number of Contributors The impact of overestimating the number of contributors (N + 1) with DSCS samples was determined by analyzing the 35 single source subsamples ( ). For the majority of samples, comparable results were obtained when analyzing samples according to the correct NOC as well as N + 1 (as indicated by the blue (STRmix TM ) and green (EFM) data points along the y = x line). With STRmix TM , three subsamples returned an LR of zero when analyzed as N = 1. However, when ran as N + 1 (i.e., N = 2) high LRs were achieved (shown boxed in ). These were samples with drop-out of a heterozygous allele that resulted in an LR = 0 at a specific locus (e.g., ). By increasing the NOC, the system seems to account for the possibility of this drop-out where it did not appear when ran as N = 1. Misidentifying the number of contributors by N + 1 does not appear to negatively impact analysis. The impact of underestimating the number of contributors (N − 1) with DSCS samples was determined by analyzing the 35× 2-cell mini-mixture subsamples ( ). As indicated by the blue and green data points along the y=x line, underestimating the NOC has a detrimental impact on analysis as true donors may be falsely excluded. This is not entirely surprising as underestimating the NOC in these instances would be misidentifying a mini-mixture as a single source, which would not occur in most instances. Non-contributor profiles were not analyzed with EuroForMix due to the slow speed of the database search function. The NOC estimation trends observed correlate to those seen with previous PG studies . 3.11. Replicates Utilizing the replicate analysis function of PG systems allows for the combination of multiple samplings originating from the same bulk sample into a single analysis. For standard bulk analysis, this takes the form of multiple amplifications of the same original extract . With respect to low template DNA, this has been shown to result in greater profile recovery for known donors as opposed to a single amplification and differentiates robust signals from un-replicated artifacts, thus decreasing stochastic effects . The replicate analysis capability of the PG systems was therefore evaluated for use with single or few cell subsamples. Up to 6 cell subsamples originating from the same donor and collected from complex 2–6 person mixtures were utilized with replicate analysis with an average of 4 subsamples (±1) used. An initial screening of these subsamples was conducted in which samples that provided an inclusionary log(LR) (i.e., log(LR) > 1) for a specific donor were considered for use. An alternative approach to cluster individual subsamples by donor may be to utilize the mixture to mixture and common donor applications in the DBLR software , or employ classic clustering algorithms such as K-means or EM . A comparison of the log(LR)s obtained from individual subsamples compared to replicate analysis is provided in (single source) and (two-cell mini-mixtures). In nearly all instances, replicate analysis resulted in improved LRs compared to individual subsamples as indicated by the green datapoints above the dashed line (y = x), while the false positives orange data points decreased. Occasionally, replicate analysis with EFM resulted in a failed analysis presumably due to modeling issues arising from subsamples with differing degrees of degradation. However, the newly developed EuroForMix extension, EFMrep , prevented these failed investigations. Similar results were obtained with STRmix TM . Because the aim of direct single cell subsampling is to obtain single source (or less complex mini-mixtures) from a bulk mixture, the PG system’s accuracy was tested by intentionally misclassifying a subsample for replicate analysis. Single source and mini-mixture subsamples with non-inclusionary LRs (LR < 1) for the true donor(s) (i.e., incorrect donor for the replicate grouping) were added to the sample subset used for replicate analysis. Poor quality profiles (i.e., ~4/42 alleles) had relatively no impact on the replicate log(LR) obtained. However, as profile quality increased (by improved allele recovery), a significant decline occurred in the log(LR)s attained (i.e., LRs < 0) ( ). For mini-mixtures specifically, inclusionary LRs were occasionally still seen for the true donors when a misclassified subsample was added. However, these LRs were either comparable to the LRs obtained without the misclassified sample or decreased ( ). To briefly explain, provides the replicate log(LR) obtained when three two-cell mini-mixture subsamples comprising donors S5 and CM31 were utilized. Misclassification was tested by adding an additional ‘incorrect’ subsample to the replicate test (i.e., the three correctly classified S5CM31 mini-mixtures and one misclassified single source subsample) for donors SA10, S3, and S8. Misclassified mini-mixtures were additionally added to the S5CM31 mini-mixture replicate analysis (i.e., the three correctly classified S5CM31 mini-mixtures and one or two misclassified S3SA10 mini-mixtures). Encouragingly, the data from the EuroForMix analyses using the same samples were in close agreement with STRmix TM . Minor discrepancies between the two systems may be partially due to the fact EFM only models back and forward stutter; therefore, any half-back or double back stutter was removed prior to analysis with EFM. Furthermore, all back/forward stutters with EFM are all modeled with the same distribution rather than separately as with STRmix TM , and the possibility of different loci having different amplification efficiencies is not considered. 3.12. DSCS Deconvolution Compared to Standard Bulk Analysis In a single mixture, the DSCS procedure produces up to six different subsample types with associated LRs from separate putative contributors. These subsample types include one- and two-cell subsamples, two-cell mini-mixture subsamples, one- and two-cell replicates, and mini-mixture replicates. After replicate analysis, the maximum replicate likelihood ratio obtained is reported. Seven equimolar complex mixtures were deconvoluted using this DSCS process. , and (EFM) compare the ‘ground-truth’ or maximum possible log(LR) per known donor (i.e., log(1/RMP) from single source reference profiles (black bars)) to the standard analysis results (i.e., log(LR) per donor, where LR = known donor and N-1 unrelated individuals contributing to the mixture vs. N unknown unrelated individuals contributing to the mixture) (grey bars)) and the DSCS replicate log(LR)s per donor (green bars). Improved genotype information was achieved per donor for each mixture with the most significant improvement occurring with the complex five- and six-person mixtures ( and ). Standard analysis was unable to be conducted for these mixtures due to limitations of the software as it is only recommended on mixtures comprising up to four people . The STRmix TM results for these same mixtures have been previously reported . The two different PG software systems returned remarkably similar individual contributor LR values with no significant discordant results. The EFM log(LR) results ranged from 6 to 29, while STRmix TM results ranged from 8 to 28. This represented an additional diagnostic check on the accuracy and reliability of the DSCS method’s performance with the sample set studied. 3.13. Upgraded Version of PG Software (STRmix TM Update to Version 2.9.1) PG software is being continuously updated. It behooves the user to test and evaluate the new version before relying on it to perform additional analyses. Since the initial validation and comparison of the two PG systems reported in this study, an updated version of STRmix TM has been released. This section is included to illustrate to readers how the performance of a new version of PG software (using STRmix TM as an example) can be compared to the one used for previous work. Upgrading to STRmix TM v2.9.1 requires remodeling of the Model Maker parameters . For the same data previously used to model v2.8, the updated v2.9.1 Model Maker parameter results are listed in . Fifty-five subsamples were then re-analyzed with version 2.9.1 ( ) showing there is close agreement in the log(LR)s obtained with versions 2.8 and 2.9.1.
Analysis of single cells was posited as an approach to deconvolute complex DNA mixtures by obtaining single source DNA profiles from all donors [ , , , , , , , , , , , , ]. To test the applicability of such an approach which we term DSCS (direct single cell subsampling) , the creation of known composition complex DNA mixtures was required. One such way to accomplish this is to create individual donor cell suspensions; then, the appropriate concentration of each of the donor’s cells can be combined to create defined mixture ratios. Because the structural integrity of individual cells will depend upon their surrounding environment, different suspension mediums were tested including 1× PBS, water, and TE −4 . PBS buffer, which although routinely emphasized as a requirement for single cell analysis had a detrimental effect on the resulting DNA profiles. This may be due to the fact that cells analyzed in forensic applications are already dehydrated/lysed and therefore PBS buffer, normally required to maintain a balanced isotonic state between the inside and outside of living, viable cells, may be redundant as well as lacking nuclease protection components. TE −4 was therefore chosen as the cell suspension medium of choice due to improved allele recovery compared to 1× PBS and nuclease-free water ( ). This finding is not unsurprising as TE −4 does contain nuclease-inhibiting activity. Additionally, aged samples in the form of dried saliva stains and dried stains originating from TE −4 cell suspensions were compared in which relatively comparable allele recovery was obtained from both sample types. Although a slight decrease in allele recovery was seen from the cell suspension stains, cell suspensions were concluded to be a viable convenient alternative to saliva to use as a means of preparing defined composition mixtures ( ).
STRmix TM requires optimized stutter parameters for individual loci and allows for modeling of any stutter type. EuroForMix, on the other hand, only models forward and reverse stutter and models all stutter with the same distribution as opposed to per loci and does not require optimization on the part of the user. A graph of SR vs. allele number per loci and stutter type was created ( ). provide the slope and intercept (linear stutter) or the average (non-linear stutter) for each locus and stutter type. Due to the presence of elevated stutter, R 2 values obtained for single cell subsamples are lower than typically observed with standard analysis and high maximum allowable stutter percentages were utilized (back stutter 0.7, forward stutter 0.7, double back stutter 0.3, half-back stutter 0.5, half-forward stutter 0.15, −1.5 repeat stutter 0.15). The stutter distribution parameters were further determined utilizing the STRmix TM Model Maker function.
The saturation threshold for the 3500 Genetic Analyzer was determined (30,000 RFUs) by comparing the expected allele height (Ea) and the observed allele height (Oa) from single source samples. The point at which (Ea) and (Oa) begin to consistently diverge is the saturation limit of the instrument. Therefore, any profiles with allele heights > 30,000 RFUs should be diluted and capillary electrophoresis reconducted to prevent the software from incorrectly estimating the Ea in DNA profiles. Because cell subsamples are low template in nature, stochastic effects such as elevated stutter are commonly seen. Therefore, the instrument saturation limit was determined using standard DNA samples (i.e., 25 μL rxn volume amplification of 62.5 pg to 1 ng DNA extracts n = 38) ( A). Loci with simple repeat structure were utilized (D16S539, CSF1PO, TPOX, D5S818, D13S317, D7S820, and D10S1248). The expected allele heights are determined using the back stutter regression lines determined when creating stutter files for each loci and Equation (3). (3) E a = s t u t t e r p e a k h e i g h t s t u t t e r r a t i o For example, by using the DSCS stutter regression line for D16S539 (𝑆𝑅 = (0.0177 × 𝐴𝑙𝑙𝑒𝑙𝑒 𝑛𝑢𝑚𝑏𝑒𝑟) − 0.1249) ( ), the stutter ratio expected for allele 11 at D16S539 is determined to be SR= 0.0698. However, due to the aforementioned elevated stutter exhibited with DSCS subsamples (n = 189), divergence in the Ea and Oa for our DSCS subsamples is occasionally still seen even below the 30,000 RFUs threshold ( B).
Drop-in can be described as 1–2 low-level non-reproducible peaks present within a DNA profile . However, because of the low template nature of single cell analysis, drop-in peaks with DSCS samples may have higher peak heights than seen with standard analysis methods. Therefore, a drop-in cap of 30,000 RFUs was used with STRmix TM to allow for any height allele to be considered as possible drop-in. However, re-evaluation of all single-cell subsamples analyzed within this DSCS PG validation as well as subsequent experiments [ , , ] indicated the highest drop-in peak observed was around 4800 RFU, indicating a drop-in cap of 5000 RFU may be more appropriate and may prevent the software from falsely identifying true allele peaks as potential drop-in. To determine the rate of drop-in occurrences, thirty-five negative controls (i.e., zero-cell subsamples or amplification blanks) were analyzed for the presence of drop-in alleles using Equation (4). (4) D r o p i n r a t e = n u m b e r o f d r o p i n e v e n t s n u m b e r o f l o c i s c o r e d × n u m b e r o f s a m p l e s The drop-in data did not fit a gamma distribution and limited drop-in data were available. Therefore, a uniform distribution was utilized with STRmix TM . The drop-in rate was determined to be 0.0164 ( ). The same drop-in rate was used for EuroForMix analysis with the default lambda of 0.01.
Utilization of the STRmix TM software for sample analysis requires creation of a specific protocol-based statistical model that includes the determination of various empirically determined parameter variance probability distribution plots (i.e., stutter, allele, and locus specific amplification efficiency (LSAE)). This requires utilization of the Model Maker function in which single source subsamples are used. It is important for samples utilized within Model Maker to be consistent with the type and quality of samples expected to be encountered in case work. If pristine quality samples were used, then the system may not be able to account for the extreme heterozygous peak height imbalance that is commonly seen with single cell analysis due to the low template nature of the samples ( ). Because of this, 161 × −1, −2, −3, −4, or 5-cell subsamples comprised of fresh buccal cells as well as cells collected from cell suspension mixtures were utilized. The EFM software does not utilize a Model Maker type function. Model Maker performance DSCS versus standard bulk analysis was compared by plotting the log(Hb) vs. Average Peak Height of single source bulk or single cell samples ( ) with the dashed red lines indicating the 95% bounds (+/− 2 standard deviation of the mean) calculated using Equation (5). (5) ± 2 × 1.96 C 2 A P H C 2 = the 50th percentile of the allelic peak height variance Utilizing a normal distribution, 87.6% of the single source cell data fell within the 95% bounds. Ideally ≥95% of the data would have been encapsulated within the 95% bounds. A comparison of A and B shows that the 95% bounds for the DSCS samples are much broader than standard analysis parameters to account for the low template and stochastic effects seen with single cells (i.e., the extreme peak height imbalance).
With STRmix TM , various post burn-in accepts (5000, 50,000, and 500,000) were tested 5× with a single source three-cell subsample ( ). As the number of burn-in accepts increases, so does the time required for analysis as well as the precision of the LR results. STRmix TM default setting is 50,000 burn-in accepts. However, when analyzing single cell subsamples, the most consistent log (LR) results occurred at 500,000 post burn-in accepts.
To test the accuracy of the optimized PG parameters, a three-cell subsample (“S5-3C-1”) was analyzed with both PG software systems and the log(LR) was reported. The sample was then artificially degraded by decreasing the peak heights of high molecular weight alleles by 80% and low molecular weight alleles by 5%. An inhibited version of the sample in which the alleles at loci D22S1045, D21S11, D13S317, and D2S1338 were artificially inhibited by decreasing their peak heights by 40% was analyzed. The sample was also tested in which a drop-in eight allele with 15,000 RFUs was artificially added at loci D5S818. With EFM, the same log(LR) results were obtained for all four profiles (i.e., original, with drop in, inhibited, degraded) (EFM ). A figure showing a single dye channel of these artificially created samples was previously published in along with the STRmix TM results demonstrating the sample log(LR) for each of the four sample state scenarios (i.e., log(LR) = 27). With STRmix TM , the LSAE and APH for the subsample S5-3C-1 are shown in according to increasing molecular weight of the STR loci, indicating that similar trends are seen with the two different parameters. EFM, on the other hand, does not allow for the possibility of different loci having different amplification efficiencies. A comparison of the LSAE obtained for the regular sample and the inhibited sample is then shown in where a decrease in the amplification efficiency trend is obtained for the artificially inhibited loci. The degradation curves obtained from STRmix TM , are provided in for the regular sample, degraded sample, and inhibited sample, indicating that only the artificially degraded sample exhibited a characteristic degradation curve that indicated the presence of degradation.
The log (LR) for a full single source three-cell profile was calculated by hand and with the PG systems STRmix TM and EuroForMix demonstrating concordant results (i.e., Hand: 27.45, STRmix TM 27.54, and EFM 27.55). The equations utilized to calculate the loci LRs are listed in Equations (6) and (7) . The total log(LR) for each method indicates similar results. (6) H e t e r o z y g o u s l o c i = 2 [ θ + ( 1 − θ ) p i ] [ θ + ( 1 − θ ) p j ] ( 1 + θ ) ( 1 + 2 θ ) (7) H o m o z y g o u s l o c i = [ 3 θ + ( 1 − θ ) p i ] [ 2 θ + ( 1 − θ ) p i ] ( 1 + θ ) ( 1 + 2 θ ) Fst = θ = 0.01 p i = allele frequency for allele i p j = allele frequency for allele j
The sensitivity and specificity of the PG systems for DSCS was tested by analyzing 110 × 1 cell subsamples, 81 × 2-cell single source subsamples, 50 × 2-cell mini mixtures samples. With regards to sensitivity (i.e., true positive rate), each subsample was compared (within the PG software) to the known reference DNA profile and a likelihood ratio was reported indicating a strength of inclusion for that known donor. Any instance in which a log (LR) > 0 was obtained was considered inclusionary. A second threshold of log (LR) = 6 was also examined as advantageous matches of unrelated non-donors were not seen when log (LR)s ≥ 6 were achieved. A previous publication has reported the STRmix TM DSCS specificity obtained using these two thresholds. The EFM specificity is reported in . Replicate refers to ≤6 cell subsamples originating from the same donor probabilistically combined (via a joint probability function) to produce a single LR. As seen in the table and previous publication, utilization of the replicate analysis function resulted in 100% of replicate samples returning inclusionary LRs with both PG systems, while replicate log(LR)s ≥ 6 occurred in 94%, 93%, and 86% of STRmix TM 1-cell, single source 2-cell, and mixed 2-cell subsamples, respectively . A more detailed discussion of replicate analysis is provided in . In order to test the specificity (i.e., true negative rate) of the cell subsamples, a 1000 person known non-contributor database was utilized to test for false positives. False positives occurred any time an inclusionary LR (i.e., log(LR) > 0) was obtained for a known non-donor. So, for every subsample analyzed, 1000 known non-contributor profiles were substituted in the H 1 position of the likelihood ratio. The same previous publication has reported the results for STRmix TM in which the log (LR) values for known donor subsamples are plotted with respect to the number of alleles detected. The EFM results are provided in where each single source subsample returns a single LR for the known contributor (green or yellow) and 1000 LRs for the non-contributors (orange). Likelihood ratios of 0 were plotted as a log(LR) = −350. As allele count increased, the log (LR) attained for known donors increased while the log(LR)s achieved for non-donors decreased. The majority of false positives occurred at log (LR)s = 1–3, indicative of ‘uninformative’ or ‘limited support’ . However, when a 10 6 LR threshold (dashed line) was utilized, non-contributor false positives were not seen above this measure, and when allele counts exceeded 15, there was generally good separation between known and non-donors. The (LR)s achieved from 2-cell mini mixture subsamples were also analyzed with no false positives being seen at or above the log(LR) = 6 threshold ( ).
The impact of overestimating the number of contributors (N + 1) with DSCS samples was determined by analyzing the 35 single source subsamples ( ). For the majority of samples, comparable results were obtained when analyzing samples according to the correct NOC as well as N + 1 (as indicated by the blue (STRmix TM ) and green (EFM) data points along the y = x line). With STRmix TM , three subsamples returned an LR of zero when analyzed as N = 1. However, when ran as N + 1 (i.e., N = 2) high LRs were achieved (shown boxed in ). These were samples with drop-out of a heterozygous allele that resulted in an LR = 0 at a specific locus (e.g., ). By increasing the NOC, the system seems to account for the possibility of this drop-out where it did not appear when ran as N = 1. Misidentifying the number of contributors by N + 1 does not appear to negatively impact analysis. The impact of underestimating the number of contributors (N − 1) with DSCS samples was determined by analyzing the 35× 2-cell mini-mixture subsamples ( ). As indicated by the blue and green data points along the y=x line, underestimating the NOC has a detrimental impact on analysis as true donors may be falsely excluded. This is not entirely surprising as underestimating the NOC in these instances would be misidentifying a mini-mixture as a single source, which would not occur in most instances. Non-contributor profiles were not analyzed with EuroForMix due to the slow speed of the database search function. The NOC estimation trends observed correlate to those seen with previous PG studies .
Utilizing the replicate analysis function of PG systems allows for the combination of multiple samplings originating from the same bulk sample into a single analysis. For standard bulk analysis, this takes the form of multiple amplifications of the same original extract . With respect to low template DNA, this has been shown to result in greater profile recovery for known donors as opposed to a single amplification and differentiates robust signals from un-replicated artifacts, thus decreasing stochastic effects . The replicate analysis capability of the PG systems was therefore evaluated for use with single or few cell subsamples. Up to 6 cell subsamples originating from the same donor and collected from complex 2–6 person mixtures were utilized with replicate analysis with an average of 4 subsamples (±1) used. An initial screening of these subsamples was conducted in which samples that provided an inclusionary log(LR) (i.e., log(LR) > 1) for a specific donor were considered for use. An alternative approach to cluster individual subsamples by donor may be to utilize the mixture to mixture and common donor applications in the DBLR software , or employ classic clustering algorithms such as K-means or EM . A comparison of the log(LR)s obtained from individual subsamples compared to replicate analysis is provided in (single source) and (two-cell mini-mixtures). In nearly all instances, replicate analysis resulted in improved LRs compared to individual subsamples as indicated by the green datapoints above the dashed line (y = x), while the false positives orange data points decreased. Occasionally, replicate analysis with EFM resulted in a failed analysis presumably due to modeling issues arising from subsamples with differing degrees of degradation. However, the newly developed EuroForMix extension, EFMrep , prevented these failed investigations. Similar results were obtained with STRmix TM . Because the aim of direct single cell subsampling is to obtain single source (or less complex mini-mixtures) from a bulk mixture, the PG system’s accuracy was tested by intentionally misclassifying a subsample for replicate analysis. Single source and mini-mixture subsamples with non-inclusionary LRs (LR < 1) for the true donor(s) (i.e., incorrect donor for the replicate grouping) were added to the sample subset used for replicate analysis. Poor quality profiles (i.e., ~4/42 alleles) had relatively no impact on the replicate log(LR) obtained. However, as profile quality increased (by improved allele recovery), a significant decline occurred in the log(LR)s attained (i.e., LRs < 0) ( ). For mini-mixtures specifically, inclusionary LRs were occasionally still seen for the true donors when a misclassified subsample was added. However, these LRs were either comparable to the LRs obtained without the misclassified sample or decreased ( ). To briefly explain, provides the replicate log(LR) obtained when three two-cell mini-mixture subsamples comprising donors S5 and CM31 were utilized. Misclassification was tested by adding an additional ‘incorrect’ subsample to the replicate test (i.e., the three correctly classified S5CM31 mini-mixtures and one misclassified single source subsample) for donors SA10, S3, and S8. Misclassified mini-mixtures were additionally added to the S5CM31 mini-mixture replicate analysis (i.e., the three correctly classified S5CM31 mini-mixtures and one or two misclassified S3SA10 mini-mixtures). Encouragingly, the data from the EuroForMix analyses using the same samples were in close agreement with STRmix TM . Minor discrepancies between the two systems may be partially due to the fact EFM only models back and forward stutter; therefore, any half-back or double back stutter was removed prior to analysis with EFM. Furthermore, all back/forward stutters with EFM are all modeled with the same distribution rather than separately as with STRmix TM , and the possibility of different loci having different amplification efficiencies is not considered.
In a single mixture, the DSCS procedure produces up to six different subsample types with associated LRs from separate putative contributors. These subsample types include one- and two-cell subsamples, two-cell mini-mixture subsamples, one- and two-cell replicates, and mini-mixture replicates. After replicate analysis, the maximum replicate likelihood ratio obtained is reported. Seven equimolar complex mixtures were deconvoluted using this DSCS process. , and (EFM) compare the ‘ground-truth’ or maximum possible log(LR) per known donor (i.e., log(1/RMP) from single source reference profiles (black bars)) to the standard analysis results (i.e., log(LR) per donor, where LR = known donor and N-1 unrelated individuals contributing to the mixture vs. N unknown unrelated individuals contributing to the mixture) (grey bars)) and the DSCS replicate log(LR)s per donor (green bars). Improved genotype information was achieved per donor for each mixture with the most significant improvement occurring with the complex five- and six-person mixtures ( and ). Standard analysis was unable to be conducted for these mixtures due to limitations of the software as it is only recommended on mixtures comprising up to four people . The STRmix TM results for these same mixtures have been previously reported . The two different PG software systems returned remarkably similar individual contributor LR values with no significant discordant results. The EFM log(LR) results ranged from 6 to 29, while STRmix TM results ranged from 8 to 28. This represented an additional diagnostic check on the accuracy and reliability of the DSCS method’s performance with the sample set studied.
TM Update to Version 2.9.1) PG software is being continuously updated. It behooves the user to test and evaluate the new version before relying on it to perform additional analyses. Since the initial validation and comparison of the two PG systems reported in this study, an updated version of STRmix TM has been released. This section is included to illustrate to readers how the performance of a new version of PG software (using STRmix TM as an example) can be compared to the one used for previous work. Upgrading to STRmix TM v2.9.1 requires remodeling of the Model Maker parameters . For the same data previously used to model v2.8, the updated v2.9.1 Model Maker parameter results are listed in . Fifty-five subsamples were then re-analyzed with version 2.9.1 ( ) showing there is close agreement in the log(LR)s obtained with versions 2.8 and 2.9.1.
Two different PG software systems, STRmix TM and EuroForMix, which employ different models to deconvolute forensic DNA mixtures were validated for single cell (and few cell) analysis. While only two software systems were evaluated, it is envisioned that single cell PG analysis should be possible with other creditable PG software systems after appropriate validation. The single cell PG validation process is comparable to those validation studies conducted for standard DNA bulk extract analysis, although there are some notable differences. Differences occur (for STRmix TM specifically) in that rather than utilizing single source DNA extracts spanning profile quality observed within casework (i.e., < 1 ng), 1-, 2-, 3-, 4-, and 5-cell single source subsamples were utilized with the Model Maker function for DSCS validation. Similarly, stutter files are created for the DSCS process using 1-5 cell subsamples. Validation for the use of a higher number of subsampled cells (i.e., >5), is possible; however, a high degree of saturation was observed when >5 cells were used with our specific single cell workflow (i.e., 5 μL rxn volume and 32 cycles). Other marked differences occur in that rather than creating sensitivity and specificity plots according to RFUs, allele count is used instead (e.g., and ). While the overall process of validating PG for DSCS analysis is broadly similar to that performed for standard workflows, many of the optimized parameter values are markedly different. For example, the default drop-in cap and rate for STRmix TM v2.8 is 100 RFU and 0.0001, while those used with DSCS are 30,000 RFUs (though this can likely be decreased to 5000 RFUs) and 0.0164. Similarly, due to the low template effects of single cell analysis, an increase in elevated stutter is experienced with DSCS samples, thus resulting in high stutter maximums being utilized (e.g., 0.7 for forward and reverse stutter). This was further exemplified in and indicates that saturation thresholds (which are largely instrument dependent) should be determined using standard samples rather than single cell subsamples. Additionally, an increase in burn-in accepts from 50,000 to 500,000 is recommended for single cell analysis as this allows the software additional time to converge on the proper result. The biggest challenge with single cell analysis arises due to enhanced allele drop-out and peak height imbalance, especially in the extreme (but not uncommon) case in which a single allele from a heterozygous pair drops out while the sister allele is present at a high RFU (e.g., ). This can often result in the system adjudging the remaining allele as a homozygous allele especially if a high RFU is observed. This extreme imbalance was exemplified in showing the much broader variance experienced with single cells compared to standard analysis. However, even so, highly probative results are still able to be obtained, particularly when the PG replicate analysis function is utilized, which often results in DNA profile LRs comparable to the inverse of the random match probability for donor reference profiles (i.e., the upper bound of the achievable LRs ). The replicate analysis function of the PG software systems is an important component of the described DSCS single cell methodology. This enables the individual genotyping results from single cell subsamples from the same donor to be probabilistically combined (via a joint probability function) to produce a single LR. The effect is to better take into account the stochastic effects of low template DNA (such as in single cells) while at the same time recovering more of the true alleles originating from an individual donor (that would otherwise be lost due to drop out in a single sample). Since the initial validation of the PG software systems, we have modified our direct lysis/PCR strategy. The Casework Direct System (Promega, Madison, WI) has been found to significantly improve the quality of DNA profiles obtained with single cells (i.e., increased number of alleles recovered), prior to their joint use with the PG software replicate analysis functions. Perhaps one of the largest impediments to labs considering testing and evaluation of single cell PG applications is the cost associated with preparing and analyzing the hundreds of single cell subsamples needed to develop the PG model. While our manual DSCS protocol is significantly cheaper than other automated single cell sub sampling methods, another approach to potentially decrease cost would be to create in silico models of single cell subsamples . Other work with standard PG analysis has demonstrated comparable success from various labs when utilizing general PG parameters as opposed to lab specific parameters for comparable workflows (i.e., same amplification kit, reaction volume, PCR cycle number, and capillary electrophoresis model) . Therefore, it is feasible that a similar approach may be applicable for PG single cell analysis in the future.
|
An Investigation into Compound Likelihood Ratios for Forensic DNA Mixtures | 6ee295ad-0a30-4504-aa11-cced877a0e6a | 10048689 | Forensic Medicine[mh] | When forensic DNA testing reveals a concordance between a crime scene sample and a person of interest’s (POI) DNA profile, it is necessary to provide a statistic to evaluate the strength of the correspondence or the weight of the evidence. The likelihood ratio ( LR ) is acknowledged as the most powerful and relevant statistic used to calculate the weight of DNA evidence and is recommended by the DNA commission of the International Society of Forensic Genetics (ISFG) in forensic DNA mixture interpretation . The LR is a ratio of two conditional probabilities, probability densities, or numbers proportional to them. The LR is not exclusively used for the interpretation of forensic DNA evidence. It is used to assign the weight of evidence for other forensic evidence and used in many other situations in statistics. It follows from Bayes’ theorem where the odds form is: Pr H p | E , I Pr H d | E , I = Pr E | H p , I Pr E | H d , I × Pr H p | I Pr H d | I where E represents the evidence, I represents relevant background information, and H p and H d (or H a ) represent alternate hypotheses or propositions. Bayes’ theorem follows directly from the laws of probability and can be expressed in words as follows: Posterior odds = likelihood ratio × prior odds. An LR greater than one means the DNA evidence supports the proposition given in the numerator. An LR less than one means the evidence supports the alternate proposition given in the denominator. In forensic casework, the LR in Bayes’ theorem is typically written as shown above, with the probability of the evidence given the prosecution hypothesis forming the numerator and the probability of the evidence given the defence hypothesis as the denominator. The prosecution proposition ( H p ) is generally known and straightforward to apply, especially when only one POI is being considered. The defence are under no requirement to offer a proposition, and often they do not. If the defence proposition is available, then that should be selected. If not, a sensible ‘alternate’ proposition consistent with exoneration should be chosen. Hence, the use of H a for an alternate proposition can be a preferred descriptor. There is a well-established hierarchy of propositions that are informed by the evidence being assessed. The original three levels within the hierarchy are offence, activity, and source-level propositions . Forensic DNA evidence is typically evaluated at the sub-source or sub-sub-source level within the hierarchy . Within this paper, we discuss LR s assigned using sub-source level proposition sets. Below, we give an example of a sub-source set of propositions for a two-person mixed DNA profile considering one POI as a contributor (set one): Set one, the simple proposition pair, sub-source propositions ( LR for a single POI, no conditioning): H p : The DNA originated from the POI and one unknown individual, unrelated to the POI H a : The DNA originated from two unknown individuals, unrelated to the POI or each other The propositions assigned in a case should be mutually exclusive, address the issue of interest and be close to exhaustive in that they take account of relevant case information and ensure no reasonable consideration is omitted . The propositions considered must be plausible or sensible within the known framework of circumstances. The use of non-sensible propositions can lead to misleading LR s . If one is transparent about the information that has been used to form the propositions and willing to consider a re-evaluation of the findings given different propositions, should the information change, then this approach is robust. A simple proposition pair is where no more than one POI considered within H p is replaced with an unknown individual within H a . Proposition set one above is an example of a simple proposition pair. In the case of circumstances where there is more than one POI, there are multiple propositions that may be considered both under H p and H a . Consider a two-person mixture where two POI both give inclusionary LR s using a simple proposition pair. In this case, it is prudent to test whether these POI could explain the profile when considered together. This could be undertaken using a compound proposition pair, defined as one where more than one POI within H p is replaced with unknown donors in H a (, hereafter the ASB (American Standards Board) draft standard and see also ). Set two, the compound proposition pair, sub-source propositions ( LR for all POI together, no conditioning): H p : The DNA originated from POI 1 and POI 2 H a : The DNA originated from two unknown individuals, unrelated to either POI or each other Although this proposition pair is highly effective in assessing whether both POI could be donors together, reported without the simple LR s for each individual, it can appear to greatly overstate the weight against a POI who gives a small inclusionary or uninformative LR when considered individually but who is carried in the compound LR by the much stronger other donors to the mixture. Another form of proposition pair assumes the contribution of all POIs under H p and all but one POI under the alternate proposition. We cannot find a definition of this proposition pair in the ASB draft standard , although this appears to come under clause 4.5.b, where they are described as a variant of the simple proposition pair. We will term these conditional proposition pairs. If the contribution of all POIs is supported by the observations, then the LR for such a conditional proposition pair is a good approximation of the exhaustive LR , as described by Buckleton et al. (their Equations (7a) and (7b)). Set three, the conditional pair, considering POI 1 for a four-person mixture ( LR for a single POI, uses conditioning profiles): Hp : The DNA originated from POI1, POI2, POI3, and POI4 H a : The DNA originated from POI 2 , POI 3 , POI 4 , and one other individual, unrelated to POI 1 , POI 2 , POI 3 , and POI 4 Three additional conditional LR s would subsequently be assigned considering POI 2 , POI 3 , and POI 4 . This isolates the evidence for the contribution of each POI in turn. Note that there are other possible combinations of conditional propositions when considering mixtures of more than two individuals. For example, conditioning on only one or two known contributors within a four-person mixture. These partial conditioned LR s are not calculated within this paper but are explored by Duke et al. (see, for example, the study’s Table 4). Given these types of scenarios, the assignment of a compound LR is advised by the ASB draft standard . However, as these LR s may overinflate the evidence, they advise that the LR s derived from simple proposition pairs are the ones reported and not the compound LR unless this is exclusionary. Bright and Coble report that for individuals who are well-represented in the mixture, the logarithm of the compound LR is approximately the sum of the logarithm of the LR s for each of the known contributors considering simple proposition pairs. This is only approximately true and then only for true donors. Within this research, we investigate the behaviour of LR s assigned for known and non-contributors to a set of mixed DNA profiles using compound, conditional, and simple proposition pairs. We demonstrate that compound likelihood ratios can be obtained as the product of conditional likelihood ratios. We also demonstrate that, on average, conditional LR s result in higher LR s for a true donor and more exclusionary LR s for non-contributors than their equivalents using simple proposition sets.
2.1. Data Two sets of mixed GlobalFiler™ DNA samples from two different laboratories (termed Lab A and Lab B) were amplified. The data comprised thirty-two mixtures (two laboratories × four NOCs × four mixtures) consisting of four each of N = 2, N = 3, N = 4, and N = 5 contributors. Samples from Lab A were amplified using 28 PCR cycles and analysed on a 3500 Genetic Analyser with 1.2 kV, 20 s injection parameters and samples from Lab B were amplified using 29 PCR cycles and analysed on a 3500 Genetic Analyser with 1.2 kV, 24 s injection parameters. All profiles were analysed in GeneMapper™ ID-X V1.6 with analytical thresholds (AT) of 125 rfu and 100 rfu for Lab A and B, respectively. 2.2. Interpretation and LR Assignment Profiles were interpreted using STRmix™ ( https://strmix.com/ , accessed on 1 September 2022 ) V2.8 (Lab A) and V2.9.1 (Lab B), assuming the apparent number of contributors, which also equalled the experimental number of contributors. A summary of the STRmix™ assigned template and mixture proportion and the experimental design for each mixture is given in . These mixtures cover a broad range of template amounts, number of contributors, and mixture proportions and are representative of DNA profiles typically encountered in casework. LR s were assigned in STRmix™ for each known contributor using a set of simple propositions. Following the nomenclature of Slooten for a two-person mixture, the LR assigned for POI 1 using a simple proposition set is: L R 1 u / u u = L 1 , u L u , u where, for example, L 1 , u = Pr ( E | H 1 , u , I ) and H 1, u is the proposition that POI 1 and an unknown person unrelated to POI 1 are the donors. Compound LR s were assigned in STRmix™ for each mixture of the type L R 12 / u u = L 1 , 2 L u , u . Compound proposition pairs that gave an LR of 0 (exclusion) for true donors were re-deconvoluted using increased numbers of burn-in and post-burn-in accepts (×10 or ×100 compared to defaults). In addition, each profile was interpreted in STRmix™ V2.9.1 N times (where N is the number of contributors assigned to each mixture), each allowing the approximation to the exhaustive LR s to be assigned of the type L R 12 / 2 u = L 1 , 2 L 2 , u and L R 12 / 1 u = L 1 , 2 L 1 , u . These are termed conditional LR s. All LR s were assigned using the NIST 1036 Caucasian allele frequencies and F ST = 0.01. 2.3. Compound LR Derivation Using DBLR™ Conditional and simple LR s were assigned in DBLR™ ( https://strmix.com/dblr/ , accessed on 1 September 2022) for the 32 mixtures from Lab A and B. Sub-source LR s were assigned conditioning on the presence of each POI in turn. The conditional LR s were built sequentially depending on the number of contributors in the mixture. A derivation of the compound LR using conditional and simple LR s is given in . For example, for a two-person mixture with two POIs, the compound LR can be written as the product of a conditional LR and a simple LR : L R 12 / u u = L 1 , 2 L u , u = L 1 , 2 | L 1 , u L u , u | L 1 , u = L R 12 / 1 u L R 1 u / u u where L 1,2 is the likelihood of POI 1 and POI 2 both being contributors, L 1, u is the likelihood of POI 1 and one unknown, and L u , u is the likelihood of two unknown contributors to the two-person mixture. L R 12 / 1 u relates to the question: what is the likelihood of POI 2 also being present in the mixture, given POI 1 is present? For a three-person mixture, we consider the following: L R 123 / u u u = L 1 , 2 , 3 L u , u , u = L 1 , 2 , 3 | L 1 , 2 , u L u , u , u | L 1 , 2 , u = L R 123 / 12 u L R 12 u / 1 u u L R 1 u u / u u u Conditional LR s in DBLR™ differ from those in the STRmix™ because the POI can be assumed to be present without the need to undertake a separate deconvolution. Conditioning on the presence of a contributor in STRmix™ is only possible during deconvolution. Hence, if a deconvolution was first undertaken without conditioning, then a second deconvolution would need to be performed with conditioning before a conditional LR could be assigned. In the DBLR™ software, however, it is possible to assume the presence of a contributor even if the deconvolution was undertaken without conditioning. This makes it possible to use a single deconvolution to evaluate various conditional likelihood ratios. A factorisation of a compound likelihood ratio as a product of conditional likelihood ratios is exact because the weights for the genotype sets remain the same between the LR assignments. If STRmix™ is used to assign the conditional likelihood ratios, then these factorisations will hold only approximately as the weights for the genotype sets change. LR s were assigned using the NIST 1036 Caucasian allele frequencies and F ST = 0.01. 2.4. Non-Contributor LRs 2.4.1. Compound Propositions In addition to the LR s for known contributors, LR s for non-contributors were assigned. Three mixtures from each lab were selected: a three-, a four- and a five-person mixture where the compound propositions using all known donors produced a log 10 ( LR ) exceeding the sum of the sub-source log 10 ( LR ) for each individual donor. Two non-contributor genotypes were selected for each of the six mixtures. These non-contributors were either selected from a set of random donors where they resulted in inclusionary LR s using a simple proposition set or were constructed using genotypes from the known donor profiles. The inclusionary LR s ranged from two to over 38 million. Compound LR s were assigned for each mixture where the non-contributor replaced a true donor under H p . This was repeated with the non-donor replacing each true donor in each mixture set. For example, for the four-person mixtures, four compound LR calculations were undertaken where H p was considering: Donor 1, Donor 2, Donor 3, Non-donor A Donor 1, Donor 2, Non-donor A, Donor 4 Donor 1, Non-donor A, Donor 3, Donor 4 Non-donor A, Donor 2, Donor 3, Donor 4. 2.4.2. High-Risk Database, Simple and Conditional Propositions In addition, two high-risk databases of non-contributors were generated by randomly sampling alleles from the known contributors to each of the mixtures for each laboratory. In this manner, 1000 profiles were generated. LR s were assigned for each of the 32 mixtures using a simple and conditional proposition set. The conditional proposition set is conditioned on N-1 known donors. The simple proposition set considered the POI (the non-contributor from the high-risk database) under H p and an unknown under H a . The two sets of 1000 LRs were calculated within STRmix™ using the NIST 1036 Caucasian allele frequencies and F ST = 0.01. Approximate compound LR s were additionally assigned for non-contributors who gave non-exclusions ( LR ≠ 0) when using the conditional proposition set. These compound LR s were approximated by summing the log 10 ( LR ) for one known contributor using the simple proposition set and conditional log 10 ( LR )s, as shown in .
Two sets of mixed GlobalFiler™ DNA samples from two different laboratories (termed Lab A and Lab B) were amplified. The data comprised thirty-two mixtures (two laboratories × four NOCs × four mixtures) consisting of four each of N = 2, N = 3, N = 4, and N = 5 contributors. Samples from Lab A were amplified using 28 PCR cycles and analysed on a 3500 Genetic Analyser with 1.2 kV, 20 s injection parameters and samples from Lab B were amplified using 29 PCR cycles and analysed on a 3500 Genetic Analyser with 1.2 kV, 24 s injection parameters. All profiles were analysed in GeneMapper™ ID-X V1.6 with analytical thresholds (AT) of 125 rfu and 100 rfu for Lab A and B, respectively.
Profiles were interpreted using STRmix™ ( https://strmix.com/ , accessed on 1 September 2022 ) V2.8 (Lab A) and V2.9.1 (Lab B), assuming the apparent number of contributors, which also equalled the experimental number of contributors. A summary of the STRmix™ assigned template and mixture proportion and the experimental design for each mixture is given in . These mixtures cover a broad range of template amounts, number of contributors, and mixture proportions and are representative of DNA profiles typically encountered in casework. LR s were assigned in STRmix™ for each known contributor using a set of simple propositions. Following the nomenclature of Slooten for a two-person mixture, the LR assigned for POI 1 using a simple proposition set is: L R 1 u / u u = L 1 , u L u , u where, for example, L 1 , u = Pr ( E | H 1 , u , I ) and H 1, u is the proposition that POI 1 and an unknown person unrelated to POI 1 are the donors. Compound LR s were assigned in STRmix™ for each mixture of the type L R 12 / u u = L 1 , 2 L u , u . Compound proposition pairs that gave an LR of 0 (exclusion) for true donors were re-deconvoluted using increased numbers of burn-in and post-burn-in accepts (×10 or ×100 compared to defaults). In addition, each profile was interpreted in STRmix™ V2.9.1 N times (where N is the number of contributors assigned to each mixture), each allowing the approximation to the exhaustive LR s to be assigned of the type L R 12 / 2 u = L 1 , 2 L 2 , u and L R 12 / 1 u = L 1 , 2 L 1 , u . These are termed conditional LR s. All LR s were assigned using the NIST 1036 Caucasian allele frequencies and F ST = 0.01.
Conditional and simple LR s were assigned in DBLR™ ( https://strmix.com/dblr/ , accessed on 1 September 2022) for the 32 mixtures from Lab A and B. Sub-source LR s were assigned conditioning on the presence of each POI in turn. The conditional LR s were built sequentially depending on the number of contributors in the mixture. A derivation of the compound LR using conditional and simple LR s is given in . For example, for a two-person mixture with two POIs, the compound LR can be written as the product of a conditional LR and a simple LR : L R 12 / u u = L 1 , 2 L u , u = L 1 , 2 | L 1 , u L u , u | L 1 , u = L R 12 / 1 u L R 1 u / u u where L 1,2 is the likelihood of POI 1 and POI 2 both being contributors, L 1, u is the likelihood of POI 1 and one unknown, and L u , u is the likelihood of two unknown contributors to the two-person mixture. L R 12 / 1 u relates to the question: what is the likelihood of POI 2 also being present in the mixture, given POI 1 is present? For a three-person mixture, we consider the following: L R 123 / u u u = L 1 , 2 , 3 L u , u , u = L 1 , 2 , 3 | L 1 , 2 , u L u , u , u | L 1 , 2 , u = L R 123 / 12 u L R 12 u / 1 u u L R 1 u u / u u u Conditional LR s in DBLR™ differ from those in the STRmix™ because the POI can be assumed to be present without the need to undertake a separate deconvolution. Conditioning on the presence of a contributor in STRmix™ is only possible during deconvolution. Hence, if a deconvolution was first undertaken without conditioning, then a second deconvolution would need to be performed with conditioning before a conditional LR could be assigned. In the DBLR™ software, however, it is possible to assume the presence of a contributor even if the deconvolution was undertaken without conditioning. This makes it possible to use a single deconvolution to evaluate various conditional likelihood ratios. A factorisation of a compound likelihood ratio as a product of conditional likelihood ratios is exact because the weights for the genotype sets remain the same between the LR assignments. If STRmix™ is used to assign the conditional likelihood ratios, then these factorisations will hold only approximately as the weights for the genotype sets change. LR s were assigned using the NIST 1036 Caucasian allele frequencies and F ST = 0.01.
2.4.1. Compound Propositions In addition to the LR s for known contributors, LR s for non-contributors were assigned. Three mixtures from each lab were selected: a three-, a four- and a five-person mixture where the compound propositions using all known donors produced a log 10 ( LR ) exceeding the sum of the sub-source log 10 ( LR ) for each individual donor. Two non-contributor genotypes were selected for each of the six mixtures. These non-contributors were either selected from a set of random donors where they resulted in inclusionary LR s using a simple proposition set or were constructed using genotypes from the known donor profiles. The inclusionary LR s ranged from two to over 38 million. Compound LR s were assigned for each mixture where the non-contributor replaced a true donor under H p . This was repeated with the non-donor replacing each true donor in each mixture set. For example, for the four-person mixtures, four compound LR calculations were undertaken where H p was considering: Donor 1, Donor 2, Donor 3, Non-donor A Donor 1, Donor 2, Non-donor A, Donor 4 Donor 1, Non-donor A, Donor 3, Donor 4 Non-donor A, Donor 2, Donor 3, Donor 4. 2.4.2. High-Risk Database, Simple and Conditional Propositions In addition, two high-risk databases of non-contributors were generated by randomly sampling alleles from the known contributors to each of the mixtures for each laboratory. In this manner, 1000 profiles were generated. LR s were assigned for each of the 32 mixtures using a simple and conditional proposition set. The conditional proposition set is conditioned on N-1 known donors. The simple proposition set considered the POI (the non-contributor from the high-risk database) under H p and an unknown under H a . The two sets of 1000 LRs were calculated within STRmix™ using the NIST 1036 Caucasian allele frequencies and F ST = 0.01. Approximate compound LR s were additionally assigned for non-contributors who gave non-exclusions ( LR ≠ 0) when using the conditional proposition set. These compound LR s were approximated by summing the log 10 ( LR ) for one known contributor using the simple proposition set and conditional log 10 ( LR )s, as shown in .
In addition to the LR s for known contributors, LR s for non-contributors were assigned. Three mixtures from each lab were selected: a three-, a four- and a five-person mixture where the compound propositions using all known donors produced a log 10 ( LR ) exceeding the sum of the sub-source log 10 ( LR ) for each individual donor. Two non-contributor genotypes were selected for each of the six mixtures. These non-contributors were either selected from a set of random donors where they resulted in inclusionary LR s using a simple proposition set or were constructed using genotypes from the known donor profiles. The inclusionary LR s ranged from two to over 38 million. Compound LR s were assigned for each mixture where the non-contributor replaced a true donor under H p . This was repeated with the non-donor replacing each true donor in each mixture set. For example, for the four-person mixtures, four compound LR calculations were undertaken where H p was considering: Donor 1, Donor 2, Donor 3, Non-donor A Donor 1, Donor 2, Non-donor A, Donor 4 Donor 1, Non-donor A, Donor 3, Donor 4 Non-donor A, Donor 2, Donor 3, Donor 4.
In addition, two high-risk databases of non-contributors were generated by randomly sampling alleles from the known contributors to each of the mixtures for each laboratory. In this manner, 1000 profiles were generated. LR s were assigned for each of the 32 mixtures using a simple and conditional proposition set. The conditional proposition set is conditioned on N-1 known donors. The simple proposition set considered the POI (the non-contributor from the high-risk database) under H p and an unknown under H a . The two sets of 1000 LRs were calculated within STRmix™ using the NIST 1036 Caucasian allele frequencies and F ST = 0.01. Approximate compound LR s were additionally assigned for non-contributors who gave non-exclusions ( LR ≠ 0) when using the conditional proposition set. These compound LR s were approximated by summing the log 10 ( LR ) for one known contributor using the simple proposition set and conditional log 10 ( LR )s, as shown in .
For each mixture, the compound log 10 LR assigned in STRmix™ was the same as the sum of the conditional log 10 LR s and one simple log 10 LR assigned in DBLR™ (the log 10 ( LR ) was compared to six decimal places). This is the expected result. A summary of the sub-source LR s assigned using the simple proposition set and compound proposition set for the Lab A and Lab B mixtures is given in . LR s using simple proposition sets and the true donors are given as stacked columns where the LR for each contributor is given as a different colour. The compound LR is given for each mixture as a red asterisk. Exclusions ( LR = 0) were obtained for nine of the 32 mixtures using a compound proposition set and the true donors. These included one four-person mixture and all eight five-person mixtures. This is not unexpected when there are multiple unknown contributors. The sample space is so vast that it can be inadequately sampled by the number of default accepts. These profiles were re-interpreted in STRmix™ with ×10 or ×100 the default accepts (100,000 or 1,000,000 burn-in and 500,000 or 5,000,000 post-burn-in accepts) per chain to better explore the probability space in the deconvolution (see ). Following reinterpretation, compound LR s > 1 were assigned for all nine mixtures. These are the results shown in . Inspection of shows that the compound log 10 ( LR ) were larger than the sum of the individual log 10 ( LR )s using the simple proposition set for each known contributor for all but one sample. This is more pronounced for the high-order mixtures (N = 3 and greater). This is an overrepresentation of the weight of evidence against each individual contributor. The five-person mixture (Lab B, sample number 3), designed with donor ratios of 10:2:2:1:1 and with a 100 pg template for the lowest contributors interpreted using ×100 accepts resulted in a compound log 10 ( LR ) that was less than the sum of the individual log 10 ( LR )s (52.26 versus 57.47). The mixture proportions assigned by STRmix™ were 64%, 16%, 11%, 8% and 1%. The contributor position with the highest LR for two of the contributors to this mixture using simple proposition sets differed from the contributor order they aligned with for the compound LR . The sub-sub-source LR for one contributor was approximately 20 times lower in its compound LR position. The sub-sub-source LR for the other contributor was around 17 orders of magnitude lower. This contributor best aligned in the third contributor position using simple propositions with an approximate mixture proportion of 11% but was aligned as the trace fifth contributor with an approximate mixture proportion of 1% using the compound proposition set. This individual is one of the two lowest template donors. Their alignment in the third contributor position using simple propositions is likely due to the presence of a D2S1338 18.3 peak not originating from any actual donor and likely drop-in, which is favoured as an allele for the fifth contributor, and also given the amount of allele sharing between donors. The sum of the individual log 10 ( LR ) for each donor with simple propositions when in their experimentally designed contributor positions was 40.67. 3.1. Conditional LRs A plot of the log 10 ( LR ) assigned for the true donors to the 32 mixtures using the simple proposition set (per contributor) versus the conditional log 10 LRs (alternatively described as Slooten and Buckleton et al.’s approximation to the exhaustive ( LR )) is given in the top pane of . The LR s assigned given conditional propositions were larger than the LR s assigned using simple proposition sets for the same POI for all but one comparison. This was the five-person mixture from Lab B, sample number 3, discussed above. The data points for samples on the x = y line are for mixtures that were fully or close to fully resolved, and conditioning did not add any extra information to the interpretation. A plot of the log 10 ( LR ) assigned for the mixtures using compound propositions versus log 10 ( LR )s for the conditional propositions is given in the bottom pane of . The LR s assigned given conditional propositions were smaller than the LR s assigned using compound proposition sets. The data points at [~28, ~0] and [~28, ~27] and indicated as filled data points in are considering two different POIs contributing to the same mixture. The major is (almost) fully resolved, whereas the minor is very ambiguous. The major carries the minor in the log 10 ( LR ) considering compound propositions. When conditioning on the major (in the approximation of exhaustive propositions), no information is gained in relation to the minor’s genotype. Vice versa, when conditioning on the minor, no information is gained in relation to the major’s genotype. 3.2. Non-Contributor Tests 3.2.1. Compound Propositions The twelve non-contributors (two for each of the six mixtures tested in ) that had previously given inclusionary LR s using simple proposition sets resulted in exclusions ( LR = 0) when using compound propositions, where they replaced, one by one, each of the true donors in the proposition. 3.2.2. High-Risk Database, Simple and Conditional Propositions A plot of log 10 ( LR ) given a simple proposition set versus the template assigned in STRmix™ (in rfu) for the high-risk database of non-contributors is given in . Overall, 56% of comparisons were exclusions ( LR = 0) and are plotted around log 10 ( LR ) = −40 in . A plot of log 10 ( LR ) given a conditional proposition set versus the template assigned in STRmix™ (in rfu) for the high-risk database of 1000 non-contributors is given in . The conditioned individual(s) was a known donor, and the POI was a database individual. Over 99% of comparisons resulted in LR = 0. Compound log 10 ( LR ) values for non-contributors within the high-risk database, which resulted in LR > 0 when assigned using a conditional proposition set, are plotted against the corresponding conditional log 10 ( LR ) values in . The compound LR is always greater than the conditional LR for the non-donors.
A plot of the log 10 ( LR ) assigned for the true donors to the 32 mixtures using the simple proposition set (per contributor) versus the conditional log 10 LRs (alternatively described as Slooten and Buckleton et al.’s approximation to the exhaustive ( LR )) is given in the top pane of . The LR s assigned given conditional propositions were larger than the LR s assigned using simple proposition sets for the same POI for all but one comparison. This was the five-person mixture from Lab B, sample number 3, discussed above. The data points for samples on the x = y line are for mixtures that were fully or close to fully resolved, and conditioning did not add any extra information to the interpretation. A plot of the log 10 ( LR ) assigned for the mixtures using compound propositions versus log 10 ( LR )s for the conditional propositions is given in the bottom pane of . The LR s assigned given conditional propositions were smaller than the LR s assigned using compound proposition sets. The data points at [~28, ~0] and [~28, ~27] and indicated as filled data points in are considering two different POIs contributing to the same mixture. The major is (almost) fully resolved, whereas the minor is very ambiguous. The major carries the minor in the log 10 ( LR ) considering compound propositions. When conditioning on the major (in the approximation of exhaustive propositions), no information is gained in relation to the minor’s genotype. Vice versa, when conditioning on the minor, no information is gained in relation to the major’s genotype.
3.2.1. Compound Propositions The twelve non-contributors (two for each of the six mixtures tested in ) that had previously given inclusionary LR s using simple proposition sets resulted in exclusions ( LR = 0) when using compound propositions, where they replaced, one by one, each of the true donors in the proposition. 3.2.2. High-Risk Database, Simple and Conditional Propositions A plot of log 10 ( LR ) given a simple proposition set versus the template assigned in STRmix™ (in rfu) for the high-risk database of non-contributors is given in . Overall, 56% of comparisons were exclusions ( LR = 0) and are plotted around log 10 ( LR ) = −40 in . A plot of log 10 ( LR ) given a conditional proposition set versus the template assigned in STRmix™ (in rfu) for the high-risk database of 1000 non-contributors is given in . The conditioned individual(s) was a known donor, and the POI was a database individual. Over 99% of comparisons resulted in LR = 0. Compound log 10 ( LR ) values for non-contributors within the high-risk database, which resulted in LR > 0 when assigned using a conditional proposition set, are plotted against the corresponding conditional log 10 ( LR ) values in . The compound LR is always greater than the conditional LR for the non-donors.
The twelve non-contributors (two for each of the six mixtures tested in ) that had previously given inclusionary LR s using simple proposition sets resulted in exclusions ( LR = 0) when using compound propositions, where they replaced, one by one, each of the true donors in the proposition.
A plot of log 10 ( LR ) given a simple proposition set versus the template assigned in STRmix™ (in rfu) for the high-risk database of non-contributors is given in . Overall, 56% of comparisons were exclusions ( LR = 0) and are plotted around log 10 ( LR ) = −40 in . A plot of log 10 ( LR ) given a conditional proposition set versus the template assigned in STRmix™ (in rfu) for the high-risk database of 1000 non-contributors is given in . The conditioned individual(s) was a known donor, and the POI was a database individual. Over 99% of comparisons resulted in LR = 0. Compound log 10 ( LR ) values for non-contributors within the high-risk database, which resulted in LR > 0 when assigned using a conditional proposition set, are plotted against the corresponding conditional log 10 ( LR ) values in . The compound LR is always greater than the conditional LR for the non-donors.
The conditional LR showed that, on average, the LR assigned to true donors was larger than the LR assigned using simple proposition sets for the same POI. This is because conditioning on another true donor adds information to the interpretation allowing for better resolution of the remaining genotypes. This is the known effect of conditional LR s. The data points on or about the line of equality in (top pane) are profiles that were fully resolved (or close to fully resolved), where conditioning on a contributor did not add extra information to the interpretation. The conditional LR was always lower than (or equal to) the LR using compound propositions ( bottom pane). The rate of adventitious matches for high-risk non-contributors created by sampling alleles from known contributors was significantly higher when using simple proposition sets compared with conditional proposition sets ( versus ). Conditional LR s have an increased power to differentiate between true and false donors. Conditioning on a true donor should increase LR s for other true donors (as demonstrated in ) and lower them for false donors. This is demonstrated again within this work. The high-risk non-contributors represent a ‘worst case’ scenario not typically encountered in casework other than when mixtures of relatives are involved. In relation to simple proposition sets, Slooten states , “The hypotheses for L R 1 u / u u only use the person of interest under investigation here. It may seem at first sight as an unbiased way to present the evidence, not using any other POI whose contribution is also disputed as assumed contributors. But it is easily overlooked that in fact one then assumes that the other POI did not contribute, and that this assumption is not at all supported by the data.” The simple proposition under H p might also not represent the most logical scenario for the prosecution given the case circumstances. The assignment of a compound LR is a natural extension if multiple POIs give inclusionary statistics when using simple proposition sets. We have shown that the logarithm of the compound LR is the sum of conditional log 10 ( LR ) and a simple log 10 ( LR ) for the individual contributors. However, the compound LR is only useful as a test of whether two or more POI can both be donors. In the overwhelming majority of cases, they are an inappropriate expression of the weight of evidence for any individual donor and may be too high or too low. Compound proposition sets have a higher chance of both false inclusionary support (non-donor carried by strong LR s of other donors), as shown in and false exclusionary support ( LR = 0 due to the vast sampling space and computing limitations). In general, if multiple POIs can be included in a mixture individually, and the ground truth is that all POIs have contributed, we expect the compound log 10 ( LR ) to be greater than the sum of the log 10 ( LR )s assigned using simple proposition sets ( ). Mixtures with the greatest ambiguity (or least well resolved) will typically have the greatest difference between the compound log 10 ( LR ) and the sum of the individual simple log 10 ( LR )s (refer to ). This is because, in the compound LR , the LR s for the individual contributors (for example, LR 1 and LR 2 for a two-person mixture) are not independent. Conditioning on a POI adds information to the interpretation, reducing the number of genotype combinations possible for the remaining contributor/s. Fully resolved mixtures are a special case where the compound log 10 ( LR ), the sum of conditional log 10 ( LR )s, and the sum of the simple log 10 ( LR ) for each true donor POI will all be equal, as long as sub-sub-source propositions are considered. This is because when the mixture is fully resolved in the compound LR calculation L R 1 u / u u and L R 2 u / u u are now independent, i.e., conditioning on a POI being present does not add any extra information to the calculation. We have demonstrated that, for some samples with a high number of contributors, the compound LR is zero even though for each true donor POI the simple LR is inclusionary. In these samples, the genotype combinations of all true donor contributors individually were accepted at least once across the posterior burn-in iterations, but the genotype combination explaining all true donor contributors in combination was not accepted within one iteration. This is not unexpected and arises because the sample space is vast. In these cases, we recommend the use of extended MCMC accepts within the interpretation. This allowed for more time to explore the sample space and was also a finding of Duke et al. . Where it is necessary to determine if multiple POIs could together be donors to relatively high template complex mixtures comprising four or five contributors, this may require additional MCMC accepts to fully explore the range of possible genotype combinations at each locus. The additional accepts may allow a wider range of genotype combinations to be accepted, thereby preventing an exclusion.
When assigning LRs in forensic casework, an analyst may have some idea of the most appropriate prosecution proposition but very rarely has knowledge of the most appropriate defence proposition. In the absence of this information, a reasonable set may be selected in a way that maintains the legitimate interests of the defence. This can be informed by case circumstances. An understanding of the performance of the LR under certain proposition sets can also help an analyst make this decision. It may be worthwhile benchmarking two of the recommendations in the draft ASB standard; recommendations 4.4 and 4.5 . We note that these recommendations are in draft. Recommendation 4.4: A profile should be assigned as a conditioning profile to a mixture when an individual is identified as an intimate contributor or when it is reasonable to assume their presence based on case-specific information and the associated data supports the assumption. The conditioning profile could be from the complainant, POI, or other individuals, depending on the case scenario. In the published guidelines for setting sub-source propositions , the DNA Commission of the International Society for Forensic Genetics define relevant case circumstances as those that “include only the case information that is needed for the formulation of the propositions and for assigning the probabilities of the results”. Buckleton et al. describe forensically relevant case circumstances for a DNA case as “information that will help formulate the appropriate alternative, determine the number(s) of contributors, and select the relevant population”. They do not consider “information such as prior conviction, motive, presence of other types of evidence, or a confession as relevant forensic information”. As much relevant case information should be gathered as practical before formulating the propositions. The conclusions of this work suggest that this recommendation should be greatly strengthened. We reprise Slooten’s insightful comment that not conditioning is also an assumption : That the profile being considered for conditioning is not a donor and that this is not at all supported by the data. It is very tempting to feel that not assuming is somehow safe or conservative. But the choice is between assuming that the conditioning profile is, or is not, a donor. If the data support the presence of this profile, it can be very detrimental not to assume their presence. This is because of the much-enhanced ability to differentiate true from false donors when conditioning is applied. A useful way through this issue is to use the approximation to the exhaustive LR . This enables a balanced approach that assumes that the conditioning profile either is or is not a donor. However, in the event that two or more donors cannot both or all be donors (or that the compound LR is much less than 1), it is still necessary to state this explicitly. Recommendation 4.5: The analysis should separate the propositions into their simplified constituents (i.e., simple proposition pairs—recall that ASB describes both simple and conditional propositions as simple) when an LR favouring H p has resulted from a compound proposition pair incorporating multiple POIs under H p and none of the POIs under H a , in order to establish the weighting and the consequent probative value of the evidence per contributor under H p . The conclusions of this work very strongly support this statement. Compound proposition pairs can misrepresent the weight of the evidence against an individual strongly in either direction. This work strongly favours the use of conditional proposition pairs rather than simple proposition pairs whenever the data support the presence of an individual as the conditioning profile since this increases the ability to differentiate true from false donors. We have demonstrated by calculating conditional LR s in DBLR™ that the compound LR can be obtained as a product of simple and conditional LR s. This is also approximately true using LR s produced by STRmix™. The use of conditional LR s, described as an approximation to the exhaustive LR by Buckleton et al. , resulted in higher LR s for the known contributors and lower LR s for the non-donors than when using a simple proposition pair. This statistic makes the best use of the DNA profiling information.
|
Responses of Soil Collembolans to Land Degradation in a Black Soil Region in China | a7585529-4be0-4b78-95ef-c5a296674102 | 10048822 | Microbiology[mh] | Land degradation, a global issue affecting billions of people, is a phenomenon caused by a combination of natural forces and artificial activities . Erosion by wind or water, salination, deforestation, high farming intensity and the overuse of pesticides and fertilizers are the main causes of land degradation . Land degradation has many effects on the planet, with such effects felt all over the world. The main effects of land degradation can be soil erosion, salinization, acidification, and alkalization of the land, and finally desertification, and these can cause the lack of farmable land and the decline in land productivity . Therefore, it is necessary to provide some insights into the mechanisms underlying degenerated soil, in order to prevent or reverse the effects of land degradation. Land degradation can negatively impact the physical, chemical and biological quality of soil, with these effects today considered pervasive and systemic . Generally, a visible effect of land degradation is decreases in soil thickness due to accelerated erosion . With the thinning of soil, the repercussions of land degradation for soil environments start systematically from the soil’s physical properties. A previous study reveal that negative influences may have been observed in the formation and stability of soil aggregates along land-degradation gradients, and that the soil structure may have deteriorated . After this, generally, soil nutrient stocks are changed and nutrient depletion is observed, and thus the amount, distribution and turnover rate of soil may be substantially influenced. Previous studies have demonstrated that land degradation may lead to a decline in soil C stocks and a high loss of soil N, as well as aggravating P deficiency [ , , ]. These soil environmental changes can significantly affect the ecological processes and ecosystem multifunctionality of belowground systems, and thus soil organisms are also impacted. In this context, some previous studies have found that land degradation may significantly decrease the diversity of soil microbial communities . Soil fauna is a critical component of soil organisms and has an irreplaceable role in the soil food web . Nevertheless, a relatively small number of studies have investigated the effects of land degradation on soil fauna. Collembolans (springtails), which inhabit soil and litter, belong to the three lineages of modern hexapods, and are the key consumers and decomposers in the terrestrial ecosystem . Having a large population and a wide variety of species, they are one of the main components of soil fauna . As they have different feeding guilds, Collembolans are found in various ecological niches, and these can directly cause them to act on ecological processes and ecosystem multifunctionality, thus playing crucial roles in belowground systems . At the same time, their miniscule body size, fragile body strength and poor ability to migrate make Collembolans sensitive to changes in soil environments . Once environmental factors lead to new microchanges, some significant responses are found in the taxonomic composition, community structure, diversity characteristics and distribution patterns of Collembolans. For instance, the restoration of vegetation can increase the abundance of soil Collembolans and cause their taxonomic composition to vary according to the different stages of vegetation restoration . A previous study found a systematic shift in the composition traits of Collembolan communities across different land uses . Consequently, Collembolans can adequately indicate environmental changes in belowground systems. It is an indisputable fact that land degradation can change the environment of belowground systems, thereby significantly altering soil properties. This inevitably affects soil Collembolan communities. However, there is currently a knowledge gap in the literature regarding the responses of soil Collembolans to land degradation. Black soil, or mollisol in the USDA soil taxonomy, is characterized by its dark color and rich texture . Due to its abundance of organic matter and nutrient elements, this type of soil is essential for the agricultural industry and has a variety of uses. Black soil can be found in many parts of the world, mainly the Central United States, Ukraine, Northeast China and Argentina. Songnen Plain, a typical black soil area, is one of the most important grain-producing regions in China. Recently, land degradation has been rapidly increasing in the Songnen Plain because of its flood-prone landform, high-intensity agricultural activities and non-conservation tillage system . As a result, its soil environment is getting considerably worse. In light of this background, in order to better understand the responses of soil Collembolans to land degradation in a black soil region, we selected the Songnen Plain as the experimental site. Soil Collembolans were collected from four degrees of land-degradation habitats. We hypothesized that (H1) the changes in the soil Collembolan communities would be consistent with the land degradation in the black soil region, and (H2) that the soil Collembolan responses to the land degradation would differ across each of their taxa.
2.1. Site Description In this study, the investigation was conducted in the northern part of the Songnen Plain (47°17′ N–47°54′ N, 125°92′ E–126°10′ E), where a great deal of black soil is widely distributed. This region is situated in the Xiao Hinggan Mountains and Songnen Plain intersection, and low hills account for a large proportion of its area. It is characterized by a continental monsoon climate, with long cold winters and short cool summers, and has a mean annual average temperature of 1.2 °C and mean annual precipitation of 490 mm. The soil type is mollisol, which is neutral or slightly acidic, with rich humus. Maize is most widely cultivated in the non-conservation tillage system used in the region, and thus land degradation is very common in the area. 2.2. Sampling Design To evaluate the responses of soil Collembolans to land degradation in a black soil region, habitats with different degrees of land degradation were selected among maize planting fields. Land degradation was divided into four degrees following the classification method of Sun et al. . The habitats consisted of a no land-degradation habitat (NLD), light land-degradation habitat (LLD), moderate land-degradation habitat (MLD) and severe land-degradation habitat (SLD). Soil thickness was measured in the field, and the yield was monitored after harvest. The ecological parameters of each are shown in . In this study, we collected samples in the spring (May), summer (July) and autumn (September) of 2021. Three independent stands (300 m 2 ) and five replicate sample plots (1 m 2 ) were randomly established. Soil core samples (25 × 25 × 20 cm) were taken in order to collect Collembolans in each plot. A total of 180 soil core samples (4 degrees × 3 replicated stands × 5 replicated plots × 3 sampling periods) were obtained. At the same time, soil samples were collected using soil augers in each plot. 2.3. Soil Collembolan Extraction and Identification Soil core samples were extracted over 96 h via Tullgren funnel extractors. All of the Collembolan samples were preserved in 75% alcohol. After being cleared using Nesbitt’s fluid and mounted using Hoyer’s medium, soil Collembolan samples were made into slide specimens. Collembolans were counted using a Nikon SMZ745T stereoscopic microscope (Nikon, Tokyo, Japan) and a Nikon E200 biomicroscope (Nikon, Tokyo, Japan), and Collembolan specimens were identified at a species level [ , , ]. 2.4. Soil Analysis After litters, roots and gravels were removed from the soil samples, each sample was air-dried and stored at room temperature. Soil properties were measured using conventional standards . Briefly, K 2 Cr 2 O 7 –H 2 SO 4 digestion and FeSO 4 titration were applied to determine soil organic matter (SOM); the total nitrogen (TN) and available phosphorus (AP) values of the soil were measured using a SEAL AA3 auto analyzer (SEAL, Mequon, WI, USA); the pipette method was applied to measure contents of soil clay particles and sand grains. 2.5. Statistical Analysis A Venn diagram was created to illustrate the differences in soil Collembolan taxonomic composition within the four land-degradation degrees, with endemic species being manually screened and a diagram being created using the “VennDiagram” R package . To analyze the effects of land-degradation degrees on the soil Collembolan community complexity, a network analysis was performed based on Spearman’s rank correlation matrix, using Cytoscape 3.9.1 . To compare the differences in the soil Collembolan abundance (individuals per m −2 ) caused by land degradation, the Tukey’s HSD test was conducted using the “car” R package . To test the effects of land-degradation degrees on soil Collembolan richness and diversity, rarefaction curves were created using EstimateS 9.1 . A heatmap of the hierarchical clustering was created to evaluate the distribution patterns of soil Collembolan species among the four land-degradation habitats, using the “pheatmap” R package . The hierarchical clustering in the heatmap was based on the unweighted pair-group method with a mathematical mean algorithm (UPGMA), and the cluster analysis was conducted using the “stats” R package . In addition, Pearson’s correlations were calculated to reveal the relationships between each soil Collembolan genus and different environmental variables via the “corrplot” R package . In order to determine direct and indirect interaction effects between independent and measured variables in a single model, structural equation models (SEMs) were employed using the “lavaan” R package .
In this study, the investigation was conducted in the northern part of the Songnen Plain (47°17′ N–47°54′ N, 125°92′ E–126°10′ E), where a great deal of black soil is widely distributed. This region is situated in the Xiao Hinggan Mountains and Songnen Plain intersection, and low hills account for a large proportion of its area. It is characterized by a continental monsoon climate, with long cold winters and short cool summers, and has a mean annual average temperature of 1.2 °C and mean annual precipitation of 490 mm. The soil type is mollisol, which is neutral or slightly acidic, with rich humus. Maize is most widely cultivated in the non-conservation tillage system used in the region, and thus land degradation is very common in the area.
To evaluate the responses of soil Collembolans to land degradation in a black soil region, habitats with different degrees of land degradation were selected among maize planting fields. Land degradation was divided into four degrees following the classification method of Sun et al. . The habitats consisted of a no land-degradation habitat (NLD), light land-degradation habitat (LLD), moderate land-degradation habitat (MLD) and severe land-degradation habitat (SLD). Soil thickness was measured in the field, and the yield was monitored after harvest. The ecological parameters of each are shown in . In this study, we collected samples in the spring (May), summer (July) and autumn (September) of 2021. Three independent stands (300 m 2 ) and five replicate sample plots (1 m 2 ) were randomly established. Soil core samples (25 × 25 × 20 cm) were taken in order to collect Collembolans in each plot. A total of 180 soil core samples (4 degrees × 3 replicated stands × 5 replicated plots × 3 sampling periods) were obtained. At the same time, soil samples were collected using soil augers in each plot.
Soil core samples were extracted over 96 h via Tullgren funnel extractors. All of the Collembolan samples were preserved in 75% alcohol. After being cleared using Nesbitt’s fluid and mounted using Hoyer’s medium, soil Collembolan samples were made into slide specimens. Collembolans were counted using a Nikon SMZ745T stereoscopic microscope (Nikon, Tokyo, Japan) and a Nikon E200 biomicroscope (Nikon, Tokyo, Japan), and Collembolan specimens were identified at a species level [ , , ].
After litters, roots and gravels were removed from the soil samples, each sample was air-dried and stored at room temperature. Soil properties were measured using conventional standards . Briefly, K 2 Cr 2 O 7 –H 2 SO 4 digestion and FeSO 4 titration were applied to determine soil organic matter (SOM); the total nitrogen (TN) and available phosphorus (AP) values of the soil were measured using a SEAL AA3 auto analyzer (SEAL, Mequon, WI, USA); the pipette method was applied to measure contents of soil clay particles and sand grains.
A Venn diagram was created to illustrate the differences in soil Collembolan taxonomic composition within the four land-degradation degrees, with endemic species being manually screened and a diagram being created using the “VennDiagram” R package . To analyze the effects of land-degradation degrees on the soil Collembolan community complexity, a network analysis was performed based on Spearman’s rank correlation matrix, using Cytoscape 3.9.1 . To compare the differences in the soil Collembolan abundance (individuals per m −2 ) caused by land degradation, the Tukey’s HSD test was conducted using the “car” R package . To test the effects of land-degradation degrees on soil Collembolan richness and diversity, rarefaction curves were created using EstimateS 9.1 . A heatmap of the hierarchical clustering was created to evaluate the distribution patterns of soil Collembolan species among the four land-degradation habitats, using the “pheatmap” R package . The hierarchical clustering in the heatmap was based on the unweighted pair-group method with a mathematical mean algorithm (UPGMA), and the cluster analysis was conducted using the “stats” R package . In addition, Pearson’s correlations were calculated to reveal the relationships between each soil Collembolan genus and different environmental variables via the “corrplot” R package . In order to determine direct and indirect interaction effects between independent and measured variables in a single model, structural equation models (SEMs) were employed using the “lavaan” R package .
3.1. Taxonomic Composition and Community Structure During the study period, 10,341 individual soil Collembolans were collected from the four land-degradation habitats: these belonged to 30 species, 26 genera, 12 families and three orders ( ). In total, the dominant species were found to be Proisotoma minima (30.20%) and Isotomiella minor (13.46%), and the common species included Parisotoma dichaeta (8.89%), Proisotoma sp.1 (8.89%), Ceratophysella sp.1 (5.81%), Folsomia candida (5.42%), Bionychiurus sp.1 (5.24%), Protaphorura sp.1 (4.18%), Micronychiurus sp.1 (3.89%), Homidia quadrimaculata (3.47%), Xenylla sp.1 (1.91%), Metaphorura affinis (1.26%) and Tomocerus kinoshitai (1.12%). Additionally, the residual 17 species were rare species, accounting for 6.28% of the total number of individuals. At the species level, there were both similar dominant species and different taxonomic compositions among the four land-degradation habitats ( A). Proisotoma minima were always a dominant species during the study period. Isotomiella minor were observed to be a dominant species in the no land-degradation habitat (NLD), light land-degradation habitat (LLD) and moderate land-degradation habitat (MLD), whereas they were a common species (4.32%) in the severe land-degradation habitat (SLD). Parisotoma dichaeta were only a dominant species in the LLD (12.64%) and SLD (10.07%). At the same time, Proisotoma sp.1 were only a dominant species in the NLD (11.15%) and SLD (17.45%). Ceratophysella sp.1 was a dominant species (12.34%) only in the MLD. Across the different land-degradation habitats, both unique and shared species of soil Collembolans were found, as illustrated in the Venn diagram shown in B. All four habitats had 16 species ( Proisotoma minima , Isotomiella minor , Parisotoma dichaeta , Proisotoma sp.1, Ceratophysella sp.1, Folsomia candida , Bionychiurus sp.1, Protaphorura sp.1, Micronychiurus sp.1, Homidia quadrimaculata , Xenylla sp.1, Metaphorura affinis , Desoria spatiosa , Anurida assimilis , Desoria pseudomaritima and Sminthurides sp.1) in common. Pseudosinella sp.1 was only collected in the LLD and MLD, while Friesea sp.1 was only found in the LLD and SLD. Bourletiella sp.1 was the only unique species found in the NLD. The community structure and key topological features of the soil Collembolans were summarized using network analysis ( ). It was found that community complexity declined as the degree of land degradation increased; among the habitats, this was most obvious in the SLD. At the same time, we found that various interspecific relationships existed between each habitat. For instance, Proisotoma minima was negatively correlated with a majority of the species of soil Collembolans found in the habitats with low levels of land degradation (NLD and LLD), yet it was positively correlated with most of the other species in the habitats with high levels of land degradation (MLD and SLD). 3.2. Diversity Characteristics and Distribution Patterns During the study period, 337,300 m −2 (27 species) Collembolan specimens were collected in the NLD; 333,200 m −2 (29 species) in the LLD; 308,000 m −2 (28 species) in the MLD; and 55,600 m −2 (18 species) in the SLD. Among these habitats, the NLD had the most abundance, and the LLD displayed the most richness. In contrast, the SLD had the least abundance and richness. During the different seasons, the diversity characteristics of the soil Collembolans differed across the four habitats ( ). The abundance data of the soil Collembolans are shown in A–C. The abundance of the Collembolans from the SLD was consistently at the lowest levels during the study period, with these levels being significant in summer and autumn ( p < 0.05). During spring, significant differences were not found across all the four habitats ( p > 0.05), and the highest abundance was observed in the MLD. In summer, there were no significant differences among the NLD, LLD and MLD ( p > 0.05), and the MLD again had the greatest abundance. During autumn, the abundance of the NLD was the highest, and was significantly greater than the abundance of the MLD and SLD ( p < 0.05). The rarefaction curves of the soil Collembolan richness and diversity are shown in D–I. All the curves approached plateaus, and this indicated that the majority of the soil Collembolan species were collected in these areas. The richness and diversity of the soil Collembolans varied across the different land-degradation habitats. In spring and summer, the richness and diversity of the SLD were both obviously lower than the richness and diversity of other habitats, whereas distinctions in the richness and diversity of the habitats were not clear during autumn. The heat map shown in illustrates the distribution patterns of the soil Collembolans from the four land-degradation habitats during the study period. These four land degradation habitats were divisible into three clusters, specifically as follows: major similarities existed among the spring MLD, autumn NLD and LLD; most of the habitats (except the SLD) in summer and the MLD in autumn were divisible into a cluster; the residue sites were similar to each other. At the same time, the soil Collembolan communities were also divisible into three clusters. Proisotoma minima was only able to make up one cluster. Isotomiella minor , Proisotoma sp.1 and Parisotoma dichaeta formed another cluster. Finally, other species were divisible into a third cluster, and this indicated that the majority of Collembolan species were distributed relatively evenly. 3.3. Correlations of Soil Collembolans with Land Degradation As illustrated by Pearson’s correlations, different environmental variables were correlated with each genus of soil Collembolans in numerous ways ( ). Pseudachorutes , Bionychiurus , Micronychiurus , Protaphorura , Metaphorura , Tomocerus , Desoria and Isotomiella were significantly correlated with the majority of the soil environmental variables caused by land degradation ( p < 0.05). This revealed that these genera might exhibit relatively sensitive responses to land degradation. At the same time, we found that most of the genera were significantly negatively correlated with the contents of soil sand grains ( p < 0.05). This indicated that land sandification might be a major limiting factor for soil Collembolans. In addition, soil available P was found not to be correlated with most of the genera, except Arrhopalites , which had a significantly positive correlation with soil available P ( p < 0.05). A structural equation model (SEM) was designed to evaluate the correlations among the land degradation (soil thickness and land productivity), soil fertility (soil organic matter, total N, available P and available K), soil texture (clay particles and sand grains) and soil Collembolan communities (abundance and richness) ( ). The final model provided an excellent fit (χ 2 /df = 2.145, p = 0.365, GFI = 0.907, CFI = 0.919 and RMSEA = 0.017). The SEM found that land degradation significantly negatively affected the soil fertility (regression weight = −0.8877) and soil texture (regression weight = −0.8877). At the same time, the soil Collembolan communities were positively influenced by the soil fertility (regression weight = 0.2521) and soil texture (regression weight = 0.1991). Therefore, this model indicated that land degradation may have impacted the soil Collembolan communities by affecting the soil fertility and soil texture.
During the study period, 10,341 individual soil Collembolans were collected from the four land-degradation habitats: these belonged to 30 species, 26 genera, 12 families and three orders ( ). In total, the dominant species were found to be Proisotoma minima (30.20%) and Isotomiella minor (13.46%), and the common species included Parisotoma dichaeta (8.89%), Proisotoma sp.1 (8.89%), Ceratophysella sp.1 (5.81%), Folsomia candida (5.42%), Bionychiurus sp.1 (5.24%), Protaphorura sp.1 (4.18%), Micronychiurus sp.1 (3.89%), Homidia quadrimaculata (3.47%), Xenylla sp.1 (1.91%), Metaphorura affinis (1.26%) and Tomocerus kinoshitai (1.12%). Additionally, the residual 17 species were rare species, accounting for 6.28% of the total number of individuals. At the species level, there were both similar dominant species and different taxonomic compositions among the four land-degradation habitats ( A). Proisotoma minima were always a dominant species during the study period. Isotomiella minor were observed to be a dominant species in the no land-degradation habitat (NLD), light land-degradation habitat (LLD) and moderate land-degradation habitat (MLD), whereas they were a common species (4.32%) in the severe land-degradation habitat (SLD). Parisotoma dichaeta were only a dominant species in the LLD (12.64%) and SLD (10.07%). At the same time, Proisotoma sp.1 were only a dominant species in the NLD (11.15%) and SLD (17.45%). Ceratophysella sp.1 was a dominant species (12.34%) only in the MLD. Across the different land-degradation habitats, both unique and shared species of soil Collembolans were found, as illustrated in the Venn diagram shown in B. All four habitats had 16 species ( Proisotoma minima , Isotomiella minor , Parisotoma dichaeta , Proisotoma sp.1, Ceratophysella sp.1, Folsomia candida , Bionychiurus sp.1, Protaphorura sp.1, Micronychiurus sp.1, Homidia quadrimaculata , Xenylla sp.1, Metaphorura affinis , Desoria spatiosa , Anurida assimilis , Desoria pseudomaritima and Sminthurides sp.1) in common. Pseudosinella sp.1 was only collected in the LLD and MLD, while Friesea sp.1 was only found in the LLD and SLD. Bourletiella sp.1 was the only unique species found in the NLD. The community structure and key topological features of the soil Collembolans were summarized using network analysis ( ). It was found that community complexity declined as the degree of land degradation increased; among the habitats, this was most obvious in the SLD. At the same time, we found that various interspecific relationships existed between each habitat. For instance, Proisotoma minima was negatively correlated with a majority of the species of soil Collembolans found in the habitats with low levels of land degradation (NLD and LLD), yet it was positively correlated with most of the other species in the habitats with high levels of land degradation (MLD and SLD).
During the study period, 337,300 m −2 (27 species) Collembolan specimens were collected in the NLD; 333,200 m −2 (29 species) in the LLD; 308,000 m −2 (28 species) in the MLD; and 55,600 m −2 (18 species) in the SLD. Among these habitats, the NLD had the most abundance, and the LLD displayed the most richness. In contrast, the SLD had the least abundance and richness. During the different seasons, the diversity characteristics of the soil Collembolans differed across the four habitats ( ). The abundance data of the soil Collembolans are shown in A–C. The abundance of the Collembolans from the SLD was consistently at the lowest levels during the study period, with these levels being significant in summer and autumn ( p < 0.05). During spring, significant differences were not found across all the four habitats ( p > 0.05), and the highest abundance was observed in the MLD. In summer, there were no significant differences among the NLD, LLD and MLD ( p > 0.05), and the MLD again had the greatest abundance. During autumn, the abundance of the NLD was the highest, and was significantly greater than the abundance of the MLD and SLD ( p < 0.05). The rarefaction curves of the soil Collembolan richness and diversity are shown in D–I. All the curves approached plateaus, and this indicated that the majority of the soil Collembolan species were collected in these areas. The richness and diversity of the soil Collembolans varied across the different land-degradation habitats. In spring and summer, the richness and diversity of the SLD were both obviously lower than the richness and diversity of other habitats, whereas distinctions in the richness and diversity of the habitats were not clear during autumn. The heat map shown in illustrates the distribution patterns of the soil Collembolans from the four land-degradation habitats during the study period. These four land degradation habitats were divisible into three clusters, specifically as follows: major similarities existed among the spring MLD, autumn NLD and LLD; most of the habitats (except the SLD) in summer and the MLD in autumn were divisible into a cluster; the residue sites were similar to each other. At the same time, the soil Collembolan communities were also divisible into three clusters. Proisotoma minima was only able to make up one cluster. Isotomiella minor , Proisotoma sp.1 and Parisotoma dichaeta formed another cluster. Finally, other species were divisible into a third cluster, and this indicated that the majority of Collembolan species were distributed relatively evenly.
As illustrated by Pearson’s correlations, different environmental variables were correlated with each genus of soil Collembolans in numerous ways ( ). Pseudachorutes , Bionychiurus , Micronychiurus , Protaphorura , Metaphorura , Tomocerus , Desoria and Isotomiella were significantly correlated with the majority of the soil environmental variables caused by land degradation ( p < 0.05). This revealed that these genera might exhibit relatively sensitive responses to land degradation. At the same time, we found that most of the genera were significantly negatively correlated with the contents of soil sand grains ( p < 0.05). This indicated that land sandification might be a major limiting factor for soil Collembolans. In addition, soil available P was found not to be correlated with most of the genera, except Arrhopalites , which had a significantly positive correlation with soil available P ( p < 0.05). A structural equation model (SEM) was designed to evaluate the correlations among the land degradation (soil thickness and land productivity), soil fertility (soil organic matter, total N, available P and available K), soil texture (clay particles and sand grains) and soil Collembolan communities (abundance and richness) ( ). The final model provided an excellent fit (χ 2 /df = 2.145, p = 0.365, GFI = 0.907, CFI = 0.919 and RMSEA = 0.017). The SEM found that land degradation significantly negatively affected the soil fertility (regression weight = −0.8877) and soil texture (regression weight = −0.8877). At the same time, the soil Collembolan communities were positively influenced by the soil fertility (regression weight = 0.2521) and soil texture (regression weight = 0.1991). Therefore, this model indicated that land degradation may have impacted the soil Collembolan communities by affecting the soil fertility and soil texture.
4.1. Effects of Land Degradation on Soil Collembolan Communities In this study, we observed that the soil Collembolan specimens from the severe land-degradation habitat (SLD), whether regarding their abundance, richness, diversity or community complexity, were always at the lowest levels compared to the other land-degradation habitats (NLD, LLD and MLD). Further, the community complexity found among the different habitats was consistent with the land degradation. This indicated that land degradation may have limited the soil Collembolan communities in the black soil study region. These findings were in line with our hypothesis that changes in soil Collembolan communities would be consistent with the land degradation present in the black soil study region (H1), and agreed with the previous study of soil bacterial communities . Previous studies revealed that land surface organic matter could provide abundant food resources for Collembolans, and that it was the key factor that directly affects soil Collembolan communities . In this study, we observed that the land degradation led to visible declines in the contents of soil organic matter ( ). Consequently, the Collembolan living environment was short on food resources, which resulted in a deficiency of the Collembolans in the SLD. Additionally, a previous study revealed that Collembolans partially breathe through their skin, and thus need a moist environment . In the SLD, we found that the proportion of sand grains was greater than that of other habitats, which might have rapidly increased the soil water runoff and created relatively drier conditions. As a result, the soil Collembolan communities were further restricted in this severe land-degradation habitat. Further, in this study, 27 species of Collembolans were collected in the NLD, 29 species in the LLD and 28 species in the MLD. These findings were partially out of line with our first hypothesis (H1). The “intermediate disturbance hypothesis” predicts that maximum levels of biodiversity should be observed under some intermediate disturbance frequency in ecological communities, which permits more species invasions and colonization . Compared with the other land degradation habitats, the LLD habitat was at the intermediate disturbance level, which may have promoted the prosperity of the species of soil Collembolans. At the same time, the Venn diagram and heat map constructed in the study illustrated that the majority of the Collembolans were distributed relatively evenly, indicating that land degradation did not cause the obvious change in fauna composition observed in the study region ( B and ). Previous studies demonstrated that the fauna composition of soil Collembolans was dependent on environmental conditions on the regional scale . Since all the sites included in this study were located in the Songnen Plain, they had the same environmental conditions. Consequently, similarities across the soil Collembolans were observed. 4.2. Responses of Soil Collembolans to Land Degradation We found that different soil environmental variables were correlated with each genus of soil Collembolans in numerous ways, and this confirmed our hypothesis that responses to land degradation would differ across the soil Collembolan taxa (H2). Previous studies revealed that life histories, feeding guilds, propagation characteristics and adaptability mechanisms differed dramatically across soil Collembolan taxa [ , , ]. Consequently, in this study, a variety of responses to land degradation were observed across the different genera of soil Collembolans. At the same time, we found that Pseudachorutes , Bionychiurus , Micronychiurus , Protaphorura , Metaphorura , Tomocerus , Desoria and Isotomiella were significantly correlated with the majority of the soil environmental variables caused by land degradation ( p < 0.05) ( ). Soil Collembolans could be categorized into three life forms: euedaphic (subsurface active); hemiedaphic (within the soil and partly on the surface) and epedaphic (surface-active) [ , , ]. In this study, Bionychiurus , Micronychiurus , Protaphorura and Metaphorura are examples of euedaphic Collembolans, which prefer to live in deep soils. Tomocerus , Desoria and Isotomiella are examples of epedaphic Collembolans, which prefer to live on the soil surface. Meanwhile, hemiedaphic Collembolans, which were intermediate dwellers, exhibit strong endurance to environmental changes. Consequently, in this study, the euedaphic and epedaphic Collembolans exhibited relatively sensitive responses to land degradation. In addition, Bourletiella sp.1 was the only unique species found in the NLD. Bourletiella , which belongs to the order Symphypleona, are typical herbivorous and epedaphic Collembolans . Herbivorous Collembolans take fresh organic matter as their main food. Abundant soil nutrients may produce plenty of fresh organic matter, and thus habitats with low levels of land degradation may be beneficial to Symphypleona. In this study, we determined that Proisotoma minima , a dominant species in all the habitats, was negatively correlated with the majority of soil Collembolan species in the habitats with low levels of land degradation (NLD and LLD), yet they were positively correlated with most of the other species in the habitats with high levels of land degradation (MLD and SLD) ( ). Proisotoma , primary decomposers that feed on litter and some fungi, had an abundant population, and were thus located in a relatively low niche . In the habitats with low levels of land degradation, the environmental conditions were relatively better than that of other habitats; consequently, a large number of Proisotoma minima were able to occupy the living spaces of species with identical niches. As a result, they were negatively correlated with other species of soil Collembolans. In the habitats with high levels of land degradation, the environmental conditions were relatively worse, which might have released some living space for Collembolans with identical niches, and thus this might have led to the positive correlation found between Proisotoma minima and other species. Consequently, the relationship between Proisotoma minima and other Collembolans indicated that land degradation was present in the black soil study region. We found that soil Collembolan communities might have responded negatively to land degradation, because the land degradation changed the soil fertility and soil texture ( ). Land degradation could cause a great deal of soil clay to leach, and it radically reduces the ability to preserve soil and water . As a result, considerable organic matter left belowground ecosystems as root exudates and litter, which led to sequential declines in soil fertility . Soil fertility is vitally important in supplying sufficient food for Collembolans, while loam is crucial in providing relatively wetter conditions at the same time . Consequently, we found that the soil Collembolans responded negatively to land degradation in the study area. In addition, this study still has a lot of room for improvement in the responses of soil Collembolan to land degradation. The morphological characteristics of soil Collembolan have been confirmed to diversify with environmental factors, and thus these characteristics are functional . Consequently, functional traits associated with changes in soil Collembolan communities may indicate the occurrence of land degradation. However, there is a relatively small number of studies available at present. Therefore, in the future, it is necessary to strengthen the studies on the functional traits of soil Collembolans responding to land degradation, and finally put forward a deeper explanation of the relationships between soil Collembolans and environmental changes.
In this study, we observed that the soil Collembolan specimens from the severe land-degradation habitat (SLD), whether regarding their abundance, richness, diversity or community complexity, were always at the lowest levels compared to the other land-degradation habitats (NLD, LLD and MLD). Further, the community complexity found among the different habitats was consistent with the land degradation. This indicated that land degradation may have limited the soil Collembolan communities in the black soil study region. These findings were in line with our hypothesis that changes in soil Collembolan communities would be consistent with the land degradation present in the black soil study region (H1), and agreed with the previous study of soil bacterial communities . Previous studies revealed that land surface organic matter could provide abundant food resources for Collembolans, and that it was the key factor that directly affects soil Collembolan communities . In this study, we observed that the land degradation led to visible declines in the contents of soil organic matter ( ). Consequently, the Collembolan living environment was short on food resources, which resulted in a deficiency of the Collembolans in the SLD. Additionally, a previous study revealed that Collembolans partially breathe through their skin, and thus need a moist environment . In the SLD, we found that the proportion of sand grains was greater than that of other habitats, which might have rapidly increased the soil water runoff and created relatively drier conditions. As a result, the soil Collembolan communities were further restricted in this severe land-degradation habitat. Further, in this study, 27 species of Collembolans were collected in the NLD, 29 species in the LLD and 28 species in the MLD. These findings were partially out of line with our first hypothesis (H1). The “intermediate disturbance hypothesis” predicts that maximum levels of biodiversity should be observed under some intermediate disturbance frequency in ecological communities, which permits more species invasions and colonization . Compared with the other land degradation habitats, the LLD habitat was at the intermediate disturbance level, which may have promoted the prosperity of the species of soil Collembolans. At the same time, the Venn diagram and heat map constructed in the study illustrated that the majority of the Collembolans were distributed relatively evenly, indicating that land degradation did not cause the obvious change in fauna composition observed in the study region ( B and ). Previous studies demonstrated that the fauna composition of soil Collembolans was dependent on environmental conditions on the regional scale . Since all the sites included in this study were located in the Songnen Plain, they had the same environmental conditions. Consequently, similarities across the soil Collembolans were observed.
We found that different soil environmental variables were correlated with each genus of soil Collembolans in numerous ways, and this confirmed our hypothesis that responses to land degradation would differ across the soil Collembolan taxa (H2). Previous studies revealed that life histories, feeding guilds, propagation characteristics and adaptability mechanisms differed dramatically across soil Collembolan taxa [ , , ]. Consequently, in this study, a variety of responses to land degradation were observed across the different genera of soil Collembolans. At the same time, we found that Pseudachorutes , Bionychiurus , Micronychiurus , Protaphorura , Metaphorura , Tomocerus , Desoria and Isotomiella were significantly correlated with the majority of the soil environmental variables caused by land degradation ( p < 0.05) ( ). Soil Collembolans could be categorized into three life forms: euedaphic (subsurface active); hemiedaphic (within the soil and partly on the surface) and epedaphic (surface-active) [ , , ]. In this study, Bionychiurus , Micronychiurus , Protaphorura and Metaphorura are examples of euedaphic Collembolans, which prefer to live in deep soils. Tomocerus , Desoria and Isotomiella are examples of epedaphic Collembolans, which prefer to live on the soil surface. Meanwhile, hemiedaphic Collembolans, which were intermediate dwellers, exhibit strong endurance to environmental changes. Consequently, in this study, the euedaphic and epedaphic Collembolans exhibited relatively sensitive responses to land degradation. In addition, Bourletiella sp.1 was the only unique species found in the NLD. Bourletiella , which belongs to the order Symphypleona, are typical herbivorous and epedaphic Collembolans . Herbivorous Collembolans take fresh organic matter as their main food. Abundant soil nutrients may produce plenty of fresh organic matter, and thus habitats with low levels of land degradation may be beneficial to Symphypleona. In this study, we determined that Proisotoma minima , a dominant species in all the habitats, was negatively correlated with the majority of soil Collembolan species in the habitats with low levels of land degradation (NLD and LLD), yet they were positively correlated with most of the other species in the habitats with high levels of land degradation (MLD and SLD) ( ). Proisotoma , primary decomposers that feed on litter and some fungi, had an abundant population, and were thus located in a relatively low niche . In the habitats with low levels of land degradation, the environmental conditions were relatively better than that of other habitats; consequently, a large number of Proisotoma minima were able to occupy the living spaces of species with identical niches. As a result, they were negatively correlated with other species of soil Collembolans. In the habitats with high levels of land degradation, the environmental conditions were relatively worse, which might have released some living space for Collembolans with identical niches, and thus this might have led to the positive correlation found between Proisotoma minima and other species. Consequently, the relationship between Proisotoma minima and other Collembolans indicated that land degradation was present in the black soil study region. We found that soil Collembolan communities might have responded negatively to land degradation, because the land degradation changed the soil fertility and soil texture ( ). Land degradation could cause a great deal of soil clay to leach, and it radically reduces the ability to preserve soil and water . As a result, considerable organic matter left belowground ecosystems as root exudates and litter, which led to sequential declines in soil fertility . Soil fertility is vitally important in supplying sufficient food for Collembolans, while loam is crucial in providing relatively wetter conditions at the same time . Consequently, we found that the soil Collembolans responded negatively to land degradation in the study area. In addition, this study still has a lot of room for improvement in the responses of soil Collembolan to land degradation. The morphological characteristics of soil Collembolan have been confirmed to diversify with environmental factors, and thus these characteristics are functional . Consequently, functional traits associated with changes in soil Collembolan communities may indicate the occurrence of land degradation. However, there is a relatively small number of studies available at present. Therefore, in the future, it is necessary to strengthen the studies on the functional traits of soil Collembolans responding to land degradation, and finally put forward a deeper explanation of the relationships between soil Collembolans and environmental changes.
In summary, in this study, it was found that the soil Collembolan communities may have exhibited negative responses to land degradation in the study region. The Collembolan specimens’ taxonomic composition varied across the different land-degradation habitats, which were located in a black soil region of the Songnen Plain, but the majority of the Collembolan species were distributed relatively evenly across this region. The abundance, richness, diversity and community complexity of the soil Collembolans from the severe land-degradation habitat were always at the lowest levels compared to the other habitats. Further, seasonal variations were observed in the abundance, richness and diversity levels of the soil Collembolans. Proisotoma minima was negatively correlated with a majority of the species of soil Collembolans in the habitats with low levels of land degradation, whereas they were positively correlated with most of the other species in the habitats with high levels of land degradation. Additionally, each genus of soil Collembolans responded to the land degradation in numerous ways, with the epedaphic and euedaphic Collembolans responding more obviously to it. The findings of this study have implications for the study of the relationship between land degradation and soil Collembolans, and can provide some assistance in developing biodiversity guidelines for farmland protection in black soil regions.
|
Co-Designing Communication: A Design Thinking Approach Applied to Radon Health Communication | 1867b130-2254-4728-abef-226623c5e5e1 | 10048842 | Health Communication[mh] | Health intervention planning models emphasize the importance of participatory methods, thus involving community members and other relevant stakeholders in the different planning stages, from problem definition to intervention implementation . Not only does this increase the external validity of the intervention by the acceptance and acknowledgment of the input provided by the community, but it also provides broad perspectives and skills from community members, stakeholders, and the design team. Using the collective creativity of professionals and the local community in designing an intervention is referred to as co-design and can be seen as a citizen science approach . Although multiple citizen science projects were conducted within the field of radon, co-design methods have, to our knowledge, not yet been adopted in intervention design . Radon is an indoor air pollutant. It is a natural radioactive gas that is present in the soil in varying concentrations depending on the composition of the ground. Radon is invisible and has no scent, there are no visible casualties due to the gas, and since it is a natural gas, there is no culprit to blame . In high-risk areas, radon can enter houses through cracks or different installation tubes in the foundations of buildings, and the gas can accumulate indoors. Radon concentrations are one of the leading causes of lung cancer . Despite current health interventions, research shows that testing and mitigation rates remain insufficient . This raises the question of whether the current interventions tackle the right barriers and provide the right facilitators. Research specifically focused on (mass) communication interventions regarding radon has observed multiple gaps in the communication strategies adopted in the past. For instance, statistical information in leaflets or news articles prevails . To address these gaps, an exploratory co-design study was developed to first focus on general barriers and facilitators to perform radon protective behaviors and second on the ideation and designing of communication interventions, together with people with personal experience with radon. In this way, community members co-design a communication intervention, making it more personally relevant and likely more effective . 2.1. Health Interventions to Address Radon Exposure Changing behavior requires change on different levels; the behavior change wheel identifies capability, opportunity, and motivation as the main sources of behavior. Motivation reflects the individual, opportunity reflects the individual’s environment, and capability reflects a combination of the two. For behavior change to be effective and durable, the three components should be addressed with different types of interventions that often stem from the policy level . Looking at the policy level regarding radon, Europe adapted the Basic Safety Standards in 2013 and included radon protection as well . In practice, all European Member States are legally required to develop and implement a radon action plan containing information on ways to decrease radon levels at homes and workplaces. In the United States, the Indoor Radon Abatement Act (IRAA) from 1988 requires that indoor radon levels be as low as outdoors . These legislations, however, are on the highest level (namely the European level and the National level of the United States). The responsibility lies with the countries/states and their interpretation of their responsibility and legislation. Some countries/states, for instance, Estonia, only inform people about radon and place the responsibility for behavioral actions on the individual , whereas other countries, for instance, Ireland and Belgium, take the initial steps to include more specific legislation . Multiple scholars state that legislation procedures in terms of housing code requirements (comparable to energy efficiency) might increase the uptake for radon testing and mitigating , as is the case in certain States in The United States, for instance, Pennsylvania . On a European level, Austria is considering similar measures . Other policy measures are mostly concerned with reducing the economic impact of the testing and mitigating procedure—for instance, incentivizing mitigations, offering subventions, or providing free tests . A city in Ireland experimented with providing digital radon monitors in the library to facilitate the need for these monitors without the costs of buying them . Other countries, such as Bulgaria and the Czech Republic, provide free tests, and yet other countries (e.g., Belgium) sell tests at lowered prices during the heating season. Subventions for mitigation are also country-dependent; for instance, Austria, Germany, and Sweden provide financial support to those carrying out mitigation works . No real evidence is available on whether the financial aspect matters to people. Interestingly, focus groups in Ireland show that people who performed mitigation perceived the costs as not too high as it was an investment in their health. At the same time, people who did not mitigate (but had high levels of radon) perceived the costs as too high and an important barrier . Despite the interventions and measures in place, the uptake of radon protective behavior remains insufficient . It remains unclear whether the interventions in place address the barriers people experience and whether they create the right facilitating conditions. Therefore, there is a need to explore in more depth what barriers and facilitators people experience regarding radon-protective behavior. As radon is a multi-level problem, not only do the situational and the environmental factors matter, the responsibility of actually performing testing and mitigating often still lies with the individual homeowners . So, while creating the right environment for them to act is needed, they still must be motivated to act. One way to increase motivation is through communication and persuasion. Communication occurs on different levels, including interpersonal communication (e.g., an individual talking about radon with their general practitioner), stakeholder communication (e.g., general practitioners that are informed about radon on a higher level), and mass media communication (e.g., press articles about radon). A recent systematic review that focused on mass media communication about radon shows that campaigns mostly aim to increase awareness, knowledge, risk perception, and perceived susceptibility using factual communication in the form of brochures or press articles. The focus is on providing people with information about the characteristics of radon and the (technical) solutions. Although informative leaflets can be effective, they assume the full rationality of the audience, where they act upon the information they receive. The literature on behavior change has shown that people often experience bounded rationality and that other aspects, such as relevance, biases, and emotions, play an important part in the process . Other messages such as fear appeals in videos showed increased intention to request more information , and direct phone calls and letters increased intention to test . Moreover, while these communication interventions have shown to be effective to some level (e.g., low degree of increase in testing behavior), the next step, namely mitigation, remains mainly unchanged , which identifies an additional gap. In particular, Hevey identified 17 steps of behavior, from becoming informed about radon to having confirmed mitigation . However, communication interventions rarely move along these steps. The precaution adoption process model is a theory based on the different stages of behavior, from being unaware of the problem to maintaining the problem. The theory emphasizes that different stages require different communication approaches. For instance, to move from the first stage (unaware) to the second stage (unengaged), media messages about the hazards are needed, while in progressing from the second stage to the third (undecided), testimonials and personal experiences are most effective. Further, to proceed from the third stage to the fourth (decided not to act) or to the fifth (decided to act), information about personal susceptibility, likelihood, and severity of radon exposure is effective. Detailed information about ways to perform the behavior, the costs, and the resources are mainly effective when moving from the fifth stage to the final stage (maintenance) . Overall, the systematic review showed a need for more personally relevant communication efforts, as the question remains whether and to what extent the current communication approaches tackle the right determinants at the right moment and are in line with the needs of the public . This unveils the need to inquire about the their preferences of the target group regarding radon-related communication. 2.2. Co-Design in Health Interventions on Radon To answer these questions, we need to engage in dialogue with the target group themselves and, even more so, involve them actively in developing communication tools. Participatory designs include various methods; however, the mean denominator is the active engagement of the public. Different levels exist within participatory designs, from providing information (one-way) to a discussion (two-way) and active participation (multiple ways), which is the highest level of involvement. The latter often results in participatory decision-making and co-design of new products, technologies, or health interventions . Within the existing research about the health issues related to radon, participatory designs or citizen science projects have been adopted previously . The main topic investigated in previous studies was the understanding of the lack of mitigating behavior, either through interviews (i.e., providing the information) or through discussing the topic in focus groups (i.e., discussion) . Citizen science projects were related to, for instance, raising awareness, radon mapping, or radon testing and mitigating . To our knowledge, ours is the first study applying active participation in the design process of a communication intervention in the context of radon. More specifically, our study was designed to involve residents and homeowners in understanding the lack of radon protective behaviors and related general barriers and facilitators and considering solutions regarding communication campaigns. To investigate these aspects, we opted for design thinking. This participatory design framework allows for opening up the problem and inviting people to think along to identify it and create solutions based on their first-hand experiences . It is a way of creative problem-solving that is human-centered and emphasizes observation, collaboration, and visualization of ideas. It emphasizes empathizing with the issue and the context of the issue, defining the exact problem and challenge, ideating ways to solve the challenge, and testing prototypes to do so . This method, both problem- and solution-oriented, can provide new insights into why people avoid radon protective behaviors, what they think the solution would be, and even what the solution should look like. To summarize, two questions are raised: first, what are the main barriers and facilitators to engaging in radon-protective behavior experienced by homeowners, and how are these addressed in current interventions, if at all? Second, how can the communication about radon be improved to be more relevant and engaging for the target group? Changing behavior requires change on different levels; the behavior change wheel identifies capability, opportunity, and motivation as the main sources of behavior. Motivation reflects the individual, opportunity reflects the individual’s environment, and capability reflects a combination of the two. For behavior change to be effective and durable, the three components should be addressed with different types of interventions that often stem from the policy level . Looking at the policy level regarding radon, Europe adapted the Basic Safety Standards in 2013 and included radon protection as well . In practice, all European Member States are legally required to develop and implement a radon action plan containing information on ways to decrease radon levels at homes and workplaces. In the United States, the Indoor Radon Abatement Act (IRAA) from 1988 requires that indoor radon levels be as low as outdoors . These legislations, however, are on the highest level (namely the European level and the National level of the United States). The responsibility lies with the countries/states and their interpretation of their responsibility and legislation. Some countries/states, for instance, Estonia, only inform people about radon and place the responsibility for behavioral actions on the individual , whereas other countries, for instance, Ireland and Belgium, take the initial steps to include more specific legislation . Multiple scholars state that legislation procedures in terms of housing code requirements (comparable to energy efficiency) might increase the uptake for radon testing and mitigating , as is the case in certain States in The United States, for instance, Pennsylvania . On a European level, Austria is considering similar measures . Other policy measures are mostly concerned with reducing the economic impact of the testing and mitigating procedure—for instance, incentivizing mitigations, offering subventions, or providing free tests . A city in Ireland experimented with providing digital radon monitors in the library to facilitate the need for these monitors without the costs of buying them . Other countries, such as Bulgaria and the Czech Republic, provide free tests, and yet other countries (e.g., Belgium) sell tests at lowered prices during the heating season. Subventions for mitigation are also country-dependent; for instance, Austria, Germany, and Sweden provide financial support to those carrying out mitigation works . No real evidence is available on whether the financial aspect matters to people. Interestingly, focus groups in Ireland show that people who performed mitigation perceived the costs as not too high as it was an investment in their health. At the same time, people who did not mitigate (but had high levels of radon) perceived the costs as too high and an important barrier . Despite the interventions and measures in place, the uptake of radon protective behavior remains insufficient . It remains unclear whether the interventions in place address the barriers people experience and whether they create the right facilitating conditions. Therefore, there is a need to explore in more depth what barriers and facilitators people experience regarding radon-protective behavior. As radon is a multi-level problem, not only do the situational and the environmental factors matter, the responsibility of actually performing testing and mitigating often still lies with the individual homeowners . So, while creating the right environment for them to act is needed, they still must be motivated to act. One way to increase motivation is through communication and persuasion. Communication occurs on different levels, including interpersonal communication (e.g., an individual talking about radon with their general practitioner), stakeholder communication (e.g., general practitioners that are informed about radon on a higher level), and mass media communication (e.g., press articles about radon). A recent systematic review that focused on mass media communication about radon shows that campaigns mostly aim to increase awareness, knowledge, risk perception, and perceived susceptibility using factual communication in the form of brochures or press articles. The focus is on providing people with information about the characteristics of radon and the (technical) solutions. Although informative leaflets can be effective, they assume the full rationality of the audience, where they act upon the information they receive. The literature on behavior change has shown that people often experience bounded rationality and that other aspects, such as relevance, biases, and emotions, play an important part in the process . Other messages such as fear appeals in videos showed increased intention to request more information , and direct phone calls and letters increased intention to test . Moreover, while these communication interventions have shown to be effective to some level (e.g., low degree of increase in testing behavior), the next step, namely mitigation, remains mainly unchanged , which identifies an additional gap. In particular, Hevey identified 17 steps of behavior, from becoming informed about radon to having confirmed mitigation . However, communication interventions rarely move along these steps. The precaution adoption process model is a theory based on the different stages of behavior, from being unaware of the problem to maintaining the problem. The theory emphasizes that different stages require different communication approaches. For instance, to move from the first stage (unaware) to the second stage (unengaged), media messages about the hazards are needed, while in progressing from the second stage to the third (undecided), testimonials and personal experiences are most effective. Further, to proceed from the third stage to the fourth (decided not to act) or to the fifth (decided to act), information about personal susceptibility, likelihood, and severity of radon exposure is effective. Detailed information about ways to perform the behavior, the costs, and the resources are mainly effective when moving from the fifth stage to the final stage (maintenance) . Overall, the systematic review showed a need for more personally relevant communication efforts, as the question remains whether and to what extent the current communication approaches tackle the right determinants at the right moment and are in line with the needs of the public . This unveils the need to inquire about the their preferences of the target group regarding radon-related communication. To answer these questions, we need to engage in dialogue with the target group themselves and, even more so, involve them actively in developing communication tools. Participatory designs include various methods; however, the mean denominator is the active engagement of the public. Different levels exist within participatory designs, from providing information (one-way) to a discussion (two-way) and active participation (multiple ways), which is the highest level of involvement. The latter often results in participatory decision-making and co-design of new products, technologies, or health interventions . Within the existing research about the health issues related to radon, participatory designs or citizen science projects have been adopted previously . The main topic investigated in previous studies was the understanding of the lack of mitigating behavior, either through interviews (i.e., providing the information) or through discussing the topic in focus groups (i.e., discussion) . Citizen science projects were related to, for instance, raising awareness, radon mapping, or radon testing and mitigating . To our knowledge, ours is the first study applying active participation in the design process of a communication intervention in the context of radon. More specifically, our study was designed to involve residents and homeowners in understanding the lack of radon protective behaviors and related general barriers and facilitators and considering solutions regarding communication campaigns. To investigate these aspects, we opted for design thinking. This participatory design framework allows for opening up the problem and inviting people to think along to identify it and create solutions based on their first-hand experiences . It is a way of creative problem-solving that is human-centered and emphasizes observation, collaboration, and visualization of ideas. It emphasizes empathizing with the issue and the context of the issue, defining the exact problem and challenge, ideating ways to solve the challenge, and testing prototypes to do so . This method, both problem- and solution-oriented, can provide new insights into why people avoid radon protective behaviors, what they think the solution would be, and even what the solution should look like. To summarize, two questions are raised: first, what are the main barriers and facilitators to engaging in radon-protective behavior experienced by homeowners, and how are these addressed in current interventions, if at all? Second, how can the communication about radon be improved to be more relevant and engaging for the target group? To apply the participatory design, we composed a research team comprising researchers from different disciplines, such as risk communication, health communication, sociology, nuclear physics, and citizen science. This ensured the avoidance of conceptual bias. Most researchers of the team had expertise with qualitative methods and radon research; however, none had operational expertise in design thinking as a research method. Therefore, the research protocol was developed in collaboration with a Belgian company specializing in design thinking (ACOMPANY). The company also provided a full training day of the method for all researchers involved in this study. 3.1. Participants The aim was to recruit participants who already had some experience with radon so that they could speak from their own experiences rather than a hypothetical scenario. This meant that we recruited people who had already measured (high) radon levels. 3.2. Workshop Design A workshop was designed that consisted of two unstructured group sessions. Each session lasted two hours and was scheduled a week apart. More specifically, the framework of the double diamond was applied to the context of radon and the workshop design itself . The first stage of this framework, as seen in , is the challenge, which is the starting point of the workshops and describes the ideal scenario. For this research project, the challenge was defined as “would it not be nice if all houses were radon-free,” referring to the ideal scenario where radon protective behavior is performed and facilitated easily among all homeowners in radon-prone areas. In the first session, the participants used this challenge to consider why houses are not already radon-free. In other words, “would it not be nice if all houses were radon-free” was the initial prompt to discuss barriers and facilitators in the first session. Since the participants all had experience with radon, this prompt was understandable for the participants as a starting point. Participants recorded all the problems (i.e., barriers) that arose on post-it notes while discussing them. These problem statements could relate to the causes of the challenge, the importance, the target audience, and other related issues, specifically in the form of “how-to questions.” This stems from the concept of how to ensure that all houses are radon-free, formulating a barrier as a facilitator; for instance, “how to make people aware” (i.e., facilitator) refers to the lack of awareness (i.e., barrier). Once saturation was reached and no new problems were added, dot-voting allowed for defining the most pressing problem statements. In other words, the first session discovered the why of the main challenge. Between the first and second sessions, the problem was defined further. In this case, the problem definition for the second session was “how to improve radon communication.” In the workshop’s second session, this was used as the prompt to start the discussion, together with the main findings from the first session. In this session, the focus was on ideation and brainstorming. The participants discussed potential radon communication strategies, selected the ones they considered the best, and started to develop protocols for the materials, which led to a communication strategy. This session explored the how of the main challenge. Both sessions aimed to diverge first (i.e., creating options) and converge afterward (i.e., selecting options). One of the tools often used in design thinking approaches is developing a customer journey, which indicates all the steps between being aware and purchasing a product or even becoming an ambassador (i.e., as a customer actively promoting the product among peers). Based on the precaution adoption process model and the 17 steps of radon behavior developed by Hevey , a homeowner journey was developed before the workshops. Seven steps were identified: awareness, evaluation of the knowledge (i.e., engagement with the health issue), purchase of radon test kit, delivery and conducting radon test, action (i.e., mitigating home), reassuring (i.e., confirming successful mitigation by re-testing), and ambassadorship (i.e., convincing others about the importance of radon tests). For every step, barriers, motivations, emotional states, and actions were identified. Developing the homeowner journey ensured a complete overview of the available literature about radon behavior. The full homeowner journey can be found in . If the discussion dtalled, the homeowner journey was an additional prompt during the first sessions. The workshops were conducted in Belgium and Slovenia. 3.3. Workshop 1: Belgium Effects of radon are a significant health problem in Belgium. Approximately 48% of the Walloon region in Belgium is expected to be affected by radon . Radon likely contributes to approximately 480 deaths due to lung cancer per year . To prevent this, approximately 36.000 dwellings need to be mitigated . The Federal Agency of Nuclear Control (FANC) is responsible for organizing activities to apply the regulations, comply with the obligations, and raise awareness of the actors involved in radon. Therefore, FANC strives for close collaboration with multiple actors, such as the provinces, municipalities, professional organizations, academic institutions, and the public. While exposure to radon at work is regulated and the employer is responsible for mitigating the working place, mitigation of dwellings is not legally required. It remains the responsibility of the homeowner . To increase the number of radon tests in dwellings, regional authorities contribute to radon test kits, which means that the price for a test kit is reduced from 30 euros to 15 euros. Financial help from the regional government for mitigation actions is also in place. The mitigation of a dwelling in Belgium costs between 500 euros and 5000 euros. Lists of companies with expertise in radon mitigation are published online . A communication plan was defined in 2014 and is updated yearly based on the evaluation of the past year to improve awareness and increase mitigation rates. In this context, a dedicated internet page was established. The effectiveness of the communication interventions is evaluated for the most impactful activities, such as orders of test kits. Other measures such as reach (e.g., visits to internet pages) and media return are also evaluated. FANC also tested social advertising in 2021 (paid ads on Twitter). However, this campaign was not further evaluated. The results of a public opinion survey show that 32% of the population are aware of radon and that 11% of them applied some mitigation measure in their home . The first workshop was conducted in March 2022 in Belgium. Due to COVID-19 restrictions, both sessions occurred online. An online whiteboard was used as an online alternative to physical post-its. 3.3.1. Sample Recruitment was conducted through local authorities, who spread the message about the workshops on their social media and websites. The principal investigator also contacted radon mitigation companies, who, in turn, forwarded the message to people that had completed (or were in the process of completing) radon mitigation. This way, people were invited to contact the research team to enroll in the workshops. The sample of the first workshop consisted of six participants, of which four detected radon in their homes, and two were professionally engaged with radon. Three participants belonged to the same family, all living in Luxembourg. This was unforeseen and only known at the start of the first session, but due to recruitment challenges, we decided that they still could participate as their experiences could inform us as well. In every session, five participants were present, with four overlapping participants in both sessions. 3.3.2. Facilitation Facilitators of ACOMPANY moderated the workshop in Belgium. This allowed the research team to observe and learn the methods they adopted. During both sessions, the researchers observed without interfering, as the objective was to explore first-hand barriers and solutions of the participants. This workshop demonstrated some limitations to the online format; therefore, we decided to wait until the end of COVID-19 restrictions to host the second workshop face-to-face. 3.4. Workshop 2: Slovenia Due to its geology, Slovenia has many municipalities heavily influenced by radon. It is estimated that 100 people per year die due to lung cancer caused by radon . To prevent radon-related deaths, the Slovenian Radiation Protection Administration is responsible for the Radon Action Plan . Through online and face-to-face meetings, it consults with all ministries involved with radon, including the Ministry of Health and Ministry of Environment, Technical Support Organizations, and Education. Free measurements for dwellings are available for residents in radon-risk areas; however, the number of available tests is limited. The average mitigation costs for standard dwelling amount to approximately a few thousand euros. Target groups of communication interventions are employers, employees, local decision-makers, and the public in general. Communication interventions are focused on increasing awareness and are mainly developed in the form of brochures. Other strategies include news articles, seminars, expert meetings, workshops, and a comic book for children . Perko and Turcanu determined that the frequency of personal advice, dialogue, and response to radon-related questions and concerns of residents are very good in Slovenia compared to other European countries . The effectiveness of the communication interventions is not measured, and objective radon awareness measurements among residents are unavailable. In May 2022, the second workshop occurred face-to-face in Slovenia. The recruitment was also conducted through local authorities; however, it was also picked up by local media, such as the local radio and newspaper. 3.4.1. Sample The sample of the second workshop consisted of 9 participants for the first session and 8 participants in the second session. All of them were residents from a high-risk area in Slovenia who were experienced with testing their homes and detected indoor radon concentrations above the reference level of 300 Becquerel/m 3 . They all were either planning to mitigate or had already performed mitigation measures. 3.4.2. Facilitation The second workshop was moderated by two researchers of the research team, native Slovenian speakers with experience with moderating qualitative research. The researchers who conducted the second workshop were briefed by those who observed the first one to align the workshop procedures. 3.5. Data Analysis Both workshops were recorded and transcribed according to the ethical guidelines of the social sciences. The research team conducted an inductive thematic analysis, adopting a semantical approach. The participants recorded their main thoughts regarding the barriers, facilitators, and communication approaches on post-it notes. Therefore, their views, opinions, and experiences were made explicit, hence the semantic approach. These post-it notes were used to code the transcripts to provide more background information. After each session, these post-it notes (i.e., codes) were categorized thematically by the research team, until a consensus was reached. Since the approach was to explore the barriers, facilitators, and communication ideas, no pre-defined codebook was used. The aim was to recruit participants who already had some experience with radon so that they could speak from their own experiences rather than a hypothetical scenario. This meant that we recruited people who had already measured (high) radon levels. A workshop was designed that consisted of two unstructured group sessions. Each session lasted two hours and was scheduled a week apart. More specifically, the framework of the double diamond was applied to the context of radon and the workshop design itself . The first stage of this framework, as seen in , is the challenge, which is the starting point of the workshops and describes the ideal scenario. For this research project, the challenge was defined as “would it not be nice if all houses were radon-free,” referring to the ideal scenario where radon protective behavior is performed and facilitated easily among all homeowners in radon-prone areas. In the first session, the participants used this challenge to consider why houses are not already radon-free. In other words, “would it not be nice if all houses were radon-free” was the initial prompt to discuss barriers and facilitators in the first session. Since the participants all had experience with radon, this prompt was understandable for the participants as a starting point. Participants recorded all the problems (i.e., barriers) that arose on post-it notes while discussing them. These problem statements could relate to the causes of the challenge, the importance, the target audience, and other related issues, specifically in the form of “how-to questions.” This stems from the concept of how to ensure that all houses are radon-free, formulating a barrier as a facilitator; for instance, “how to make people aware” (i.e., facilitator) refers to the lack of awareness (i.e., barrier). Once saturation was reached and no new problems were added, dot-voting allowed for defining the most pressing problem statements. In other words, the first session discovered the why of the main challenge. Between the first and second sessions, the problem was defined further. In this case, the problem definition for the second session was “how to improve radon communication.” In the workshop’s second session, this was used as the prompt to start the discussion, together with the main findings from the first session. In this session, the focus was on ideation and brainstorming. The participants discussed potential radon communication strategies, selected the ones they considered the best, and started to develop protocols for the materials, which led to a communication strategy. This session explored the how of the main challenge. Both sessions aimed to diverge first (i.e., creating options) and converge afterward (i.e., selecting options). One of the tools often used in design thinking approaches is developing a customer journey, which indicates all the steps between being aware and purchasing a product or even becoming an ambassador (i.e., as a customer actively promoting the product among peers). Based on the precaution adoption process model and the 17 steps of radon behavior developed by Hevey , a homeowner journey was developed before the workshops. Seven steps were identified: awareness, evaluation of the knowledge (i.e., engagement with the health issue), purchase of radon test kit, delivery and conducting radon test, action (i.e., mitigating home), reassuring (i.e., confirming successful mitigation by re-testing), and ambassadorship (i.e., convincing others about the importance of radon tests). For every step, barriers, motivations, emotional states, and actions were identified. Developing the homeowner journey ensured a complete overview of the available literature about radon behavior. The full homeowner journey can be found in . If the discussion dtalled, the homeowner journey was an additional prompt during the first sessions. The workshops were conducted in Belgium and Slovenia. Effects of radon are a significant health problem in Belgium. Approximately 48% of the Walloon region in Belgium is expected to be affected by radon . Radon likely contributes to approximately 480 deaths due to lung cancer per year . To prevent this, approximately 36.000 dwellings need to be mitigated . The Federal Agency of Nuclear Control (FANC) is responsible for organizing activities to apply the regulations, comply with the obligations, and raise awareness of the actors involved in radon. Therefore, FANC strives for close collaboration with multiple actors, such as the provinces, municipalities, professional organizations, academic institutions, and the public. While exposure to radon at work is regulated and the employer is responsible for mitigating the working place, mitigation of dwellings is not legally required. It remains the responsibility of the homeowner . To increase the number of radon tests in dwellings, regional authorities contribute to radon test kits, which means that the price for a test kit is reduced from 30 euros to 15 euros. Financial help from the regional government for mitigation actions is also in place. The mitigation of a dwelling in Belgium costs between 500 euros and 5000 euros. Lists of companies with expertise in radon mitigation are published online . A communication plan was defined in 2014 and is updated yearly based on the evaluation of the past year to improve awareness and increase mitigation rates. In this context, a dedicated internet page was established. The effectiveness of the communication interventions is evaluated for the most impactful activities, such as orders of test kits. Other measures such as reach (e.g., visits to internet pages) and media return are also evaluated. FANC also tested social advertising in 2021 (paid ads on Twitter). However, this campaign was not further evaluated. The results of a public opinion survey show that 32% of the population are aware of radon and that 11% of them applied some mitigation measure in their home . The first workshop was conducted in March 2022 in Belgium. Due to COVID-19 restrictions, both sessions occurred online. An online whiteboard was used as an online alternative to physical post-its. 3.3.1. Sample Recruitment was conducted through local authorities, who spread the message about the workshops on their social media and websites. The principal investigator also contacted radon mitigation companies, who, in turn, forwarded the message to people that had completed (or were in the process of completing) radon mitigation. This way, people were invited to contact the research team to enroll in the workshops. The sample of the first workshop consisted of six participants, of which four detected radon in their homes, and two were professionally engaged with radon. Three participants belonged to the same family, all living in Luxembourg. This was unforeseen and only known at the start of the first session, but due to recruitment challenges, we decided that they still could participate as their experiences could inform us as well. In every session, five participants were present, with four overlapping participants in both sessions. 3.3.2. Facilitation Facilitators of ACOMPANY moderated the workshop in Belgium. This allowed the research team to observe and learn the methods they adopted. During both sessions, the researchers observed without interfering, as the objective was to explore first-hand barriers and solutions of the participants. This workshop demonstrated some limitations to the online format; therefore, we decided to wait until the end of COVID-19 restrictions to host the second workshop face-to-face. Recruitment was conducted through local authorities, who spread the message about the workshops on their social media and websites. The principal investigator also contacted radon mitigation companies, who, in turn, forwarded the message to people that had completed (or were in the process of completing) radon mitigation. This way, people were invited to contact the research team to enroll in the workshops. The sample of the first workshop consisted of six participants, of which four detected radon in their homes, and two were professionally engaged with radon. Three participants belonged to the same family, all living in Luxembourg. This was unforeseen and only known at the start of the first session, but due to recruitment challenges, we decided that they still could participate as their experiences could inform us as well. In every session, five participants were present, with four overlapping participants in both sessions. Facilitators of ACOMPANY moderated the workshop in Belgium. This allowed the research team to observe and learn the methods they adopted. During both sessions, the researchers observed without interfering, as the objective was to explore first-hand barriers and solutions of the participants. This workshop demonstrated some limitations to the online format; therefore, we decided to wait until the end of COVID-19 restrictions to host the second workshop face-to-face. Due to its geology, Slovenia has many municipalities heavily influenced by radon. It is estimated that 100 people per year die due to lung cancer caused by radon . To prevent radon-related deaths, the Slovenian Radiation Protection Administration is responsible for the Radon Action Plan . Through online and face-to-face meetings, it consults with all ministries involved with radon, including the Ministry of Health and Ministry of Environment, Technical Support Organizations, and Education. Free measurements for dwellings are available for residents in radon-risk areas; however, the number of available tests is limited. The average mitigation costs for standard dwelling amount to approximately a few thousand euros. Target groups of communication interventions are employers, employees, local decision-makers, and the public in general. Communication interventions are focused on increasing awareness and are mainly developed in the form of brochures. Other strategies include news articles, seminars, expert meetings, workshops, and a comic book for children . Perko and Turcanu determined that the frequency of personal advice, dialogue, and response to radon-related questions and concerns of residents are very good in Slovenia compared to other European countries . The effectiveness of the communication interventions is not measured, and objective radon awareness measurements among residents are unavailable. In May 2022, the second workshop occurred face-to-face in Slovenia. The recruitment was also conducted through local authorities; however, it was also picked up by local media, such as the local radio and newspaper. 3.4.1. Sample The sample of the second workshop consisted of 9 participants for the first session and 8 participants in the second session. All of them were residents from a high-risk area in Slovenia who were experienced with testing their homes and detected indoor radon concentrations above the reference level of 300 Becquerel/m 3 . They all were either planning to mitigate or had already performed mitigation measures. 3.4.2. Facilitation The second workshop was moderated by two researchers of the research team, native Slovenian speakers with experience with moderating qualitative research. The researchers who conducted the second workshop were briefed by those who observed the first one to align the workshop procedures. The sample of the second workshop consisted of 9 participants for the first session and 8 participants in the second session. All of them were residents from a high-risk area in Slovenia who were experienced with testing their homes and detected indoor radon concentrations above the reference level of 300 Becquerel/m 3 . They all were either planning to mitigate or had already performed mitigation measures. The second workshop was moderated by two researchers of the research team, native Slovenian speakers with experience with moderating qualitative research. The researchers who conducted the second workshop were briefed by those who observed the first one to align the workshop procedures. Both workshops were recorded and transcribed according to the ethical guidelines of the social sciences. The research team conducted an inductive thematic analysis, adopting a semantical approach. The participants recorded their main thoughts regarding the barriers, facilitators, and communication approaches on post-it notes. Therefore, their views, opinions, and experiences were made explicit, hence the semantic approach. These post-it notes were used to code the transcripts to provide more background information. After each session, these post-it notes (i.e., codes) were categorized thematically by the research team, until a consensus was reached. Since the approach was to explore the barriers, facilitators, and communication ideas, no pre-defined codebook was used. 4.1. Workshop 1: Belgium (Online) 4.1.1. Session 1: Problem Statements The results of the first session were oriented toward problem formulations related to the following challenge: “would it not be nice if all houses were radon-free?”. In total, 36 problem statements were formulated, identifying the underlying barriers and facilitators. Not all of them were in the “how-to” format. However, they were still valuable in emphasizing certain problem areas. The following are examples of problem statements: “How to establish an EU standard?”, “How to oblige radon measures in new buildings?”, “How to find help from the state?”, “How to facilitate the necessary steps?”, “How to shock people?”, “How to develop a decision tree? ”, etc. The full list of problem statements can be found in . Another example includes problem statements such as “How to make people aware?”, “How to ‘touch’ people?”, “How to visualize the danger?”: “… we realize that people don’t know about radon in our country. I live in the province of Luxembourg [Belgium], which is the most affected. And despite everything we do, people don’t know about it. I think that if we want to be able to act and do something, people must first know.” (P2) “One difficulty is that when we talk about the FANC [Federal Agency of Nuclear Control], we don’t know, it’s something we don’t know too much about, which is, which is not close to here. So, there is a certain distance, both physical and perhaps also in the consciousness of people.” (P3) Other problem statements included issues related to “How to get help to remediate?”, “How to find reliable information?” and “How to find the right solution for the right house?”: “To give you an example, we have a list of companies in Luxembourg [country] that should be able to deal with radon. We contacted them all, the whole list, there is nobody who really has experience on it, but they are on the list of experts.” (P5) After diverging, i.e., collecting different problem statements, and after saturation was reached, the participants converged by choosing the problems that they felt were most important, as presented in . Participants compiled their top 3 issues. To provide an overview of the prioritized issues, researchers attributed 3 points to their number 1, 2 points to their number 2, and 1 point to their number 3. The ones with the most points are therefore considered the most important. Problem definition After the first session, researchers clustered the problem statements thematically to identify the underlying facilitators. The following categories were formulated: installing standardization to ensure quality ( n = 7), clarifying a stepwise approach ( n = 4), communication through different stakeholders ( n = 4), thresholds ( n = 7), cost of mitigation ( n = 2), mitigation contractors ( n = 2), and communication ( n = 10). The full overview can be found in . Since the study aimed to co-design communication tools, the problem definition was also related to communication. Since communication was also highly represented and comprised some of the prioritized problem statements, this decision was justified. 4.1.2. Session 2: Solution Statements In the second session, the working statement concerned communication. In total, 41 ideas were presented by the participants. Examples of ideas are workshops in primary schools, including general practitioners in the communication concerning radon, creating a “radon safe” label, a testimonial of someone who easily mitigated, a catchy radio spot with humor, advertising via social media, more visibility to mitigation companies, flyers in public spaces, etc. The full list of communication ideas can be found in . After saturation during the brainstorming, participants converged by voting for their favorite ideas. They each had two votes, and the results are presented in . During this session, the facilitator prompted ideas for four steps of the homeowner journey: radon awareness, evaluation (before testing), action (i.e., mitigation), and ambassadorships. To simplify the process for the participants, the research team decided to map the ideas to the homeowner journey among themselves after the session. Some ideas were mapped in multiple stages. The full overview can be found in . Most of the ideas were mapped to the first ( n = 20) and the second step ( n = 20) with a lot of overlapping communication strategies such as an advertising campaign via social media, a catchy radio spot with humor, a booklet in schools, press articles and flyers. In the action step, fewer ideas were presented ( n = 14), and these strategies implied more specific information. Examples include a testimonial of someone who easily mitigated radon effects, flyers with information about mitigation costs, showing examples of other people who mitigated, showing pictures that emphasize the simplicity of the process, and providing more visibility to solutions and mitigators. Finally, the last step, ambassadorship, was the one with the least ideas ( n = 5); however, those ideas do emphasize the social component of communication strategies, including, for instance, an advertising campaign on social media, a testimonial, creating a “radon safe” label, or organizing a competition with prizes for people who mitigated their houses. Due to the limits of the online format in time management and lacking group dynamics, the second session of the first workshop ended with prioritizing solutions and did not further proceed with designing the solutions. 4.2. Workshop 2: Slovenia (Face-to-Face) 4.2.1. Session 1: Problem Statements Similar to the first workshop in Belgium, the first session in Slovenia was oriented toward problem formulations; however, the highly involved participants had already started formulating solutions at this stage. Despite the different formats, the solutions provided in this first session also expose underlying issues. For clarification, we rephrased the solutions from the first workshop to problem statements; however, the original formulations can still be found in . In total, 45 problem statements/solutions were formulated. A few examples include: “How to include radon as a topic in schools?”, “How to provide understandable and accessible information about mitigation?”, “How to provide accessible free dosimeters?”, “How to get subventions from the state?” “How to guarantee the quality of the mitigation works?”, etc. The full list can be found in . Another example is “How to increase awareness about radon in the population?”. Multiple participants indicated that they learned about radon through their social networks: “Well, then one of my friends was encouraged [to test], and she also said, I didn’t know either, I didn’t know, and the problem is that we ordinary people don’t even know, unless we are really terribly interested in it, to even report it so that you can measure it.” (P6) “We had a measurement done because a friend of ours had done it a couple of 500 m away, and then we had it done.” (P9) After diverging, and when no new problems were added, the participants converged by voting for the most important problem statements in their opinion. They each cast three votes. The issues with the most votes were the most important barriers. The results of the dot voting can be found in . Problem definition The problem statements were clustered thematically by the researchers, resulting in the following categories: communication, information, and awareness ( n = 10), advice after measurement ( n = 6), comprehensive/holistic approach ( n = 3), accessibility of passive and active dosimeters and measurement support ( n = 9), mitigation support ( n = 5), the financial burden of mitigation ( n = 5), the legal requirement ( n = 6), and motivation ( n = 1). The full overview can be found in . Similar to the Belgian workshop, the communication, information, and awareness category were emphasized. Again, this justified the decision to focus on communication in the second session. More specifically, the following questions were raised: How do you think radon awareness should be raised? Moreover, how should advice on mitigation be communicated? 4.2.2. Session 2: Solution Statements For the first question about awareness, 22 ideas were formulated, including advertisements on YouTube, TikTok, and Instagram, regular information about radon in mass media, personal letters to all households, an interactive portal about radon, radon education in schools, contributions about radon in TV, radio, and newspapers. The participants voted for the best ideas, which can be found in . The group then discussed the details of the personal letter (i.e., informing households by post). For instance, the participants discussed that the letter should cover the prevalence of radon, the dangers, locations and ways to order dosimeters, the concerning radon values, and an invitation to participate in the measurements. They discussed that the municipality should draft the letter with an official signature. Further, they discussed the possibility of opening a special office to manage the radon campaign. The group also discussed whom to target and whether it should be addressed or unaddressed mail. They mentioned that a special message could be printed on the envelope, such as “it’s about your health.” The participants agreed that the letter should be sent in the winter. Creating a logo or corporate identity was also discussed, using red and yellow, as these colors are associated with radon areas, and green because it is associated with a solution. In the first part, the logo should be intimidating, and reassuring in the second part, as a solution. The group also discussed that the letter should be distributed by e-mail and social media. For the second question concerning the advice on mitigation, the group formulated 13 ideas. Examples included personal testimonials of people during mitigation, a list of mitigation contractors, social media campaigns, and personal communication with a selected advisor. The full list can be found in . Results of voting for the second question, resulting in the following prioritized ideas, can be seen in . The idea that received the most support was to hear people’s testimonials about their experiences with mitigation. The stories could either include a successful experience or lessons learned from less successful experiences. There was an idea to organize this through social networks online, for instance, through municipalities on social media. The group agreed that the information should not be too technical and should not resemble a commercial. Finally, they also discussed the need to target younger generations who are buying and building houses, and that information channels should be chosen accordingly. 4.1.1. Session 1: Problem Statements The results of the first session were oriented toward problem formulations related to the following challenge: “would it not be nice if all houses were radon-free?”. In total, 36 problem statements were formulated, identifying the underlying barriers and facilitators. Not all of them were in the “how-to” format. However, they were still valuable in emphasizing certain problem areas. The following are examples of problem statements: “How to establish an EU standard?”, “How to oblige radon measures in new buildings?”, “How to find help from the state?”, “How to facilitate the necessary steps?”, “How to shock people?”, “How to develop a decision tree? ”, etc. The full list of problem statements can be found in . Another example includes problem statements such as “How to make people aware?”, “How to ‘touch’ people?”, “How to visualize the danger?”: “… we realize that people don’t know about radon in our country. I live in the province of Luxembourg [Belgium], which is the most affected. And despite everything we do, people don’t know about it. I think that if we want to be able to act and do something, people must first know.” (P2) “One difficulty is that when we talk about the FANC [Federal Agency of Nuclear Control], we don’t know, it’s something we don’t know too much about, which is, which is not close to here. So, there is a certain distance, both physical and perhaps also in the consciousness of people.” (P3) Other problem statements included issues related to “How to get help to remediate?”, “How to find reliable information?” and “How to find the right solution for the right house?”: “To give you an example, we have a list of companies in Luxembourg [country] that should be able to deal with radon. We contacted them all, the whole list, there is nobody who really has experience on it, but they are on the list of experts.” (P5) After diverging, i.e., collecting different problem statements, and after saturation was reached, the participants converged by choosing the problems that they felt were most important, as presented in . Participants compiled their top 3 issues. To provide an overview of the prioritized issues, researchers attributed 3 points to their number 1, 2 points to their number 2, and 1 point to their number 3. The ones with the most points are therefore considered the most important. Problem definition After the first session, researchers clustered the problem statements thematically to identify the underlying facilitators. The following categories were formulated: installing standardization to ensure quality ( n = 7), clarifying a stepwise approach ( n = 4), communication through different stakeholders ( n = 4), thresholds ( n = 7), cost of mitigation ( n = 2), mitigation contractors ( n = 2), and communication ( n = 10). The full overview can be found in . Since the study aimed to co-design communication tools, the problem definition was also related to communication. Since communication was also highly represented and comprised some of the prioritized problem statements, this decision was justified. 4.1.2. Session 2: Solution Statements In the second session, the working statement concerned communication. In total, 41 ideas were presented by the participants. Examples of ideas are workshops in primary schools, including general practitioners in the communication concerning radon, creating a “radon safe” label, a testimonial of someone who easily mitigated, a catchy radio spot with humor, advertising via social media, more visibility to mitigation companies, flyers in public spaces, etc. The full list of communication ideas can be found in . After saturation during the brainstorming, participants converged by voting for their favorite ideas. They each had two votes, and the results are presented in . During this session, the facilitator prompted ideas for four steps of the homeowner journey: radon awareness, evaluation (before testing), action (i.e., mitigation), and ambassadorships. To simplify the process for the participants, the research team decided to map the ideas to the homeowner journey among themselves after the session. Some ideas were mapped in multiple stages. The full overview can be found in . Most of the ideas were mapped to the first ( n = 20) and the second step ( n = 20) with a lot of overlapping communication strategies such as an advertising campaign via social media, a catchy radio spot with humor, a booklet in schools, press articles and flyers. In the action step, fewer ideas were presented ( n = 14), and these strategies implied more specific information. Examples include a testimonial of someone who easily mitigated radon effects, flyers with information about mitigation costs, showing examples of other people who mitigated, showing pictures that emphasize the simplicity of the process, and providing more visibility to solutions and mitigators. Finally, the last step, ambassadorship, was the one with the least ideas ( n = 5); however, those ideas do emphasize the social component of communication strategies, including, for instance, an advertising campaign on social media, a testimonial, creating a “radon safe” label, or organizing a competition with prizes for people who mitigated their houses. Due to the limits of the online format in time management and lacking group dynamics, the second session of the first workshop ended with prioritizing solutions and did not further proceed with designing the solutions. The results of the first session were oriented toward problem formulations related to the following challenge: “would it not be nice if all houses were radon-free?”. In total, 36 problem statements were formulated, identifying the underlying barriers and facilitators. Not all of them were in the “how-to” format. However, they were still valuable in emphasizing certain problem areas. The following are examples of problem statements: “How to establish an EU standard?”, “How to oblige radon measures in new buildings?”, “How to find help from the state?”, “How to facilitate the necessary steps?”, “How to shock people?”, “How to develop a decision tree? ”, etc. The full list of problem statements can be found in . Another example includes problem statements such as “How to make people aware?”, “How to ‘touch’ people?”, “How to visualize the danger?”: “… we realize that people don’t know about radon in our country. I live in the province of Luxembourg [Belgium], which is the most affected. And despite everything we do, people don’t know about it. I think that if we want to be able to act and do something, people must first know.” (P2) “One difficulty is that when we talk about the FANC [Federal Agency of Nuclear Control], we don’t know, it’s something we don’t know too much about, which is, which is not close to here. So, there is a certain distance, both physical and perhaps also in the consciousness of people.” (P3) Other problem statements included issues related to “How to get help to remediate?”, “How to find reliable information?” and “How to find the right solution for the right house?”: “To give you an example, we have a list of companies in Luxembourg [country] that should be able to deal with radon. We contacted them all, the whole list, there is nobody who really has experience on it, but they are on the list of experts.” (P5) After diverging, i.e., collecting different problem statements, and after saturation was reached, the participants converged by choosing the problems that they felt were most important, as presented in . Participants compiled their top 3 issues. To provide an overview of the prioritized issues, researchers attributed 3 points to their number 1, 2 points to their number 2, and 1 point to their number 3. The ones with the most points are therefore considered the most important. Problem definition After the first session, researchers clustered the problem statements thematically to identify the underlying facilitators. The following categories were formulated: installing standardization to ensure quality ( n = 7), clarifying a stepwise approach ( n = 4), communication through different stakeholders ( n = 4), thresholds ( n = 7), cost of mitigation ( n = 2), mitigation contractors ( n = 2), and communication ( n = 10). The full overview can be found in . Since the study aimed to co-design communication tools, the problem definition was also related to communication. Since communication was also highly represented and comprised some of the prioritized problem statements, this decision was justified. In the second session, the working statement concerned communication. In total, 41 ideas were presented by the participants. Examples of ideas are workshops in primary schools, including general practitioners in the communication concerning radon, creating a “radon safe” label, a testimonial of someone who easily mitigated, a catchy radio spot with humor, advertising via social media, more visibility to mitigation companies, flyers in public spaces, etc. The full list of communication ideas can be found in . After saturation during the brainstorming, participants converged by voting for their favorite ideas. They each had two votes, and the results are presented in . During this session, the facilitator prompted ideas for four steps of the homeowner journey: radon awareness, evaluation (before testing), action (i.e., mitigation), and ambassadorships. To simplify the process for the participants, the research team decided to map the ideas to the homeowner journey among themselves after the session. Some ideas were mapped in multiple stages. The full overview can be found in . Most of the ideas were mapped to the first ( n = 20) and the second step ( n = 20) with a lot of overlapping communication strategies such as an advertising campaign via social media, a catchy radio spot with humor, a booklet in schools, press articles and flyers. In the action step, fewer ideas were presented ( n = 14), and these strategies implied more specific information. Examples include a testimonial of someone who easily mitigated radon effects, flyers with information about mitigation costs, showing examples of other people who mitigated, showing pictures that emphasize the simplicity of the process, and providing more visibility to solutions and mitigators. Finally, the last step, ambassadorship, was the one with the least ideas ( n = 5); however, those ideas do emphasize the social component of communication strategies, including, for instance, an advertising campaign on social media, a testimonial, creating a “radon safe” label, or organizing a competition with prizes for people who mitigated their houses. Due to the limits of the online format in time management and lacking group dynamics, the second session of the first workshop ended with prioritizing solutions and did not further proceed with designing the solutions. 4.2.1. Session 1: Problem Statements Similar to the first workshop in Belgium, the first session in Slovenia was oriented toward problem formulations; however, the highly involved participants had already started formulating solutions at this stage. Despite the different formats, the solutions provided in this first session also expose underlying issues. For clarification, we rephrased the solutions from the first workshop to problem statements; however, the original formulations can still be found in . In total, 45 problem statements/solutions were formulated. A few examples include: “How to include radon as a topic in schools?”, “How to provide understandable and accessible information about mitigation?”, “How to provide accessible free dosimeters?”, “How to get subventions from the state?” “How to guarantee the quality of the mitigation works?”, etc. The full list can be found in . Another example is “How to increase awareness about radon in the population?”. Multiple participants indicated that they learned about radon through their social networks: “Well, then one of my friends was encouraged [to test], and she also said, I didn’t know either, I didn’t know, and the problem is that we ordinary people don’t even know, unless we are really terribly interested in it, to even report it so that you can measure it.” (P6) “We had a measurement done because a friend of ours had done it a couple of 500 m away, and then we had it done.” (P9) After diverging, and when no new problems were added, the participants converged by voting for the most important problem statements in their opinion. They each cast three votes. The issues with the most votes were the most important barriers. The results of the dot voting can be found in . Problem definition The problem statements were clustered thematically by the researchers, resulting in the following categories: communication, information, and awareness ( n = 10), advice after measurement ( n = 6), comprehensive/holistic approach ( n = 3), accessibility of passive and active dosimeters and measurement support ( n = 9), mitigation support ( n = 5), the financial burden of mitigation ( n = 5), the legal requirement ( n = 6), and motivation ( n = 1). The full overview can be found in . Similar to the Belgian workshop, the communication, information, and awareness category were emphasized. Again, this justified the decision to focus on communication in the second session. More specifically, the following questions were raised: How do you think radon awareness should be raised? Moreover, how should advice on mitigation be communicated? 4.2.2. Session 2: Solution Statements For the first question about awareness, 22 ideas were formulated, including advertisements on YouTube, TikTok, and Instagram, regular information about radon in mass media, personal letters to all households, an interactive portal about radon, radon education in schools, contributions about radon in TV, radio, and newspapers. The participants voted for the best ideas, which can be found in . The group then discussed the details of the personal letter (i.e., informing households by post). For instance, the participants discussed that the letter should cover the prevalence of radon, the dangers, locations and ways to order dosimeters, the concerning radon values, and an invitation to participate in the measurements. They discussed that the municipality should draft the letter with an official signature. Further, they discussed the possibility of opening a special office to manage the radon campaign. The group also discussed whom to target and whether it should be addressed or unaddressed mail. They mentioned that a special message could be printed on the envelope, such as “it’s about your health.” The participants agreed that the letter should be sent in the winter. Creating a logo or corporate identity was also discussed, using red and yellow, as these colors are associated with radon areas, and green because it is associated with a solution. In the first part, the logo should be intimidating, and reassuring in the second part, as a solution. The group also discussed that the letter should be distributed by e-mail and social media. For the second question concerning the advice on mitigation, the group formulated 13 ideas. Examples included personal testimonials of people during mitigation, a list of mitigation contractors, social media campaigns, and personal communication with a selected advisor. The full list can be found in . Results of voting for the second question, resulting in the following prioritized ideas, can be seen in . The idea that received the most support was to hear people’s testimonials about their experiences with mitigation. The stories could either include a successful experience or lessons learned from less successful experiences. There was an idea to organize this through social networks online, for instance, through municipalities on social media. The group agreed that the information should not be too technical and should not resemble a commercial. Finally, they also discussed the need to target younger generations who are buying and building houses, and that information channels should be chosen accordingly. Similar to the first workshop in Belgium, the first session in Slovenia was oriented toward problem formulations; however, the highly involved participants had already started formulating solutions at this stage. Despite the different formats, the solutions provided in this first session also expose underlying issues. For clarification, we rephrased the solutions from the first workshop to problem statements; however, the original formulations can still be found in . In total, 45 problem statements/solutions were formulated. A few examples include: “How to include radon as a topic in schools?”, “How to provide understandable and accessible information about mitigation?”, “How to provide accessible free dosimeters?”, “How to get subventions from the state?” “How to guarantee the quality of the mitigation works?”, etc. The full list can be found in . Another example is “How to increase awareness about radon in the population?”. Multiple participants indicated that they learned about radon through their social networks: “Well, then one of my friends was encouraged [to test], and she also said, I didn’t know either, I didn’t know, and the problem is that we ordinary people don’t even know, unless we are really terribly interested in it, to even report it so that you can measure it.” (P6) “We had a measurement done because a friend of ours had done it a couple of 500 m away, and then we had it done.” (P9) After diverging, and when no new problems were added, the participants converged by voting for the most important problem statements in their opinion. They each cast three votes. The issues with the most votes were the most important barriers. The results of the dot voting can be found in . Problem definition The problem statements were clustered thematically by the researchers, resulting in the following categories: communication, information, and awareness ( n = 10), advice after measurement ( n = 6), comprehensive/holistic approach ( n = 3), accessibility of passive and active dosimeters and measurement support ( n = 9), mitigation support ( n = 5), the financial burden of mitigation ( n = 5), the legal requirement ( n = 6), and motivation ( n = 1). The full overview can be found in . Similar to the Belgian workshop, the communication, information, and awareness category were emphasized. Again, this justified the decision to focus on communication in the second session. More specifically, the following questions were raised: How do you think radon awareness should be raised? Moreover, how should advice on mitigation be communicated? For the first question about awareness, 22 ideas were formulated, including advertisements on YouTube, TikTok, and Instagram, regular information about radon in mass media, personal letters to all households, an interactive portal about radon, radon education in schools, contributions about radon in TV, radio, and newspapers. The participants voted for the best ideas, which can be found in . The group then discussed the details of the personal letter (i.e., informing households by post). For instance, the participants discussed that the letter should cover the prevalence of radon, the dangers, locations and ways to order dosimeters, the concerning radon values, and an invitation to participate in the measurements. They discussed that the municipality should draft the letter with an official signature. Further, they discussed the possibility of opening a special office to manage the radon campaign. The group also discussed whom to target and whether it should be addressed or unaddressed mail. They mentioned that a special message could be printed on the envelope, such as “it’s about your health.” The participants agreed that the letter should be sent in the winter. Creating a logo or corporate identity was also discussed, using red and yellow, as these colors are associated with radon areas, and green because it is associated with a solution. In the first part, the logo should be intimidating, and reassuring in the second part, as a solution. The group also discussed that the letter should be distributed by e-mail and social media. For the second question concerning the advice on mitigation, the group formulated 13 ideas. Examples included personal testimonials of people during mitigation, a list of mitigation contractors, social media campaigns, and personal communication with a selected advisor. The full list can be found in . Results of voting for the second question, resulting in the following prioritized ideas, can be seen in . The idea that received the most support was to hear people’s testimonials about their experiences with mitigation. The stories could either include a successful experience or lessons learned from less successful experiences. There was an idea to organize this through social networks online, for instance, through municipalities on social media. The group agreed that the information should not be too technical and should not resemble a commercial. Finally, they also discussed the need to target younger generations who are buying and building houses, and that information channels should be chosen accordingly. By setting up a qualitative co-design workshop with homeowners, we aimed at gaining more in-depth knowledge about the barriers that people experience in mitigating their house, on the one hand, and collecting their creative input and insights about ways to communicate the dangers of radon could be improved on the other. First of all, the results show that the barriers people experience are situated within different levels of interventions and different steps of behavior, as described in the literature review. The stages discussed in this section are simplified and focus on awareness, testing, and mitigating behavior for clarification purposes. Barriers related to the first stages of behavior were focused on a lack of awareness and engaging communication. The participants agreed that awareness should be the first step. In Belgium, the focus was placed on more attention-grabbing awareness campaigns, such as social media campaigns and humor, while Slovenia focused on personalized letters. This is in line with the research of Weinstein et al., where they tested whether personalized phone calls and letters affected perceived susceptibility and self-protective behavior (i.e., intention to test). They determined that personal susceptibility did increase significantly for those who received the phone call and the letter; however, no differences were detected in terms of intention to test. This could indicate that the proposed letters by the participants could successfully increase engagement with the health topic, yet that other communication strategies are needed to address the further steps in the mitigating process . These results also show the nuance of the concept of awareness, where a discrepancy between being aware and making a personal risk assessment remains. As Poortinga et al. reported, high levels of awareness do not always result in higher levels of concern; therefore, raising awareness could be focused more on grabbing attention and raising curiosity rather than merely informing. Barriers associated with testing behavior include the lack of available active and passive dosimeters in Slovenia. According to the participants, communication in this stage should be more specific than in the awareness stage; for instance, a comprehensive website with information, workshops, or newspaper articles would provide them with the information they need without overwhelming. Moreover, information from different stakeholders, such as medical doctors, could help emphasize the importance of radon testing. Apart from the accessibility of tests in Slovenia, no issues were mentioned regarding the costs of test kits. When examining the next stage, it can be observed that many barriers are related to mitigating behavior. Participants highlighted the importance of personalized advice after testing, with a clear step-wise approach on what steps to take next and how to do so. Finding mitigating companies with radon experience was challenging, according to the participants in Belgium and Slovenia. Moreover, the lack of guaranteed results after mitigation was a particularly important barrier in Belgium. Participants indicated that this had to be the state’s responsibility to implement regulations for these companies, as that would facilitate the process of the homeowners finding the best help for their particular radon problem. This could be achieved by certifying certain mitigating companies or involving inspections at mitigating companies, as proposed by the participants. Further, the financial burden of mitigation was mentioned in both workshops, emphasizing the need for subventions or financial aid from the government. Regarding mitigation behavior, the participants indicated a need for communication on different levels, for instance, stakeholder communication. They felt the involved stakeholders (e.g., medical professionals, mitigating companies, local authorities) are not sufficiently up to date in helping homeowners accordingly with radon issues. Especially in this stage, participants expressed a need for detailed and clear information, and both countries suggested using testimonials. The participants emphasized that the testimonial should contain a story of someone who mitigated their house or what lessons could be learned from unsuccessful mitigations. In that way, both the problem and the solution were addressed. This idea is already supported by the literature on narratives, stating that narratives could help in facilitating information processing, comprehension, and recall . Overarching barriers were related to legislation and regulation. On the policy level, both workshops showed a need for obligatory radon measures in new buildings; moreover, a need for a European standard was also expressed in Belgium. Despite the European Basic Safety Standards and the inclusion of radon measures in the building permit in Belgium, participants still expressed these aspects as a need for future policy-level interventions (Council Directive 2013/59/EURATOM, of 5 December 2013). This aligns with the current policy measures; however, the policy must be implemented sufficiently to impact homeowners’ barriers. Further, policy changes in adding radon levels to the energy certificate to regulate radon levels in the housing market were proposed, which agrees with previous research on mitigating . Regarding communication, the participants highlighted the need for a holistic step-wise approach, where communication follows the different stages of behavior and a consistent message is conveyed across stakeholders, channels, and time. Generally, it is important to note that behavior change will only occur if the environment is ready. In other words, barriers related to, for instance, the availability of dosimeters and mitigation companies should be addressed first before communicating about the health risks to ensure fitting solutions are available. This study indicated that co-design workshops and participatory research are crucial to gaining the users’ perspectives and ideas early in the intervention design. The face-to-face workshop was preferred when comparing both workshops, especially since this setting increased the group dynamic and collaboration efforts. The online format was, given the circumstances, still valuable in understanding the barriers and collaborating on communication ideas, yet a face-to-face setting was needed to conduct an even more in-depth inquiry. Design thinking workshops have shown to be valuable in the intervention design process related to radon; however, other health topics could and should also be addressed with participatory methods, such as design thinking, early on to maximize the involvement and input of the target group. 5.1. Limitations Just like any study, this study also experienced some limitations. Ideally, both workshops would be conducted in a face-to-face setting instead of the online setting in Belgium. This would facilitate even more creativity and sharing experiences among the participants. Moreover, recruitment challenges limited us to one workshop with two sessions in each country. Although we gained many new perspectives and ideas, more workshops with more participants would allow for saturation among the population instead of saturation among the sample. Regarding the sample of these workshops, we focused on homeowners that had measured (high) radon levels in their homes. Although this was the purpose of the study, it created selection bias. 5.2. Future Research Future research should explore more participatory research designs, both in intervention design research and radon health communication, emphasizing different social categories and countries. Moreover, scholars could investigate more comprehensive communication strategies with adapted messages depending on the sample’s behavior change stage. Finally, researchers could explore the ideas provided by the participants further in terms of theoretical framework, but also in terms of effectiveness in a lab setting. Just like any study, this study also experienced some limitations. Ideally, both workshops would be conducted in a face-to-face setting instead of the online setting in Belgium. This would facilitate even more creativity and sharing experiences among the participants. Moreover, recruitment challenges limited us to one workshop with two sessions in each country. Although we gained many new perspectives and ideas, more workshops with more participants would allow for saturation among the population instead of saturation among the sample. Regarding the sample of these workshops, we focused on homeowners that had measured (high) radon levels in their homes. Although this was the purpose of the study, it created selection bias. Future research should explore more participatory research designs, both in intervention design research and radon health communication, emphasizing different social categories and countries. Moreover, scholars could investigate more comprehensive communication strategies with adapted messages depending on the sample’s behavior change stage. Finally, researchers could explore the ideas provided by the participants further in terms of theoretical framework, but also in terms of effectiveness in a lab setting. In this study, the questions were raised: what are the main barriers and facilitators to engaging in radon-protective behavior experienced by homeowners, and how are these addressed in current interventions? Second, how can the communication about radon be improved to be more relevant and engaging for the target group? To investigate these questions, we designed a participatory co-design research method with homeowners in Belgium and Slovenia. The findings of these workshops show that participants require more policy and legislation, for instance, about certifying mitigation companies or including radon measurement on the energy certificate. Moreover, they experience a need for support from the state during radon testing and mitigating procedures, both in terms of financial aid and communication or advice. Furthermore, they indicated a need for more awareness among the general public and, more specifically, a lack of engagement. A holistic communication approach is also needed, including by stakeholders such as general practitioners and architects. When looking at communication specifically, both workshops suggested that communication strategies should be amended to match the stage from awareness to having a radon-safe home. Communication tools such as radio spots with humor or personalized letters to raise awareness and engagement were proposed. Further, testimonials were pointed out as an effective way to highlight the issues and solutions of people who reported similar experiences. Further research should adopt co-design methods, both in research about radon health communication and in different fields. Further, scholars could test the effectiveness of some of these ideas in a controlled setting and in an integrated, multi-stage intervention. |
Predicting Fecal Indicator Bacteria Using Spatial Stream Network Models in A Mixed-Land-Use Suburban Watershed in New Jersey, USA | d69ec1e9-d4c4-41a2-9871-312ce61551c4 | 10049084 | Microbiology[mh] | Good water quality ensures the well-being of urban and suburban residents, as it safeguards public health and provides economic benefits through recreational opportunities . However, microbial water quality is often impaired due to expanded impervious areas and the improper management of sanitary infrastructures, such as combined sewer overflows and failing septic tanks, resulting in elevated concentrations of fecal indicator bacteria and waterborne pathogens in urban and suburban waterways . Therefore, it is essential to monitor microbial water quality regularly. Microbial water quality is assessed using fecal indicator bacteria (FIB), such as E. coli , fecal coliform, and enterococci. These FIBs originate from the gastrointestinal tract of human or warm-blooded animals and are introduced into the urban and suburban waterbodies through fecal matter . Although FIBs are not pathogens per se, their presence indicates the potential cooccurrence of true pathogens . Water quality criteria were hence established to protect public health based on epidemiological studies that investigate the relationships between the rates of gastrointestinal illness among swimmers and the levels of FIBs [ , , , ]. For example, the most recent USEPA’s Recreational Water Quality Criteria uses a geometric mean of 126 CFU of E. coli per 100 mL to indicate an estimated illness rate of 36 per 1000 primary contact recreators in fresh water . Exceedance of the water quality threshold indicates an elevated risk of waterborne illness, and the waterbody is deemed impaired and needs to be listed on the List of Impaired Waters (i.e., Clean Water Act 303 (d) List). Water quality restoration plans, such as total maximum daily load (TMDL) plans, are then to be developed to help guide pollution reduction efforts in order to attain water quality standards . Development of a TMDL plan requires identifying and quantifying sources of pollutants and determining the degree of reduction needed to meet the applicable water quality standards . Mathematic models are heavily involved in the process of TMDL development for pollutant load estimation and source allocation in order to identify and quantify the degree of contamination in a waterbody for better remedial plans . Mechanistic models that are based on underlying hydrological and biogeochemical processes, such as streambed bacterial deposition and decay in the natural system, have been widely utilized for TMDL development, such as Soil and Water Assessment Tool (SWAT) and Hydrological Simulation Program-FORTRAN (HSPF) [ , , ]. These process-based models often demand more expertise in understanding detailed mechanisms and require more time in developing models . On the other hand, empirical models that are based on statistical approaches to estimate pollutant loads or identify associations between stressors and response variables are also widely used in practice . Without needing to first understand the underlying complex processes regarding the fate and transport of contaminants, empirical approaches serve as a statistical alternative to support TMDL development . A variety of regression models have been applied to identify key environmental variables linked to deteriorating water quality indicators, including linear, logistic, and Poisson regressions [ , , , ]. Linear regressions are highly accessible, easy to implement and have been widely applied in water quality data analysis. They are broadly accepted among water resource stakeholders . Building a valid regression model, however, often requires a relatively large number of observations and enhanced knowledge to identify the relationships between the response and input variables. While the latter can often be established under a theoretical framework and multivariate and/or nonparametric exploration, acquiring a large number of observations could prove to be difficult, especially when sample collection and processing are costly and take a long time. Oftentimes, compromises have to be made, accepting a larger variance for the sake of a relatively small sample size to proceed with the intended analysis. In addition, linear regression assumes no heteroscedasticity among the residuals and the independence of observations . In reality, since “near things are more related than distant things” , observations from adjacent geographic locations tend to have similar characteristics. The term “spatial autocorrelation” or “autocovariance” describes the similarity of measurements as a function of the distance separating them . Studies that consider spatial autocorrelation often use Euclidean distance to measure the straight-line distance between measurements and apply it in the terrestrial ecosystems . However, due to the constraints of movement of aquatic organisms and materials within stream channels, for stream water quality modeling, along-channel in-stream distance might be more appropriate . Modeling water quality with stream distance had not previously been well explored until the last decade. Money and Carter developed a predictive model of E. coli based on turbidity data with river-based distance. In addition to turbidity, other water quality and quantity parameters and watershed characteristics, such as land use type and locations of point source pollutions, often play essential roles in affecting microbial water quality, and have been widely considered in TMDL implementation plan development . To take into consideration the potential spatial autocorrelation in the modeling process, spatial stream network (SSN) models were developed and applied in recent studies [ , , , , ]. An SSN model employs a pre-established ArcGIS toolset and open source R package to provide an empirical approach to modeling stream water quality in response to watershed attributes with the consideration of spatial autocorrelation to better reflect stream characteristics . Autocovariance of the SSN models was constructed based on moving average functions to address spatial autocorrelation and measure the degree of dependence among observations in the stream systems . The conventional autocovariance model assumes isotrophy, indicating that the autocorrelation among observations depends only on the distance separating them, but not on the direction . Moving average functions address the direction-based autocorrelation in the stream systems in tail-up or tail-down models . Since SSN modeling relies less on understanding underlying hydrological and biogeochemical processes contributing to water quality but utilizes readily available land use/land cover (LULC) data, it provides greater potential for wider applications by stakeholders engaging in water quality issues in urban and suburban environments with limited resources . SSN modeling can also predict FIB at a predetermined interval along the stream channel, providing opportunities to identify hot spots for further water quality monitoring and restoration plans . The Musconetcong River, from which we took water samples for the current study, is located in northwestern New Jersey and forms part of the National Wild and Scenic River System in recognition of its remarkable recreational resources . Urban and suburban water recreation provides opportunities to promote physical and mental health among residents, foster environmental stewardship and community engagement, and support tourism and the local economy . However, the water quality of the Musconetcong River has been historically impaired due to intensive urbanization and roadway development . In fact, it has been listed in the New Jersey Integrated Water Quality Monitoring and Assessment Report (303 (d) List) due to the high abundance of fecal coliforms in northwestern New Jersey. Historically, human wastewater was not connected to sewer systems in the watershed, and various types of domestic and industrial wastewater effluents were discharged into the river through dysfunctional septic tanks. Other non-point source originated from livestock operations, wildlife, etc., contributing considerably to fecal contaminations throughout this mixed-land-use watershed . In 2003, a TMDL plan was established to address microbial water quality deterioration here, requiring a 93% reduction in fecal coliforms . Since then, various water quality restoration projects have been undertaken, including riparian buffers, prescribed grazing, herd reduction, sinkhole closure, and green infrastructure . In 2018, a water quality monitoring study concluded that there had been a significant improvement in microbial water quality over the past decade; however, further reductions in E. coli loads were still required to meet the TMDL goal . Local water quality stakeholders have been developing additional monitoring strategies and water quality restoration plans, with delisting of the Musconetcong River from the 303 (d) List as the goal. In this study, we aimed to develop empirical models with consideration of the spatial autocorrelation and associated key watershed variables with fecal indicator bacteria in the suburban mixed-land-use Musconetcong River watershed. Specifically, our goal was to analyze patterns of E. coli concentrations in response to upstream land use attributes through spatial stream network models. We hypothesize that the elevated E. coli concentrations in the Musconetcong River were strongly associated with the upstream urban land use. The results from this study will not only provide insights into identifying key land use stressors to guide future water quality restoration directions in this suburban Musconetcong River watershed, but also serve as a potential water quality modeling framework for different watersheds in the USA and beyond when resources are limited to urban and suburban water quality stakeholders.
2.1. Water Quality, Land Use and Precipitation Data E. coli was used as a fecal indicator bacterium as it is more reflective of recent fecal contamination from warm-blooded animals compared to fecal coliforms or total coliforms . Surface water samples were collected twice a month from May to October 2018, with three additional samplings from June to August for a total of 21 events at 5 mainstem and 6 tributary sites located throughout the middle section of the Musconetcong River watershed ( ). No samples were collected when stream channels were dry. The sites were selected based on past water quality monitoring activities showing the significant potential of fecal contaminations. Grab samples were collected aseptically into 1 L sterilized polypropylene bottles and were placed immediately in a cooler during transit. Enumeration of E. coli was based on an EPA-approved method using mColiBlue24 ® Broth (Hach Method 10029). In brief, samples were filtered through mixed cellulose ester membrane filters (0.45 mm, 47 mm, Hach, Loveland Colorado) and then soaked with mColiBlue24 ® Broth selective medium. The filters were subsequently incubated at 35 °C for 24 h. After incubation, colonies showing a blue/indigo color were recorded as E. coli . Tentative E. coli colonies were further verified using brilliant green bile (BGB) and lauryl tryptose broth (LTB). The final results were reported as colony forming units (CFU) per 100 mL of water samples. After the processing of the samples, we checked the weather data on the days when samples were collected and determined if certain samples were collected during a storm event based on the weather data. All E. coli data were uploaded to the Water Quality Portal ( https://www.waterqualitydata.us/ , accessed on 5 December 2021). The 2015 land use/land cover data were acquired from the New Jersey Department of Environmental Protection (NJDEP), Bureau of Geographic Information Systems. A modified Anderson coding system was used to categorize land use data . The total drainage area for the whole Musconetcong River watershed was 120,941.71 acres, with 53.2% forest, 21.2% urban, and 14.0% agricultural land. Four land use categories were extracted as explanatory variables for spatial stream network (SSN) modeling, including urban (1000), pasture (2120), forest (4000) and wetland (6000). Due to the sampling cost and length, the sample size in the current study was not particularly large (21 events). While more detailed land use land cover products, such as the USGS’s Multi-Resolution Land Characteristics (MRLC) land cover, are available, to avoid losing too many degrees of freedom in the empirical study, we elected to use the relatively coarse land use land cover categorization. Urban land use classification covered land use characteristics from rural single units to high-density multiple residential dwellings, commercial buildings, and industrial areas. Houses with various dwelling units or a high density in the study area were all connected to septic tanks instead of sewage pipes for waste disposal (Personal Communication with the Musconetcong Watershed Association). The study’s purpose was to investigate in general how urbanization within a watershed impacts water quality. The pasture subgroup was selected from the agricultural land use categories (2000), as failing septic systems and poor livestock management practices were previously listed as major threats to microbial water quality . Precipitation data were obtained from NJDEP, the Department of Water Monitoring & Standards ( https://njdep.rutgers.edu/rainfall/ , accessed on 5 December 2021). The rainfall threshold for a storm event was defined as 12.7 mm (0.5 inch) within 36 h, the same definition used in the Watershed Restoration and Protection Plan for the Musconetcong River Watershed created by the Rutgers Cooperative Extension Water Resources Program . A summary of analytical methods and data sources used in this study is shown in . 2.2. Spatial Stream Network Modeling To explore the impact of land use in the watershed on water quality, we extended from traditional regression analysis to adopt the spatial stream network model in this study [ , , ]. Spatial stream network models are generalized linear mixed models allowing explanation of variance in observations with both fixed and random effects due to spatial autocorrelation among geographically observed events , which can be expressed as follows: y = Xβ + z + ε where y is a vector of observations, X is a matrix for fixed effects that explain the variance of the observations that can be captured by general spatial patterns, β is a vector of coefficients for X, z accounts for random effects resulting from spatial autocorrelation that cannot be explained by fixed effects, and ε is a vector of independent random errors . The spatial autocorrelation structures that describe the random effects can be modeled with linear with sill, Mariah, exponential, Epanechnikov, and spherical models . Parameters for spatial autocovariance functions include nugget, partial sill, and range. The nugget of the autocovariance function describes the portion of variance that cannot be explained, the partial sill accounts for the variance that depends on the distance among observations, and the range indicates the minimal distance at which observations are no longer spatially correlated . SSN models can also incorporate autocovariance models based on Euclidean distance, accounting for terrestrial and atmospheric factors that are stream-independent . Both tail-up and tail-down autocovariance structures were established to estimate spatial autocorrelation. In tail-up models, moving average functions run upstream only and split at confluences. Spatial weights are required to estimate the proportion of upstream tributary influence based on flow volume, watershed area, or other relevant attributes . Tail-up models are only applicable to flow-connected sites, where water directly flows from an upstream to a downstream site characterized by passive downstream movement, e.g., temperature, bacteria, or sediment . In tail-down models, moving average functions run in the downstream direction unilaterally until reaching the most downstream location of the stream network, indicating the possibility of autocorrelation among all locations . In addition to flow-connected sites, tail-down models can be applied to flow-unconnected sites, where some upstream movement is expected to facilitate the connectivity for a given attribute, e.g., fish or macroinvertebrates . In order to perform spatial stream network (SSN) modeling, an SSN object was first created using STARS geoprocessing toolset in ArcGIS 10.8 (ESRI, Redlands, CA, USA). STARS is used to build a landscape network (LSN) as a personal geodatabase to represent spatial relationships, such as flow connectivity, direction, and distance and transform the geometry, attribute data, and topological relationships among features of GIS datasets into an SSN object that can be easily accessed and analyzed in R statistical software with the ssn package for SSN modeling. The process included building a landscape network (LSN), creating reach contributing areas , calculating RCA attributes and area, accumulating watershed attributes, incorporating the sampling locations into the LSN, calculating watershed attributes, and calculating upstream distance, segment proportion influence, and additive function . In the watershed of the current study, we obtained a total of 427 nodes in the stream network. The entire process of generating the SSN object was automated, and once generated in ArcGIS, the SSN object was then exported to R statistical software for SSN modeling. In this study, the ssn package in R was used to fit spatial stream network models to the observation data based on 11 sampling locations. Two models were established and calibrated. The first model used the geometric mean concentrations of E. coli for all sampling events as the response variable; the second model used the geometric mean concentrations of E. coli during storm events only as the response variable. Shapiro–Wilk tests were used to test the normality of the distributions of the original values . If the normality assumption was not satisfied, the values of geometric mean concentrations were log-transformed. A nonspatial model based on an ordinary least squares regression and two spatial models based on Euclidean and stream distance, respectively, were established to incorporate four upstream watershed attributes as explanatory variables, including urban, pasture, forest, and wetland. Percentage upstream land use values were derived from dividing the total accumulated area of upstream land use by the total upstream catchment area processed and generated via the STARS toolset. Variance inflation factors (VIF) were calculated to check multicollinearity among the four land use variables. Tail-up models were applied in the spatial autocovariance functions for the SSN modeling based on stream distance, as the movement of bacteria along the stream networks is characterized by passive downstream transport . Likelihood approaches were utilized to estimate the regression coefficients b for the matrix of fixed effects . Maximum likelihood (ML) estimators have been shown to be greatly biased when the number of observations is small , and thus restricted maximum likelihood (REML) was used. An alpha value of 0.05 was used to evaluate whether the relationship between watershed variables (urban, agricultural, forest, or wetland) and E. coli concentrations in the models was statistically significant. The root mean square prediction error (RMSPE) of the leave one out cross validation (LOOCV) was used as the criterion to compare the predictive capability among models. LOOCV was performed by excluding one observation at a time. Submodels were calculated and compared with the excluded samples . A lower RMSPE of the LOOCV indicated a better model performance . The coefficient of determination (R 2 ) was used to measure the proportion of variance in the observations explained by the fixed or random effects of each model. 2.3. Predicting E. coli Concentrations One of the practical uses for establishing the empirical SSN model for the Musconetcong River watershed is that we can use the model to predict the concentration of E. coli along the river. Predictions were performed for the geometric mean concentrations of E. coli for all sampling events and during storm events based on Euclidean distance and stream distance models, respectively. A total of 816 predictive sites were assigned evenly across the Musconetcong River stream network at a 300-m interval. Determination of predictive sites was conducted using the STARS geoprocessing toolset in ArcGIS 10.8. Predictions of E. coli concentrations were performed in the R statistical software with SSN package . The results were illustrated on stream maps illustrating both the log 10 geometric mean concentrations of E. coli . The capability of predicting the concentration of E. coli provides both a means to assess modeling performance and a supplement to the regular sampling routine for water quality monitoring and management.
E. coli was used as a fecal indicator bacterium as it is more reflective of recent fecal contamination from warm-blooded animals compared to fecal coliforms or total coliforms . Surface water samples were collected twice a month from May to October 2018, with three additional samplings from June to August for a total of 21 events at 5 mainstem and 6 tributary sites located throughout the middle section of the Musconetcong River watershed ( ). No samples were collected when stream channels were dry. The sites were selected based on past water quality monitoring activities showing the significant potential of fecal contaminations. Grab samples were collected aseptically into 1 L sterilized polypropylene bottles and were placed immediately in a cooler during transit. Enumeration of E. coli was based on an EPA-approved method using mColiBlue24 ® Broth (Hach Method 10029). In brief, samples were filtered through mixed cellulose ester membrane filters (0.45 mm, 47 mm, Hach, Loveland Colorado) and then soaked with mColiBlue24 ® Broth selective medium. The filters were subsequently incubated at 35 °C for 24 h. After incubation, colonies showing a blue/indigo color were recorded as E. coli . Tentative E. coli colonies were further verified using brilliant green bile (BGB) and lauryl tryptose broth (LTB). The final results were reported as colony forming units (CFU) per 100 mL of water samples. After the processing of the samples, we checked the weather data on the days when samples were collected and determined if certain samples were collected during a storm event based on the weather data. All E. coli data were uploaded to the Water Quality Portal ( https://www.waterqualitydata.us/ , accessed on 5 December 2021). The 2015 land use/land cover data were acquired from the New Jersey Department of Environmental Protection (NJDEP), Bureau of Geographic Information Systems. A modified Anderson coding system was used to categorize land use data . The total drainage area for the whole Musconetcong River watershed was 120,941.71 acres, with 53.2% forest, 21.2% urban, and 14.0% agricultural land. Four land use categories were extracted as explanatory variables for spatial stream network (SSN) modeling, including urban (1000), pasture (2120), forest (4000) and wetland (6000). Due to the sampling cost and length, the sample size in the current study was not particularly large (21 events). While more detailed land use land cover products, such as the USGS’s Multi-Resolution Land Characteristics (MRLC) land cover, are available, to avoid losing too many degrees of freedom in the empirical study, we elected to use the relatively coarse land use land cover categorization. Urban land use classification covered land use characteristics from rural single units to high-density multiple residential dwellings, commercial buildings, and industrial areas. Houses with various dwelling units or a high density in the study area were all connected to septic tanks instead of sewage pipes for waste disposal (Personal Communication with the Musconetcong Watershed Association). The study’s purpose was to investigate in general how urbanization within a watershed impacts water quality. The pasture subgroup was selected from the agricultural land use categories (2000), as failing septic systems and poor livestock management practices were previously listed as major threats to microbial water quality . Precipitation data were obtained from NJDEP, the Department of Water Monitoring & Standards ( https://njdep.rutgers.edu/rainfall/ , accessed on 5 December 2021). The rainfall threshold for a storm event was defined as 12.7 mm (0.5 inch) within 36 h, the same definition used in the Watershed Restoration and Protection Plan for the Musconetcong River Watershed created by the Rutgers Cooperative Extension Water Resources Program . A summary of analytical methods and data sources used in this study is shown in .
To explore the impact of land use in the watershed on water quality, we extended from traditional regression analysis to adopt the spatial stream network model in this study [ , , ]. Spatial stream network models are generalized linear mixed models allowing explanation of variance in observations with both fixed and random effects due to spatial autocorrelation among geographically observed events , which can be expressed as follows: y = Xβ + z + ε where y is a vector of observations, X is a matrix for fixed effects that explain the variance of the observations that can be captured by general spatial patterns, β is a vector of coefficients for X, z accounts for random effects resulting from spatial autocorrelation that cannot be explained by fixed effects, and ε is a vector of independent random errors . The spatial autocorrelation structures that describe the random effects can be modeled with linear with sill, Mariah, exponential, Epanechnikov, and spherical models . Parameters for spatial autocovariance functions include nugget, partial sill, and range. The nugget of the autocovariance function describes the portion of variance that cannot be explained, the partial sill accounts for the variance that depends on the distance among observations, and the range indicates the minimal distance at which observations are no longer spatially correlated . SSN models can also incorporate autocovariance models based on Euclidean distance, accounting for terrestrial and atmospheric factors that are stream-independent . Both tail-up and tail-down autocovariance structures were established to estimate spatial autocorrelation. In tail-up models, moving average functions run upstream only and split at confluences. Spatial weights are required to estimate the proportion of upstream tributary influence based on flow volume, watershed area, or other relevant attributes . Tail-up models are only applicable to flow-connected sites, where water directly flows from an upstream to a downstream site characterized by passive downstream movement, e.g., temperature, bacteria, or sediment . In tail-down models, moving average functions run in the downstream direction unilaterally until reaching the most downstream location of the stream network, indicating the possibility of autocorrelation among all locations . In addition to flow-connected sites, tail-down models can be applied to flow-unconnected sites, where some upstream movement is expected to facilitate the connectivity for a given attribute, e.g., fish or macroinvertebrates . In order to perform spatial stream network (SSN) modeling, an SSN object was first created using STARS geoprocessing toolset in ArcGIS 10.8 (ESRI, Redlands, CA, USA). STARS is used to build a landscape network (LSN) as a personal geodatabase to represent spatial relationships, such as flow connectivity, direction, and distance and transform the geometry, attribute data, and topological relationships among features of GIS datasets into an SSN object that can be easily accessed and analyzed in R statistical software with the ssn package for SSN modeling. The process included building a landscape network (LSN), creating reach contributing areas , calculating RCA attributes and area, accumulating watershed attributes, incorporating the sampling locations into the LSN, calculating watershed attributes, and calculating upstream distance, segment proportion influence, and additive function . In the watershed of the current study, we obtained a total of 427 nodes in the stream network. The entire process of generating the SSN object was automated, and once generated in ArcGIS, the SSN object was then exported to R statistical software for SSN modeling. In this study, the ssn package in R was used to fit spatial stream network models to the observation data based on 11 sampling locations. Two models were established and calibrated. The first model used the geometric mean concentrations of E. coli for all sampling events as the response variable; the second model used the geometric mean concentrations of E. coli during storm events only as the response variable. Shapiro–Wilk tests were used to test the normality of the distributions of the original values . If the normality assumption was not satisfied, the values of geometric mean concentrations were log-transformed. A nonspatial model based on an ordinary least squares regression and two spatial models based on Euclidean and stream distance, respectively, were established to incorporate four upstream watershed attributes as explanatory variables, including urban, pasture, forest, and wetland. Percentage upstream land use values were derived from dividing the total accumulated area of upstream land use by the total upstream catchment area processed and generated via the STARS toolset. Variance inflation factors (VIF) were calculated to check multicollinearity among the four land use variables. Tail-up models were applied in the spatial autocovariance functions for the SSN modeling based on stream distance, as the movement of bacteria along the stream networks is characterized by passive downstream transport . Likelihood approaches were utilized to estimate the regression coefficients b for the matrix of fixed effects . Maximum likelihood (ML) estimators have been shown to be greatly biased when the number of observations is small , and thus restricted maximum likelihood (REML) was used. An alpha value of 0.05 was used to evaluate whether the relationship between watershed variables (urban, agricultural, forest, or wetland) and E. coli concentrations in the models was statistically significant. The root mean square prediction error (RMSPE) of the leave one out cross validation (LOOCV) was used as the criterion to compare the predictive capability among models. LOOCV was performed by excluding one observation at a time. Submodels were calculated and compared with the excluded samples . A lower RMSPE of the LOOCV indicated a better model performance . The coefficient of determination (R 2 ) was used to measure the proportion of variance in the observations explained by the fixed or random effects of each model.
One of the practical uses for establishing the empirical SSN model for the Musconetcong River watershed is that we can use the model to predict the concentration of E. coli along the river. Predictions were performed for the geometric mean concentrations of E. coli for all sampling events and during storm events based on Euclidean distance and stream distance models, respectively. A total of 816 predictive sites were assigned evenly across the Musconetcong River stream network at a 300-m interval. Determination of predictive sites was conducted using the STARS geoprocessing toolset in ArcGIS 10.8. Predictions of E. coli concentrations were performed in the R statistical software with SSN package . The results were illustrated on stream maps illustrating both the log 10 geometric mean concentrations of E. coli . The capability of predicting the concentration of E. coli provides both a means to assess modeling performance and a supplement to the regular sampling routine for water quality monitoring and management.
3.1. Microbial Water Quality and Land Use All 21 sampling events were recorded for most of the locations, except for T1 (3) and T3 (19). Among them, three were defined as storm events (accumulated rainfall within 36 h was 13.2, 22.9, and 41.9 mm, respectively), except for T1 (one event, accumulated rainfall was 22.9 mm). The geometric mean concentrations of E. coli for all sampling events among the mainstem sites were similar (240 to 296 CFU/100 mL), while those among the tributary sites had a greater variation (73.3 to 1355.6 CFU/100 mL). Four of the six tributary sites exceeded the geometric mean threshold of E. coli (126 CFU/100 mL) compared to one of the five mainstem sites for all sampling events ( ). During storm events, the geometric mean concentrations of E. coli among the mainstem sites ranged from 303 to 811 CFU/100 mL), while those among the tributary sites also had a greater variation (585.7 to 7300 CFU/100 mL). All of the sites exceeded the geometric mean threshold of E. coli ( ). The upstream land use characteristics of sampling locations were similar among the mainstream sites, while those of tributary sites exhibited a greater variety as each site demonstrated unique patterns in this suburban mixed-land-use watershed ( ). The percentage of upstream urban land ranged from 21.3% to 22.5% and from 8.2% to 57.4% for the mainstem and tributary sites, respectively. The percentage of upstream pasture varied from 0.0% to 0.2% and 0.1% to 20.3% for the mainstem and tributary sites, respectively. The percentage of upstream forest ranged from 41.0% to 42.9% and from 14.6% to 58.7% for the mainstem and tributary sites, respectively. The percentage of upstream wetland ranged from 6.3% to 6.7% and 0.3% to 6.4% for the mainstem and tributary sites, respectively. Overall, pasture land dominated the study area, with urban lands clustered toward the upper portion. Forest land is concentrated toward the upstream area of tributaries in the study area ( ). 3.2. Spatial Stream Network Modeling Three individual spatial stream network (SSN) models with three distinct autocovariance structures were constructed for each response variable in this study, including no spatial correlation (ordinary least squares), straight-line distance (Euclidean distance), and along-channel in-stream distance (stream distance). Variance inflation factors (VIF) for all four land use variables were below 2, indicating that no multicollinearity among them was detected. The original values of the geometric mean concentrations of E. coli were not normally distributed. Therefore, the original values were log-transformed to satisfy the normal distribution assumption. shows the SSN model autocovariance structures, coefficients for explanatory variables, and performance. Applying SSN models identified that upstream urban land was positively and significantly associated with the log 10 geometric mean concentrations of E. coli for all events and during storm events, respectively ( p < 0.05). Upstream pasture land was positively and significantly correlated with the log 10 geometric mean concentrations of E. coli during storm events only ( p < 0.05). Although insignificantly, the geometric mean concentrations of E. coli for all sampling events demonstrated negative correlations with upstream wetland ( p = 0.06). SSN models improved the overall coefficient of determination (R 2 ) as the additional portion of variance can be explained by the random effects attributed to the spatial autocorrelation. In fact, based on stream distance, the overall SSN models explained nearly 100% of variance compared to the ordinary least squares models, with 20% of variance being unaccounted for. The lowest level of unexplained variance was found for SSN models based on stream distance for both response variables (Nugget < 0.0001). Prediction errors also improved from non-spatial to spatial models. SSN models based on Euclidean distance showed the lowest value of LOOCV RMSPE for log 10 geometric mean concentrations of E. coli for all sampling events (1.110), while the lowest value of LOOCV RMSPE was seen for log 10 geometric mean concentrations of E. coli during storm events modeled based on stream distance (1.290). Overall, the distance with no spatial autocorrelation observed for log 10 geometric mean concentrations of E. coli for all sampling events was shorter than that for the storm events based on either Euclidean (Range 2.31 vs. 3.11 km) or stream distance (Range 2.87 vs. 3.37). The best stream distance autocovariance functions were linear with sill tail-up for both log 10 geometric mean concentrations of E. coli for all sampling events and during storm events. 3.3. Predicting E. coli Concentrations Euclidean distance and stream distance models were chosen to predict the log 10 geometric means concentrations of E. coli for all events and during storm events, respectively, due to lower root mean square percentage error (RMSPE) values ( ). shows both the geometric mean concentrations of E. coli of the 11 sampling locations (stars) as well as the predicted geometric mean concentrations of E. coli (solid circles) in the study area. Overall, the predicted geometric mean concentrations of E. coli were higher during storm events ( b) than for all sampling events ( a). Predictive values for the geometric mean concentrations of E. coli ranged from 1 to 5012 CFU/100 mL and from 43 to 30,903 CFU/100 mL for all events and during storm events, respectively.
All 21 sampling events were recorded for most of the locations, except for T1 (3) and T3 (19). Among them, three were defined as storm events (accumulated rainfall within 36 h was 13.2, 22.9, and 41.9 mm, respectively), except for T1 (one event, accumulated rainfall was 22.9 mm). The geometric mean concentrations of E. coli for all sampling events among the mainstem sites were similar (240 to 296 CFU/100 mL), while those among the tributary sites had a greater variation (73.3 to 1355.6 CFU/100 mL). Four of the six tributary sites exceeded the geometric mean threshold of E. coli (126 CFU/100 mL) compared to one of the five mainstem sites for all sampling events ( ). During storm events, the geometric mean concentrations of E. coli among the mainstem sites ranged from 303 to 811 CFU/100 mL), while those among the tributary sites also had a greater variation (585.7 to 7300 CFU/100 mL). All of the sites exceeded the geometric mean threshold of E. coli ( ). The upstream land use characteristics of sampling locations were similar among the mainstream sites, while those of tributary sites exhibited a greater variety as each site demonstrated unique patterns in this suburban mixed-land-use watershed ( ). The percentage of upstream urban land ranged from 21.3% to 22.5% and from 8.2% to 57.4% for the mainstem and tributary sites, respectively. The percentage of upstream pasture varied from 0.0% to 0.2% and 0.1% to 20.3% for the mainstem and tributary sites, respectively. The percentage of upstream forest ranged from 41.0% to 42.9% and from 14.6% to 58.7% for the mainstem and tributary sites, respectively. The percentage of upstream wetland ranged from 6.3% to 6.7% and 0.3% to 6.4% for the mainstem and tributary sites, respectively. Overall, pasture land dominated the study area, with urban lands clustered toward the upper portion. Forest land is concentrated toward the upstream area of tributaries in the study area ( ).
Three individual spatial stream network (SSN) models with three distinct autocovariance structures were constructed for each response variable in this study, including no spatial correlation (ordinary least squares), straight-line distance (Euclidean distance), and along-channel in-stream distance (stream distance). Variance inflation factors (VIF) for all four land use variables were below 2, indicating that no multicollinearity among them was detected. The original values of the geometric mean concentrations of E. coli were not normally distributed. Therefore, the original values were log-transformed to satisfy the normal distribution assumption. shows the SSN model autocovariance structures, coefficients for explanatory variables, and performance. Applying SSN models identified that upstream urban land was positively and significantly associated with the log 10 geometric mean concentrations of E. coli for all events and during storm events, respectively ( p < 0.05). Upstream pasture land was positively and significantly correlated with the log 10 geometric mean concentrations of E. coli during storm events only ( p < 0.05). Although insignificantly, the geometric mean concentrations of E. coli for all sampling events demonstrated negative correlations with upstream wetland ( p = 0.06). SSN models improved the overall coefficient of determination (R 2 ) as the additional portion of variance can be explained by the random effects attributed to the spatial autocorrelation. In fact, based on stream distance, the overall SSN models explained nearly 100% of variance compared to the ordinary least squares models, with 20% of variance being unaccounted for. The lowest level of unexplained variance was found for SSN models based on stream distance for both response variables (Nugget < 0.0001). Prediction errors also improved from non-spatial to spatial models. SSN models based on Euclidean distance showed the lowest value of LOOCV RMSPE for log 10 geometric mean concentrations of E. coli for all sampling events (1.110), while the lowest value of LOOCV RMSPE was seen for log 10 geometric mean concentrations of E. coli during storm events modeled based on stream distance (1.290). Overall, the distance with no spatial autocorrelation observed for log 10 geometric mean concentrations of E. coli for all sampling events was shorter than that for the storm events based on either Euclidean (Range 2.31 vs. 3.11 km) or stream distance (Range 2.87 vs. 3.37). The best stream distance autocovariance functions were linear with sill tail-up for both log 10 geometric mean concentrations of E. coli for all sampling events and during storm events.
Euclidean distance and stream distance models were chosen to predict the log 10 geometric means concentrations of E. coli for all events and during storm events, respectively, due to lower root mean square percentage error (RMSPE) values ( ). shows both the geometric mean concentrations of E. coli of the 11 sampling locations (stars) as well as the predicted geometric mean concentrations of E. coli (solid circles) in the study area. Overall, the predicted geometric mean concentrations of E. coli were higher during storm events ( b) than for all sampling events ( a). Predictive values for the geometric mean concentrations of E. coli ranged from 1 to 5012 CFU/100 mL and from 43 to 30,903 CFU/100 mL for all events and during storm events, respectively.
4.1. Spatial Stream Network Modeling Performance Spatial stream network modeling provides an empirical approach for urban and suburban water quality stakeholders to analyze the spatial distribution of parameters without the need to first understand the underlying hydrological and biogeochemical processes that lead to water quality impairment. The application of SSN models has been introduced to a wide range of water quality indicators, such as temperature , dissolved oxygen , total phosphorus , and macroinvertebrates . However, the modeling of fecal indicator bacteria using SSN models had not been documented until recently . Holcomb and Messier modeled fecal coliforms in a mixed-land-use watershed in North Carolina and identified agricultural land use, forest cover, antecedent precipitation, and temperature as being strongly associated with mean fecal coliform concentrations. Neill and Tetzlaff used E. coli as a fecal indicator bacterium in an agriculture-dominant watershed with sporadic urban development in Scotland and found that the anthropogenic impact index (lumped indicator for potential contamination from human point sources) was significantly correlated with 5th, 50th, and 95th percentile E. coli concentrations. Similar to Neill and others (2018), modeling based on stream distance in this study demonstrated improvement in the total R square and root mean square percentage error (RMSPE) for geometric mean concentrations of E. coli during storm events. However, only an improvement in the total R square was observed for the model based on stream distance for geometric mean concentrations of E. coli for all sampling events. This may be due to limitations regarding the number of sites as well as the locations of the mainstem sites, as they were relatively highly clustered compared to the overall watershed. An improvement in RMSPE was not observed in Holcomb et al. (2018) either. In this study, the best autocovariance functions were linear with sill with stream distance for both log 10 geometric mean concentrations of E. coli for all sampling events and during storm events, indicating a pattern in which the variability among observations increases linearly with the separation distance until it reaches the maximum difference . Overall, the range for the SSN models for storm events was greater than that for all sampling events, and the stream distance models had a greater range than the Euclidean distance models did, indicating heavier impacts on E. coli concentrations from upstream areas during storm events. 4.2. Suburban Land Use and Microbial Water Quality Various factors could lead to the deterioration of microbial water quality in urban and suburban waterways . In this study, upstream urban land was identified as being positively correlated with either log 10 geometric mean concentrations of E. coli for all sampling events or during storm events only. Using the same SSN modeling approach, Neill and Tetzlaff identified human (leaking sewage pipes and failing septic tanks) and farmyard sources as being significantly associated with 5th, 50th, and 95th percentile concentrations of E. coli . A variety of mechanisms, such as failing onsite wastewater treatment systems (i.e., septic tanks), combined sewer overflows (CSOs), sanitary sewer overflows (SSOs) and urban runoffs can introduce human fecal contamination into adjacent waterways in urban and suburban watersheds . Increased risks of childhood emergency department visits and infectious diarrhea associated with combined sewer overflow and septic sites have been documented . In addition to urban land use, upstream pasture land was found to be positively and significantly associated with log 10 geometric mean concentrations of E. coli during storm events in this study. Holcomb et al. (2018) also reported similar significant relationships between agricultural land use and fecal coliform. Ill-considered agricultural practices in suburban watersheds, such as concentrated animal feeding operations, overgrazing and manure applications could lead to water quality deterioration . Davies-Colley et al. (2004) also demonstrated that a dairy cow herd produced more than 50,000 CFU/100 mL of E. coli when crossing a stream. A significant relationship was found between upstream pastureland use and E. coli for storm events only in this study, illustrating the impact of agricultural runoff on microbial water quality during storm events. For instance, higher concentrations of E. coli and Salmonella were identified in irrigation ponds at two produce farms after rain events . Noteworthily, a negative association (although only marginally significant, p = 0.06) was identified between upstream wetlands and geometric mean concentrations of E. coli for all sampling events. Wetlands can provide essential ecosystem services, such as water purification and runoff reductions . Reductions in fecal indicator bacteria or pathogens provided by wetlands have been well documented [ , , ]. Hsu et al. (2017) reported an average 22.3% reduction in E. coli across two wetlands receiving inflow from urban waterways . The results from this study also reinforce the concept of using wetlands as the best management practice for microbial water quality restoration. For example, a variety of constructed wetlands have been implemented to treat domestic and agricultural wastewater as well as urban stormwater runoff . 4.3. Extreme Weather Conditions and Microbial Water Quality All of the geometric mean concentrations of E. coli during storm events exceeded the geometric mean threshold of E. coli of 126 CFU/100 mL in this study. Extreme weather events, such as heavy precipitation, have been well documented to degrade water quality by elevating the concentrations of FIB in urban and suburban waterways through stormwater runoff and combined and sanitary overflows . These heavy rainfall events significantly contributed to waterborne disease outbreaks in the United States and Canada . In the United States, among the 548 reported outbreaks analyzed from 1948 through 1994, 68% of them were preceded by the highest 20% of precipitation events . In Canada, the heaviest 7% of precipitation events increased the relative odds of an outbreak by 2.3-fold . In fact, more than 400,000 cases of acute gastrointestinal illness (AGI) were attributed to a drinking water treatment plant overwhelmed by high turbidity load after a period of heavy precipitation in Milwaukee, Wisconsin in 1993 . Downstream of the Musconetcong River in Philadelphia, where the Delaware River serves as the drinking water source, a significant increase in waterborne AGI following precipitation above the 95th percentile was documented . Under the current trend of global climate change, the National Climate Assessment projected an almost 50% increase in the total annual precipitation falling in the heaviest cases (1%) by the late 21st century under the higher scenario (RCP 8.5) in the northeastern United States . This will further increase the frequency and intensity of urban stormwater runoffs and combined and sanitary overflows, degrading microbial water quality and increasing potential public health risk in the Musconetcong River and downstream drainage area in the future if no additional water quality protection action is implemented. The land use variables identified from the SSN modeling results can be taken into consideration when developing water quality restoration or climate resiliency plans to address water quality and public health concerns in urban and suburban areas, such as establishing municipal separate storm sewer system (MS4), expanding public sanitary sewer systems, regulating private sewage disposal practice, and promoting the use of green infrastructure. 4.4. Predicting Fecal Contamination Hot Spots and Future Directions Since field water sampling is time-consuming and costly, water quality modeling can serve as a cost-effective and timely approach to enhance existing monitoring programs in conjunction with actual field samples. It can also provide stakeholders with a mean to quickly identify “hot spots” during the site selection process for further water monitoring and water quality enhancement actions. Spatial stream network predictive modeling provides a feasible approach to predict riverine E. coli concentrations with a predetermined stream-distance interval. The predicted results can be used to prioritize sites for further monitoring and subsequent restoration practices. Combined with local knowledge, the modeling results can be tailored to form a localized site-specific management approach. In this study, predictive sites near T1 in a catchment basin with a high percentage of urban land showed elevated E. coli concentrations for all sampling events and during storm events. Another example is shown in the next four predictive sites downstream of T4. The predicted value for E. coli for all sampling events was below the geometric mean threshold (126 CFU/100 mL). However, the predicted value for storm events was above the single-sample maximum threshold (235 CFU/100 mL), indicating the significant impact of agricultural runoff as a result of heavy precipitation on microbial water quality. Different localized water quality restoration approaches should be developed for both areas, such as improving the maintenance of septic systems in urbanized areas and establishing riparian buffers in agriculture-dominant sites. For TMDL process development, SSN models become a cost-effective approach along with field sampling programs for water quality stakeholders to engage in watershed management with the support of readily available land use land cover data, open-source R packages, and geographic information systems. In fact, the current study area was located in one of the TMDL segments of the Musconetcong River, and additional segments are located both upstream and downstream of the current study area . Therefore, the same SSN approach can be extended to additional TMDL segments within the Musconetcong River watershed to provide a comprehensive understanding for integrated water quality management (NJDEP, 2003). Incorporating additional sampling locations is essential to encompass a larger watershed to improve SSN modeling. Money and Carter suggested “a minimum of 10–50 data points should exist to construct a correlation model depending on the watershed size”. In our current study, because of cost and the size of our watershed, we only collected 21 data points to establish the model. The sample size is on the small end and does not have sufficient degrees of freedom for us to set up a model training and testing routine. While water monitoring is an ongoing endeavor for environmental management teams at both the local and state levels, we hope in the future that expanding sampling locations will allow model validation to encompass both training and validation datasets. It was also suggested that the placement of sampling locations in relation to the whole watershed could impact the outcome of the spatial modeling . For instance, a better overall error map may be produced by placing sites close to the origins of tributaries, while a good estimate of tail-up autocovariance functions could be generated by selecting sites between confluences. In addition, a spatial stream predictive model in conjunction with other field and laboratory observations (e.g., sanitary survey or microbial source tracking) can identify sites with greater contamination potential and sources of fecal contamination. This is essential for a thorough watershed management plan to address impairment and TMDL source reductions. Although beyond the scope of SSN modeling, temporal variations also affect the outcome when modeling microbial water quality. Holcomb and Messier included a time component in their geostatistical models and obtained improved prediction errors than models considering spatial effects only. The effects of precipitation patterns could also be investigated when the temporal effects are incorporated into spatial modeling to provide a comprehensive understanding of microbial water quality dynamics in response to a variety of environmental variables and weather scenarios. Overall, the current study provides critical insights into assessing the amount of fecal contamination to guide further monitoring and management activities and serves as potential water quality modeling framework for urban and suburban watersheds in the USA and beyond.
Spatial stream network modeling provides an empirical approach for urban and suburban water quality stakeholders to analyze the spatial distribution of parameters without the need to first understand the underlying hydrological and biogeochemical processes that lead to water quality impairment. The application of SSN models has been introduced to a wide range of water quality indicators, such as temperature , dissolved oxygen , total phosphorus , and macroinvertebrates . However, the modeling of fecal indicator bacteria using SSN models had not been documented until recently . Holcomb and Messier modeled fecal coliforms in a mixed-land-use watershed in North Carolina and identified agricultural land use, forest cover, antecedent precipitation, and temperature as being strongly associated with mean fecal coliform concentrations. Neill and Tetzlaff used E. coli as a fecal indicator bacterium in an agriculture-dominant watershed with sporadic urban development in Scotland and found that the anthropogenic impact index (lumped indicator for potential contamination from human point sources) was significantly correlated with 5th, 50th, and 95th percentile E. coli concentrations. Similar to Neill and others (2018), modeling based on stream distance in this study demonstrated improvement in the total R square and root mean square percentage error (RMSPE) for geometric mean concentrations of E. coli during storm events. However, only an improvement in the total R square was observed for the model based on stream distance for geometric mean concentrations of E. coli for all sampling events. This may be due to limitations regarding the number of sites as well as the locations of the mainstem sites, as they were relatively highly clustered compared to the overall watershed. An improvement in RMSPE was not observed in Holcomb et al. (2018) either. In this study, the best autocovariance functions were linear with sill with stream distance for both log 10 geometric mean concentrations of E. coli for all sampling events and during storm events, indicating a pattern in which the variability among observations increases linearly with the separation distance until it reaches the maximum difference . Overall, the range for the SSN models for storm events was greater than that for all sampling events, and the stream distance models had a greater range than the Euclidean distance models did, indicating heavier impacts on E. coli concentrations from upstream areas during storm events.
Various factors could lead to the deterioration of microbial water quality in urban and suburban waterways . In this study, upstream urban land was identified as being positively correlated with either log 10 geometric mean concentrations of E. coli for all sampling events or during storm events only. Using the same SSN modeling approach, Neill and Tetzlaff identified human (leaking sewage pipes and failing septic tanks) and farmyard sources as being significantly associated with 5th, 50th, and 95th percentile concentrations of E. coli . A variety of mechanisms, such as failing onsite wastewater treatment systems (i.e., septic tanks), combined sewer overflows (CSOs), sanitary sewer overflows (SSOs) and urban runoffs can introduce human fecal contamination into adjacent waterways in urban and suburban watersheds . Increased risks of childhood emergency department visits and infectious diarrhea associated with combined sewer overflow and septic sites have been documented . In addition to urban land use, upstream pasture land was found to be positively and significantly associated with log 10 geometric mean concentrations of E. coli during storm events in this study. Holcomb et al. (2018) also reported similar significant relationships between agricultural land use and fecal coliform. Ill-considered agricultural practices in suburban watersheds, such as concentrated animal feeding operations, overgrazing and manure applications could lead to water quality deterioration . Davies-Colley et al. (2004) also demonstrated that a dairy cow herd produced more than 50,000 CFU/100 mL of E. coli when crossing a stream. A significant relationship was found between upstream pastureland use and E. coli for storm events only in this study, illustrating the impact of agricultural runoff on microbial water quality during storm events. For instance, higher concentrations of E. coli and Salmonella were identified in irrigation ponds at two produce farms after rain events . Noteworthily, a negative association (although only marginally significant, p = 0.06) was identified between upstream wetlands and geometric mean concentrations of E. coli for all sampling events. Wetlands can provide essential ecosystem services, such as water purification and runoff reductions . Reductions in fecal indicator bacteria or pathogens provided by wetlands have been well documented [ , , ]. Hsu et al. (2017) reported an average 22.3% reduction in E. coli across two wetlands receiving inflow from urban waterways . The results from this study also reinforce the concept of using wetlands as the best management practice for microbial water quality restoration. For example, a variety of constructed wetlands have been implemented to treat domestic and agricultural wastewater as well as urban stormwater runoff .
All of the geometric mean concentrations of E. coli during storm events exceeded the geometric mean threshold of E. coli of 126 CFU/100 mL in this study. Extreme weather events, such as heavy precipitation, have been well documented to degrade water quality by elevating the concentrations of FIB in urban and suburban waterways through stormwater runoff and combined and sanitary overflows . These heavy rainfall events significantly contributed to waterborne disease outbreaks in the United States and Canada . In the United States, among the 548 reported outbreaks analyzed from 1948 through 1994, 68% of them were preceded by the highest 20% of precipitation events . In Canada, the heaviest 7% of precipitation events increased the relative odds of an outbreak by 2.3-fold . In fact, more than 400,000 cases of acute gastrointestinal illness (AGI) were attributed to a drinking water treatment plant overwhelmed by high turbidity load after a period of heavy precipitation in Milwaukee, Wisconsin in 1993 . Downstream of the Musconetcong River in Philadelphia, where the Delaware River serves as the drinking water source, a significant increase in waterborne AGI following precipitation above the 95th percentile was documented . Under the current trend of global climate change, the National Climate Assessment projected an almost 50% increase in the total annual precipitation falling in the heaviest cases (1%) by the late 21st century under the higher scenario (RCP 8.5) in the northeastern United States . This will further increase the frequency and intensity of urban stormwater runoffs and combined and sanitary overflows, degrading microbial water quality and increasing potential public health risk in the Musconetcong River and downstream drainage area in the future if no additional water quality protection action is implemented. The land use variables identified from the SSN modeling results can be taken into consideration when developing water quality restoration or climate resiliency plans to address water quality and public health concerns in urban and suburban areas, such as establishing municipal separate storm sewer system (MS4), expanding public sanitary sewer systems, regulating private sewage disposal practice, and promoting the use of green infrastructure.
Since field water sampling is time-consuming and costly, water quality modeling can serve as a cost-effective and timely approach to enhance existing monitoring programs in conjunction with actual field samples. It can also provide stakeholders with a mean to quickly identify “hot spots” during the site selection process for further water monitoring and water quality enhancement actions. Spatial stream network predictive modeling provides a feasible approach to predict riverine E. coli concentrations with a predetermined stream-distance interval. The predicted results can be used to prioritize sites for further monitoring and subsequent restoration practices. Combined with local knowledge, the modeling results can be tailored to form a localized site-specific management approach. In this study, predictive sites near T1 in a catchment basin with a high percentage of urban land showed elevated E. coli concentrations for all sampling events and during storm events. Another example is shown in the next four predictive sites downstream of T4. The predicted value for E. coli for all sampling events was below the geometric mean threshold (126 CFU/100 mL). However, the predicted value for storm events was above the single-sample maximum threshold (235 CFU/100 mL), indicating the significant impact of agricultural runoff as a result of heavy precipitation on microbial water quality. Different localized water quality restoration approaches should be developed for both areas, such as improving the maintenance of septic systems in urbanized areas and establishing riparian buffers in agriculture-dominant sites. For TMDL process development, SSN models become a cost-effective approach along with field sampling programs for water quality stakeholders to engage in watershed management with the support of readily available land use land cover data, open-source R packages, and geographic information systems. In fact, the current study area was located in one of the TMDL segments of the Musconetcong River, and additional segments are located both upstream and downstream of the current study area . Therefore, the same SSN approach can be extended to additional TMDL segments within the Musconetcong River watershed to provide a comprehensive understanding for integrated water quality management (NJDEP, 2003). Incorporating additional sampling locations is essential to encompass a larger watershed to improve SSN modeling. Money and Carter suggested “a minimum of 10–50 data points should exist to construct a correlation model depending on the watershed size”. In our current study, because of cost and the size of our watershed, we only collected 21 data points to establish the model. The sample size is on the small end and does not have sufficient degrees of freedom for us to set up a model training and testing routine. While water monitoring is an ongoing endeavor for environmental management teams at both the local and state levels, we hope in the future that expanding sampling locations will allow model validation to encompass both training and validation datasets. It was also suggested that the placement of sampling locations in relation to the whole watershed could impact the outcome of the spatial modeling . For instance, a better overall error map may be produced by placing sites close to the origins of tributaries, while a good estimate of tail-up autocovariance functions could be generated by selecting sites between confluences. In addition, a spatial stream predictive model in conjunction with other field and laboratory observations (e.g., sanitary survey or microbial source tracking) can identify sites with greater contamination potential and sources of fecal contamination. This is essential for a thorough watershed management plan to address impairment and TMDL source reductions. Although beyond the scope of SSN modeling, temporal variations also affect the outcome when modeling microbial water quality. Holcomb and Messier included a time component in their geostatistical models and obtained improved prediction errors than models considering spatial effects only. The effects of precipitation patterns could also be investigated when the temporal effects are incorporated into spatial modeling to provide a comprehensive understanding of microbial water quality dynamics in response to a variety of environmental variables and weather scenarios. Overall, the current study provides critical insights into assessing the amount of fecal contamination to guide further monitoring and management activities and serves as potential water quality modeling framework for urban and suburban watersheds in the USA and beyond.
In this study, we successfully applied spatial stream networks (SSN) to model elevated concentrations of E. coli in the suburban mixed-land-use Musconetcong River watershed in response to upstream land use attributes, including urban, pasture, forest, and wetland. The SSN model is essentially a spatial statistical model designed for modeling the statistical relationships between variables on stream networks . The modeling results suggest that upstream urban land was positively and significantly associated with log 10 geometric mean concentrations of E. coli for all sampling events and during storm events, respectively. Upstream pasture land was positively and significantly correlated with log 10 geometric mean concentrations of E. coli during storm events only. Although only marginally significantly, the log 10 geometric mean concentrations of E. coli for all sampling events demonstrated a negative relationship with upstream wetland land use as per the SSN model calibration. Applying SSN modeling based on spatial distance also demonstrated improved model performance over non-spatial models. The SSN modeling results concur with previous findings that anthropogenic sources represent the main threat to microbial water quality in the Musconetcong River watershed. The prediction of E. coli concentrations by SSN models identified potential hot spots prone to water quality deteriorations. Future directions could benefit from incorporating additional sampling locations beyond the current section of the watershed and introducing temporal effects in the spatial modeling. Nevertheless, with the support of publicly available watershed attribute data, a pre-established ArcGIS toolset and the open-source R package, we believe that SSN models can be employed by other urban and suburban water quality stakeholders to assist in their water quality monitoring and restoration needs when the resources are constrained.
|
Transcriptomic Signatures of Single-Suture Craniosynostosis Phenotypes | e3ef8dc4-8322-4873-81e6-a148b8d5d5e3 | 10049207 | Suturing[mh] | Craniosynostosis is the premature fusion of one or more of the calvarial sutures, that occurs in syndromic and non-syndromic forms in approximately 1 in 2100–2500 live births [ , , ]. The syndromic craniosynostoses are hereditary conditions, defined as having craniosynostosis associated with reproducible features involving non-craniosynostosis phenotypes, such as limb or facial malformations. Craniosynostosis has been associated with over 150 different syndromes ; however, isolated single-suture fusions account for approximately 85% of all patients diagnosed with craniosynostosis . Single suture craniosynostosis (SSC) is defined as premature fusion of one of the four major sutures of the calvaria (metopic, sagittal, coronal, or lambdoid). Single-suture sagittal and metopic craniosynostosis have a nearly four-fold increased incidence among males, whereas in single-suture coronal cases there is a small increased incidence among females . Single-suture craniosynostosis (SSC) follows Mendelian patterns of inheritance in some families, with approximately 6–8% of patients having a positive family history that is consistent with autosomal dominant transmission . Cases with family recurrence usually involve the same suture, but there have been large pedigrees with coronal or sagittal synostosis that exhibit significant intrafamilial variability . Beyond strictly genetic causes, abnormal in utero mechanical forces [ , , , , ] and in utero exposure to valproic acid , are risk factors for the development of craniosynostosis. Craniosynostosis can be accompanied by a variety of medical conditions, most commonly including hydrocephalus, increased intracranial pressure, intellectual disabilities, and visual or hearing deficits . Therefore, craniosynostosis is one of the most clinically significant craniofacial disorders due to its prevalence, associated morbidity, complex surgical needs, and related burden on families and healthcare. Although the molecular pathogenesis for the majority of rare syndromic forms of craniosynostosis has been identified, the causes of the more common isolated single-suture forms remain effectively unknown. RNA sequencing technology has the capability to uncover novel genes and pathways related to single-suture craniosynostosis. To date there has not been a transcriptome-wide evaluation of gene expression changes in single-suture phenotype of craniosynostosis. Previous investigations of transcriptomic changes in craniosynostosis have largely utilized microarray or RT-qPCR-based approaches, which are biased toward previously identified gene associations. Conversely, RNA sequencing analysis allows for an unbiased evaluation of the whole transcriptome, with the potential to identify unknown gene associations and patterns of expression across gene family groups and biological pathways. This study will evaluate the craniosynostosis transcriptome across four single-suture phenotypes (coronal, metopic, sagittal, lambdoid) in a cohort of patients (N = 388 cases, 85 controls) recruited through four study sites in the United States between 2002 and 2016, with RNA sequencing performed on primary calvarial cell lines derived from these patients. The aim of this paper is to identify transcriptomic signatures (including sex-specific associations) of single-suture craniosynostosis phenotypes, and we hypothesize that each phenotype will be associated with a unique subset of genes and pathways, with differences related to patient sex.
2.1. Study Participant Characteristics RNA was sequenced from cultured primary calvarial cells derived from 85 controls and 388 craniosynostosis cases, that included 80 coronal, 196 sagittal, 92 metopic, and 20 lambdoid phenotypes ( ), as well as 85 control samples without craniosynostosis. As expected, the coronal phenotype was over-represented in females, and the metopic and sagittal phenotypes were over-represented in males (chi-squared test p -value < 0.05, ). There were no statistically significant differences across craniosynostosis phenotypes for proband age, at the time of sample collection. However, the age of control patients at time of collection was significantly different from all four craniosynostosis groups ( p < 0.001 for each phenotype vs. control comparisons, ANOVA with Tukey’s post hoc test) ( ). There was no statistically significant difference in cell culture time between controls and craniosynostosis phenotypes (ANOVA p > 0.05) ( ). 2.2. DEGs and Pathway Analysis by Phenotype In our primary model, we identified 72 DEGs whose expression was associated with the coronal phenotype of craniosynostosis, 33 DEGs associated with lambdoid craniosynostosis, 103 DEGs associated with metopic craniosynostosis, and 90 DEGs associated with sagittal craniosynostosis, compared to controls without craniosynostosis. The full results are presented in . Generally, we observed an even distribution of DEGs that were increased and decreased in craniosynostosis phenotypes vs. controls ( ). The top three increased DEGs associated with each phenotype were, PAX3 , GATA3 , and OLR1 (coronal); PAX3 , TFAP2A , and ALX1 (metopic); SHOX2 , HOXB13 , and CPAMD8 (sagittal); and HOXB13 , HOXB3 , and PCDH17 (lambdoid). The top three decreased DEGs associated with each phenotype were, NRN1 , MCF2L , and STEAP4 (coronal); MCF2L , SMOC2 , and ST8SIA2 (metopic); CCKAR , NRN1 , and IRX2 (sagittal); and TRH , DLX1 , and XIRP1 (lambdoid). There was a single gene, EFHD1 , that was significantly decreased across all four phenotypes of craniosynostosis compared to controls ( F). The lambdoid and metopic phenotypes had the highest percentage of unique DEGs (lambdoid = 20/33 unique DEGs, metopic = 62/103 unique DEGs) that were not significantly associated with other craniosynostosis phenotypes. There were 51 genes (57% of sagittal DEGs) uniquely associated with sagittal craniosynostosis, and 33 genes (46% of coronal DEGs) uniquely associated with coronal craniosynostosis ( F). Using rotational gene set testing, we identified three significant pathways (FDR < 0.05) whose expression was associated with the coronal phenotypes ( ). All three pathways exhibited increased expression compared to controls, and are categorized as metabolism pathways by KEGG. 2.3. Sex-Stratified DEGs and Pathways by Phenotype In our models constructed in male infants alone (N = 318), we identified 36 DEGs associated with sagittal craniosynostosis, 68 DEGs associated with metopic craniosynostosis, 6 DEGs associated with lambdoid craniosynostosis, and 9 DEGs associated with coronal craniosynostosis ( ). None of the DEGs were significant in all four phenotypes in males. We identified one significant pathway (FDR < 0.05)—“Growth hormone synthesis, secretion, and action”—with increased expression associated with metopic craniosynostosis in males ( ). There were no other pathways significantly increased or decreased for phenotypes in the male-stratified models, compared to controls. In our models constructed in female infants alone (N = 155), we identified fewer DEGs (FDR < 0.05) for each phenotype than our male-stratified analysis, including two DEGs associated with coronal ( NR3C2 , MMP11 ), one DEG associated with sagittal ( CD74 ), one DEG associated with metopic (NR3C2 ), and one DEG associated with lambdoid craniosynostosis ( RBM20 ) ( ). All DEGs showed decreased expression in the female craniosynostosis phenotypes, compared to female controls. One gene, NR3C2 , was significantly decreased in two phenotypes (coronal and metopic). There were no significant pathways identified in the female model. Across all phenotypes, more genes were significantly associated with craniosynostosis in the primary model (N = 473) than either the male (N = 318) or female (N = 155) sex-stratified models ( A). We did not identify any significant genes within a phenotype that were shared across our independent models for males, females, and the combined-sex analysis ( B–E). In all four craniosynostosis phenotypes, there were a subset of DEGs that were significant in both the male-stratified and primary models, while there was only overlap between the female-stratified and primary models for three phenotypes (coronal, lambdoid, metopic) ( B–E). Several homeobox genes were among the most differentially expressed genes (by log fold change) in the male and combined models, prompting a deeper characterization of homeobox genes. In total, there were sixteen DEGs associated with craniosynostosis phenotypes that were members of the homeobox gene family , including two DEGs associated with coronal craniosynostosis, eight DEGs associated with lambdoid craniosynostosis, five DEGs associated with metopic craniosynostosis, and nine DEGs associated with sagittal craniosynostosis ( ). All 16 homeobox genes showed directional concordance in log fold changes for each phenotype to control comparison. There were seven homeobox genes exhibiting decreased expression and nine homeobox genes exhibiting increased expression. In the primary model, the homeobox genes with the greatest overlap across phenotypes were DLX1 , DLX2 (decreased for sagittal and lambdoid), HOXB13 (increased for sagittal and lambdoid), and EN1 (increased for metopic, sagittal, and coronal). In the male-stratified model, DLX1 (decreased) was the only homeobox gene significantly associated with the three phenotypes (coronal, metopic, sagittal). There were a few genes that were only significantly associated with a single craniosynostosis phenotype, including HOXB4 (sagittal), IRX2 (sagittal and male sagittal), MESI1 , MEIS2 (metopic), IRX1 (male metopic), ALX1 (metopic and male metopic), HOXB2 (lambdoid), and HOXB3 (lambdoid and male lambdoid). 2.4. Transcription Factor Enrichment Analysis Three unique transcription factors emerged as potential regulators of DEGs, whose expression was associated with different craniosynostosis phenotypes in our primary and male-stratified models. These TFs were determined based on over-representation of TF binding sites in regulatory regions of craniosynostosis DEGs, measured through Enrichr ( A, FDR < 0.05). Two TFs, SUZ12 and EZH2, were significantly enriched as regulators of DEGs from more than one model or phenotype. These TFs are both members of the polycomb repressive complex (PRC) . SUZ12 was significantly enriched for DEGs associated with lambdoid, sagittal, metopic, and coronal craniosynostosis from the primary model, and sagittal, metopic, and lambdoid craniosynostosis in the male sex-stratified model. All downstream DEGs of SUZ12 were directionally concordant across models and phenotypes, with 37 DEGs exhibiting increased expression and 37 DEGs exhibiting decreased expression, in craniosynostosis phenotypes compared to controls ( C). EZH2 was significantly enriched for DEGs associated with lambdoid craniosynostosis in the primary model and male sex-stratified model. The downstream genes regulated by EZH2 were all upregulated ( B). Three of the DEGs downstream of EZH2 ( HOXB3 , HOXB13 , LMX1B ) are members of the HOX family.
RNA was sequenced from cultured primary calvarial cells derived from 85 controls and 388 craniosynostosis cases, that included 80 coronal, 196 sagittal, 92 metopic, and 20 lambdoid phenotypes ( ), as well as 85 control samples without craniosynostosis. As expected, the coronal phenotype was over-represented in females, and the metopic and sagittal phenotypes were over-represented in males (chi-squared test p -value < 0.05, ). There were no statistically significant differences across craniosynostosis phenotypes for proband age, at the time of sample collection. However, the age of control patients at time of collection was significantly different from all four craniosynostosis groups ( p < 0.001 for each phenotype vs. control comparisons, ANOVA with Tukey’s post hoc test) ( ). There was no statistically significant difference in cell culture time between controls and craniosynostosis phenotypes (ANOVA p > 0.05) ( ).
In our primary model, we identified 72 DEGs whose expression was associated with the coronal phenotype of craniosynostosis, 33 DEGs associated with lambdoid craniosynostosis, 103 DEGs associated with metopic craniosynostosis, and 90 DEGs associated with sagittal craniosynostosis, compared to controls without craniosynostosis. The full results are presented in . Generally, we observed an even distribution of DEGs that were increased and decreased in craniosynostosis phenotypes vs. controls ( ). The top three increased DEGs associated with each phenotype were, PAX3 , GATA3 , and OLR1 (coronal); PAX3 , TFAP2A , and ALX1 (metopic); SHOX2 , HOXB13 , and CPAMD8 (sagittal); and HOXB13 , HOXB3 , and PCDH17 (lambdoid). The top three decreased DEGs associated with each phenotype were, NRN1 , MCF2L , and STEAP4 (coronal); MCF2L , SMOC2 , and ST8SIA2 (metopic); CCKAR , NRN1 , and IRX2 (sagittal); and TRH , DLX1 , and XIRP1 (lambdoid). There was a single gene, EFHD1 , that was significantly decreased across all four phenotypes of craniosynostosis compared to controls ( F). The lambdoid and metopic phenotypes had the highest percentage of unique DEGs (lambdoid = 20/33 unique DEGs, metopic = 62/103 unique DEGs) that were not significantly associated with other craniosynostosis phenotypes. There were 51 genes (57% of sagittal DEGs) uniquely associated with sagittal craniosynostosis, and 33 genes (46% of coronal DEGs) uniquely associated with coronal craniosynostosis ( F). Using rotational gene set testing, we identified three significant pathways (FDR < 0.05) whose expression was associated with the coronal phenotypes ( ). All three pathways exhibited increased expression compared to controls, and are categorized as metabolism pathways by KEGG.
In our models constructed in male infants alone (N = 318), we identified 36 DEGs associated with sagittal craniosynostosis, 68 DEGs associated with metopic craniosynostosis, 6 DEGs associated with lambdoid craniosynostosis, and 9 DEGs associated with coronal craniosynostosis ( ). None of the DEGs were significant in all four phenotypes in males. We identified one significant pathway (FDR < 0.05)—“Growth hormone synthesis, secretion, and action”—with increased expression associated with metopic craniosynostosis in males ( ). There were no other pathways significantly increased or decreased for phenotypes in the male-stratified models, compared to controls. In our models constructed in female infants alone (N = 155), we identified fewer DEGs (FDR < 0.05) for each phenotype than our male-stratified analysis, including two DEGs associated with coronal ( NR3C2 , MMP11 ), one DEG associated with sagittal ( CD74 ), one DEG associated with metopic (NR3C2 ), and one DEG associated with lambdoid craniosynostosis ( RBM20 ) ( ). All DEGs showed decreased expression in the female craniosynostosis phenotypes, compared to female controls. One gene, NR3C2 , was significantly decreased in two phenotypes (coronal and metopic). There were no significant pathways identified in the female model. Across all phenotypes, more genes were significantly associated with craniosynostosis in the primary model (N = 473) than either the male (N = 318) or female (N = 155) sex-stratified models ( A). We did not identify any significant genes within a phenotype that were shared across our independent models for males, females, and the combined-sex analysis ( B–E). In all four craniosynostosis phenotypes, there were a subset of DEGs that were significant in both the male-stratified and primary models, while there was only overlap between the female-stratified and primary models for three phenotypes (coronal, lambdoid, metopic) ( B–E). Several homeobox genes were among the most differentially expressed genes (by log fold change) in the male and combined models, prompting a deeper characterization of homeobox genes. In total, there were sixteen DEGs associated with craniosynostosis phenotypes that were members of the homeobox gene family , including two DEGs associated with coronal craniosynostosis, eight DEGs associated with lambdoid craniosynostosis, five DEGs associated with metopic craniosynostosis, and nine DEGs associated with sagittal craniosynostosis ( ). All 16 homeobox genes showed directional concordance in log fold changes for each phenotype to control comparison. There were seven homeobox genes exhibiting decreased expression and nine homeobox genes exhibiting increased expression. In the primary model, the homeobox genes with the greatest overlap across phenotypes were DLX1 , DLX2 (decreased for sagittal and lambdoid), HOXB13 (increased for sagittal and lambdoid), and EN1 (increased for metopic, sagittal, and coronal). In the male-stratified model, DLX1 (decreased) was the only homeobox gene significantly associated with the three phenotypes (coronal, metopic, sagittal). There were a few genes that were only significantly associated with a single craniosynostosis phenotype, including HOXB4 (sagittal), IRX2 (sagittal and male sagittal), MESI1 , MEIS2 (metopic), IRX1 (male metopic), ALX1 (metopic and male metopic), HOXB2 (lambdoid), and HOXB3 (lambdoid and male lambdoid).
Three unique transcription factors emerged as potential regulators of DEGs, whose expression was associated with different craniosynostosis phenotypes in our primary and male-stratified models. These TFs were determined based on over-representation of TF binding sites in regulatory regions of craniosynostosis DEGs, measured through Enrichr ( A, FDR < 0.05). Two TFs, SUZ12 and EZH2, were significantly enriched as regulators of DEGs from more than one model or phenotype. These TFs are both members of the polycomb repressive complex (PRC) . SUZ12 was significantly enriched for DEGs associated with lambdoid, sagittal, metopic, and coronal craniosynostosis from the primary model, and sagittal, metopic, and lambdoid craniosynostosis in the male sex-stratified model. All downstream DEGs of SUZ12 were directionally concordant across models and phenotypes, with 37 DEGs exhibiting increased expression and 37 DEGs exhibiting decreased expression, in craniosynostosis phenotypes compared to controls ( C). EZH2 was significantly enriched for DEGs associated with lambdoid craniosynostosis in the primary model and male sex-stratified model. The downstream genes regulated by EZH2 were all upregulated ( B). Three of the DEGs downstream of EZH2 ( HOXB3 , HOXB13 , LMX1B ) are members of the HOX family.
This paper aimed to identify transcriptomic signatures of four single-suture phenotypes of craniosynostosis. This was accomplished through RNA sequencing of 473 patient- and control-derived primary calvarial cell lines, followed by differential gene expression and pathway analysis. The key findings of this study were that, (1) each craniosynostosis phenotype had at least 20 unique differentially expressed genes (DEGs), (2) more DEGs were identified in the male-stratified analysis compared to the female-stratified analysis, and (3) that homeobox genes were prevalent as DEGs across the primary and male-stratified models. Overall, these findings highlight the importance of assessing gene expression by phenotype and sex. Results also identified novel genes and pathways whose expression is altered in these single-suture craniosynostosis phenotypes. In the primary model, DEGs were identified for all four phenotypes with an approximately equal number of increased and decreased genes. Lambdoid craniosynostosis has the lowest number of DEGs (33 DEGs), while metopic craniosynostosis has the highest number of DEGs (103 DEGs). Several of the DEGs with the largest changes in expression have previously been associated with craniosynostosis or osteoblast differentiation. SMOC2 was among the most decreased genes in the metopic phenotype, compared to controls, and was also significantly decreased in the male-stratified metopic analysis. In canines with brachycephaly, SMOC2 is downregulated, due to alternative splicing caused by a LINE-1 insertion, which may indicate a role for SMOC2 in bicoronal (or brachycephalic) pediatric craniosynostosis . Conversely, TFAP2A was among the most increased DEGs in metopic craniosynostosis patients, compared to controls, and although it does not have a known association with craniosynostosis, its family of genes (TFAP) has been associated with other craniofacial deficiencies in tissues from a neural crest lineage , including branchio-oculo-facial syndrome , Char syndrome , and neural tube and skeletal defects . Additionally, TFAP2A expression in mesenchymal cells derived from the sagittal suture, has been measured as 1.97 times higher than in mesenchymal cells from the metopic suture . GATA3 was among the most increased DEGs associated with coronal craniosynostosis, compared to controls, and has previously been shown to have roles in osteoblast survival and the healing of bone fractures , which could explain its increased expression seen in the coronal craniosynostosis phenotype. There are known differences in incidence by patient sex across single-suture craniosynostosis phenotypes [ , , ]. We identified distinct differences in gene expression changes by patient sex, with male patients having vastly more DEGs than female patients for each craniosynostosis phenotype. Female models only identified 1–2 DEGs per phenotype, while male model DEGs ranged from 6 to 68. For every phenotype except lambdoid, there were also DEGs that were uniquely significant in the sex-stratified models compared to the primary model, for the same phenotype. The etiology of the male preponderance of craniosynostosis is not fully understood , but is an active area of research. A previous paper studying a subset of the same cohort of patients presented herein, evaluated differences in alkaline phosphatase (ALP) activity by phenotype and sex . The sagittal and metopic phenotypes in males, and the sagittal phenotype in females, was associated with increased Alp activity, and each of these phenotypes had unique gene sets, identified by microarray, that were correlated with Alp activity status . The ALPL gene, that encodes alkaline phosphatase, was significantly increased in metopic craniosynostosis in our primary and male-stratified models ( ). Multiple genetic variants of interest within ALPL have previously been identified, and are associated with occasional (<50%) occurrence of single-suture craniosynostosis . To our knowledge, the present study represents the first RNA sequencing evaluation of craniosynostosis phenotypes, in primary calvarial cell lines, to evaluate global gene differences by sex, so further research is needed to fully evaluate the differences in expression. Several of the genes identified through the sex-stratified analysis were also related to craniosynostosis or osteoblast differentiation. BMP6 was one of the most increased genes in the male sagittal phenotype and was also significant in the male metopic phenotype, and the coronal, sagittal, and metopic phenotypes in the primary model. BMP proteins comprise a gene family of over 22 members, that are components of the TGF-beta signaling pathway, where they are involved in the differentiation of mesenchymal stem cells to osteoblasts . In a case study of a prenatal diagnosis of de novo pure trisomy 6p (6p22.3 → p25.3) affected with craniosynostosis and microcephaly, the patient had a 20.88 Mb dosage increase in the genomic region containing the gene BMP6 , which led to overexpression of this gene . Recent studies have revealed that damaging de novo variants of genes within the BMP signaling pathway are associated with lambdoid craniosynostosis in human , and that augmented BMP signaling in mice neural crest cells is associated with premature fusion of intersphenoid synchondroses, resulting in craniofacial anomalies including craniosynostosis . CLDN11 was significantly increased, compared to controls, in coronal craniosynostosis in males, and was one of the top three increased genes in our male-stratified coronal craniosynostosis model. Although this gene has not previously been associated with craniosynostosis, CLDN11 is robustly expressed during osteoblast differentiation. Mouse osteoblasts that were treated with ascorbic acid for up to 24 days, to induce differentiation, exhibited a 60-fold increase in Cldn11 expression between days 0 and 8, which then decreased toward the later stages of differentiation . These results were specific to osteoblast cells, as osteoclast cells exhibited decreased CLDN11 expression, indicating that CLDN11 expression is dependent on cell type and stage of differentiation . Treatment with a recombinant CLDN11 protein has been demonstrated to prevent bone density impairment that is caused by LPS injection . MMP11 exhibited decreased expression in female coronal craniosynostosis compared to controls and was also decreased in the primary models for the coronal, metopic, and sagittal phenotypes, and male models for metopic and sagittal phenotypes. The MMP gene family degrades extracellular matrix proteins, allowing for cell migration, and also has roles in cell attachment, proliferation, differentiation, and apoptosis . MMP11 is part of the stromelysin sub-family of MMPs and has been identified, through histologic staining of the growth plate, in osteoblasts, osteocytes, and chrondrocytes in the proliferative and hypertrophic zones . Across both the primary and sex-stratified analyses, many of the DEGs for each phenotype were members of the homeobox gene family, which are a class of genes that contain a homeodomain DNA sequence . Many of the proteins encoded by these genes act as transcription factors that are important for embryonic development . While homeobox genes as a class have not been broadly associated with craniosynostosis, there is evidence for the involvement of a few homeobox genes in craniosynostosis etiology, including several identified in our transcriptome analysis. MSX2 and MSX1 are homeobox genes involved in craniofacial development , that have previously been associated with craniosynostosis. An MSX2 gain of function mutation is associated with Boston-type craniosynostosis, which most commonly affects lambdoid and coronal sutures . MSX2 overexpression in mice is associated with sagittal suture fusion . Although MSX2 was not identified as a DEG in this analysis, MSX1 , whose potential role in craniosynostosis is not well-defined, was increased in lambdoid and sagittal craniosynostosis in the primary model and sagittal craniosynostosis in the male-stratified model. MSX1+ skeletal stem cells are also able contribute to bone regeneration and ossification in response to calvarial defect injuries in rats, when treated with an optimized neurotrophic supplement . ALX4 is expressed in the mesenchyme of developing bone and has several connections to calvarial ossification and craniosynostosis. Genetic variants of ALX4 have been linked to craniosynostosis across multiple studies [ , , ]. Mutations in ALX4 have also been associated with other craniofacial defects including enlarged parietal foramina, which is an ossification defect within the parietal bone, highlighting the role of ALX4 in calvarial ossification during development . Though ALX4 was not differentially expressed between cases and controls in our study, ALX1 was significantly increased in metopic craniosynostosis in our primary and male-stratified models. ALX1 has not previously been associated with craniosynostosis, however, it is involved in osteogenesis and regulates a lncRNA transcript that is involved in bone marrow mesenchymal cell osteogenesis . Pathway analysis was performed, using FRY, on KEGG gene sets, to extend the biological significance of the findings to gene networks. The three pathways associated with coronal craniosynostosis in the primary model are all related to metabolic processes, but have not previously been associated with craniosynostosis, while the single pathway associated with metopic craniosynostosis in males was growth hormone, synthesis, secretion, and action. Although growth hormone is not directly implicated in craniosynostosis etiology, it is known to play a role in craniofacial development , and to coordinate with thyroid hormone and insulin-like growth factor (IGF) to affect bone mineral density . Transcription factor enrichment analysis was performed, to identify TFs that may regulate expression of the DEGs, based on an over-representation of genes that contain TF-specific binding motifs. Two of the significantly enriched TFs, SUZ12 and EZH2, are members of the polycomb repressive complex 2, which operates as a regulator of transcriptional silencing through deposition of H3K27me3 histone marks, with important roles in transcriptional regulation of differentiation processes during development . Of this pair, SUZ12 has not previously been associated with craniosynostosis, but there are several studies reporting associations between EZH2 and craniosynostosis. Deletion of EZH2 in undifferentiated mesenchymal cells in developing mice caused multiple structural defects related to bone and skeletal patterning including craniosynostosis, limb shortening, and clinodactyly . These defects were associated with changes to expression of HOX genes and may be attributable to early maturation of osteoblasts . In our study, EZH2 was enriched as a regulator of DEGs associated with the lambdoid phenotype in our primary and male-stratified models, while this previous study found that EZH2 deletion caused coronal and metopic craniosynostosis . EZH2 has also been implicated in premature suture closure in craniosynostosis cases that are accompanied by a mutation in TWIST1 , where EZH2 deposits fewer H3K27me3 silencing marks on osteogenic genes when TWIST1 is deleted . In our study, one of the significant downstream genes of EZH2 was HOXB13 , which interestingly also acts as a TF, where it regulates the expression of EZH2 itself, thus highlighting an interdependence of these two transcriptional regulators . More work is needed to understand the relationship between these TFs and craniosynostosis etiology. Androgen receptor (AR) was significantly enriched as a transcriptional regulator for genes that were differentially expressed in sagittal craniosynostosis. This is noteworthy due to the 3x higher incidence of this phenotype in male patients compared to female patients in our study ( ). Dysregulation of androgen hormones has previously been implicated in both syndromic and non-syndromic cases of craniosynostosis and the androgen receptor is abundantly expressed within the dura matter and calvarial bones of fetal mice . Increased androgen hormones can also affect expression of TGF-beta, leading to increased osteoblast differentiation and premature suture fusion . Multiple studies of suture closure in rabbits have indicated that treatment with the androgen-blocker flutamide can delay suture fusion and cause increased sutural separation . In vitro treatment of fetal murine dural and osteoblast cells with androgen hormone 5-alpha dihydrotestosterone, has also been demonstrated to increase differentiation and proliferation, supporting androgen involvement in suture fusion . This study should be interpreted with respect to its inherent limitations, including, that the sequencing was only performed on primary calvarial cells, consisting of mostly osteoblasts, whereas craniosynostosis likely involves gene expression changes to multiple cell types. Additionally, due to the rarity of craniosynostosis, some of the phenotypes, such as lambdoid, had limited sample size (N = 20). Another limitation is that the control group (N = 85) consisted of patients undergoing skull reconstruction surgeries for reasons unrelated to craniosynostosis, which could mean that there are baseline differences in calvaria bone gene expression between these controls and alternative non-pathological control sample populations, that did not have to undergo skull reconstruction surgery. Despite these limitations, this study had many strengths. One of the major strengths of this study is the sample size of 388 cases, spanning four phenotypes, which represents a large sample size of a relatively rare structural birth defect. This allowed us to individually analyze craniosynostosis phenotypes and perform a sex-stratified analysis for each phenotype. The use of RNA sequencing technology was also a strength, which aided in the identification of novel genes, pathways, and transcription factors related to craniosynostosis, that can be used to inform future studies.
4.1. Ethics Statement This study was HIPAA compliant and independently approved by the Institutional Review Board (IRB) associated with each of the four clinical sites, including Seattle Children’s Hospital, Northwestern University in Chicago, Children’s Health Care of Atlanta, and St. Louis Children’s Hospital. Informed consent was obtained from parents or legal guardians of all participants with craniosynostosis. Waivers of consent were approved by Seattle Children’s Hospital, for anonymous control samples. 4.2. Study Participants Three hundred and ninety-seven children, with computed tomography scans confirming diagnosis of single-suture craniosynostosis (SCC), were enrolled in the study at the time of treatment. Exclusion criteria included diagnosis of a major medical condition or presence of three or more minor extra-cranial malformations. Cases were also excluded based on the presence of causative mutations including FGFR1 , FGFR2 , FGFR3 , TWIST1 , EFNB1 , and MSX2. Calvaria and blood samples were obtained from cases during surgery. Eighty-seven children were recruited as controls, with tissue collected during craniotomy procedures for reasons other than craniosynostosis (e.g., brain tumor, isolated hydrocephalus, or at the time of autopsy). Calvaria bone fragments from cases and controls were collected from otherwise discarded tissue. All cases were screened for known pathogenic variants in FGFR1 , FGFR2 , FGFR3 , TWIST1 , EFNB1 , and MSX2 and excluded from the analysis if carrying one of these pathogenic variants or a chromosomal rearrangement. Calvaria fragments were used to establish primary calvarial cell lines as described below. We also excluded samples with failed cell growth in culture. Only cases and controls with full covariate data were included in this analysis, resulting in a total study population of 473 participants, with 85 controls and 388 cases. 4.3. Sample Processing After collection, calvaria fragments were transported in Waymouth media (WM) (Sigma, St. Louis, MO, USA), supplemented with 2× antibiotic/antimycotic solution (GE Hyclone, Logan, UT, USA) and 10% fetal bovine serum (FBS) (GE Hyclone, Logan, UT, USA). To prepare the tissue for culture, tissue was rinsed in WM and surrounding soft tissue was removed. Calvaria fragments were cut with a sterile scalpel blade into 1–2 mm pieces. These small calvaria fragments were then cultured with two pieces per well in 12-well plates, in WM at 37 °C, 5% CO 2 , and 99% humidity. After reaching confluence, approximately 3–6 weeks after plating, cells were washed with PBS, trypsinized with 0.05% Trypsin-EDTA (GE Hyclone, Logan, UT, USA), and passaged into T75 flasks. Cells were grown to confluence in the T75 flask then cryogenically stored in liquid nitrogen, using freezing media containing 90% FBS and 10% dimethyl sulfoxide . Once all samples were obtained and frozen, cells were thawed and cultured to sub-confluence in T25 flasks, where they were passaged at a density of 175,000 cells per 25 cm 2 . Once cells reached 75% confluence, they were trypsinized and washed in cold 1X PBS, followed by RNA isolation using the Roche High Pure miRNA Isolation Kit (Roche, Indianapolis, IN, USA), to isolate total RNA following the manufacturer’s protocol. The RNA Integrity Number (RIN) was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Samples with RIN > 8.6 were used for RNA Sequencing . 4.4. RNA Sequencing RNA was sequenced at Northwest Genomics Center, where next-generation sequencing libraries were prepared from 1.25 µg of total RNA in a high-throughput format, using the TruSeq Stranded mRNA kit (Illumina, San Diego, CA, USA). All the steps required for sequence library construction were automated and performed on a Sciclone NGSx Workstation (Perkin Elmer, Waltham, MA, USA). During library construction, rRNA was depleted by means of a poly-A enrichment, and first and second strand cDNA syntheses were performed. Each library was uniquely barcoded using Illumina adapters and amplified by PCR. After amplification and cleanup, library concentrations were quantified using the Quant-it dsDNA assay (Life Technologies, Carlsbad, CA, USA). Final libraries were normalized and pooled based on Agilent 2100 Bioanalyzer results (Agilent Technologies, Santa Clara, CA, USA) and size selected using a Pippin Prep (Sage Science, Beverly, MA, USA). Pooled libraries were diluted to a final concentration of 2–3 nM for sequencing on a HiSeq 4000, to a read depth of 30 million base pairs. Samples were multiplexed and sequenced on a HiSeq 4000. Lane-level sequencing reads were base quality checked using the FASTX-toolkit and FastQC, and aligned to hg19 with a reference transcriptome Ensembl v67, using TopHat2 suite followed by matefixing, as described previously . 4.5. Differential Gene Expression Analysis Genes were filtered, to remove non-protein coding genes and low expressing genes, using the EdgeR “filterbyExpr” function, with a cutoff of 20 cpm . Linear models were constructed to identify differentially expressed genes (DEGs) for each phenotype, compared to the controls, using the limma-voom pipeline . The primary model adjusted for patient sex, proband age, cell culture time, cohort, and sample city of origin. Surrogate variable analysis (SVA) was performed and included in the final model, to account for unknown confounding and cellular heterogeneity of samples . Models stratified by patient sex were constructed using the same variables as the primary model, as well as model-specific surrogate variables. Adjustment for multiple comparisons was performed using the Benjamini–Hochberg method, and genes were considered significant when the false discovery rate (FDR) was < 0.05. 4.6. Pathway Enrichment Analysis Pathway analysis was performed using the rotational self-contained gene set testing method, FRY, for each differential expression model described above . FRY accounts for inter-gene correlations, by evaluating if the T statistic for a given gene is different than expected under the null hypothesis, and determines if an overall pathway is increased or decreased . The Kyoto Encyclopedia of Genes and Genomes (KEGG) was used as the pathway database, and disease pathways were removed . Pathways with FDR < 0.05 were considered significant. 4.7. Transcription Factor Enrichment Analysis Transcription factor enrichment analysis was performed on the lists of DEGs from the primary and sex-stratified phenotype models using Enrichr, with the ENCODE/ChEA consensus TF library [ , , ]. TFs were considered significantly enriched when the FDR was < 0.05.
This study was HIPAA compliant and independently approved by the Institutional Review Board (IRB) associated with each of the four clinical sites, including Seattle Children’s Hospital, Northwestern University in Chicago, Children’s Health Care of Atlanta, and St. Louis Children’s Hospital. Informed consent was obtained from parents or legal guardians of all participants with craniosynostosis. Waivers of consent were approved by Seattle Children’s Hospital, for anonymous control samples.
Three hundred and ninety-seven children, with computed tomography scans confirming diagnosis of single-suture craniosynostosis (SCC), were enrolled in the study at the time of treatment. Exclusion criteria included diagnosis of a major medical condition or presence of three or more minor extra-cranial malformations. Cases were also excluded based on the presence of causative mutations including FGFR1 , FGFR2 , FGFR3 , TWIST1 , EFNB1 , and MSX2. Calvaria and blood samples were obtained from cases during surgery. Eighty-seven children were recruited as controls, with tissue collected during craniotomy procedures for reasons other than craniosynostosis (e.g., brain tumor, isolated hydrocephalus, or at the time of autopsy). Calvaria bone fragments from cases and controls were collected from otherwise discarded tissue. All cases were screened for known pathogenic variants in FGFR1 , FGFR2 , FGFR3 , TWIST1 , EFNB1 , and MSX2 and excluded from the analysis if carrying one of these pathogenic variants or a chromosomal rearrangement. Calvaria fragments were used to establish primary calvarial cell lines as described below. We also excluded samples with failed cell growth in culture. Only cases and controls with full covariate data were included in this analysis, resulting in a total study population of 473 participants, with 85 controls and 388 cases.
After collection, calvaria fragments were transported in Waymouth media (WM) (Sigma, St. Louis, MO, USA), supplemented with 2× antibiotic/antimycotic solution (GE Hyclone, Logan, UT, USA) and 10% fetal bovine serum (FBS) (GE Hyclone, Logan, UT, USA). To prepare the tissue for culture, tissue was rinsed in WM and surrounding soft tissue was removed. Calvaria fragments were cut with a sterile scalpel blade into 1–2 mm pieces. These small calvaria fragments were then cultured with two pieces per well in 12-well plates, in WM at 37 °C, 5% CO 2 , and 99% humidity. After reaching confluence, approximately 3–6 weeks after plating, cells were washed with PBS, trypsinized with 0.05% Trypsin-EDTA (GE Hyclone, Logan, UT, USA), and passaged into T75 flasks. Cells were grown to confluence in the T75 flask then cryogenically stored in liquid nitrogen, using freezing media containing 90% FBS and 10% dimethyl sulfoxide . Once all samples were obtained and frozen, cells were thawed and cultured to sub-confluence in T25 flasks, where they were passaged at a density of 175,000 cells per 25 cm 2 . Once cells reached 75% confluence, they were trypsinized and washed in cold 1X PBS, followed by RNA isolation using the Roche High Pure miRNA Isolation Kit (Roche, Indianapolis, IN, USA), to isolate total RNA following the manufacturer’s protocol. The RNA Integrity Number (RIN) was assessed using the Agilent 2100 Bioanalyzer (Agilent Technologies, Santa Clara, CA, USA). Samples with RIN > 8.6 were used for RNA Sequencing .
RNA was sequenced at Northwest Genomics Center, where next-generation sequencing libraries were prepared from 1.25 µg of total RNA in a high-throughput format, using the TruSeq Stranded mRNA kit (Illumina, San Diego, CA, USA). All the steps required for sequence library construction were automated and performed on a Sciclone NGSx Workstation (Perkin Elmer, Waltham, MA, USA). During library construction, rRNA was depleted by means of a poly-A enrichment, and first and second strand cDNA syntheses were performed. Each library was uniquely barcoded using Illumina adapters and amplified by PCR. After amplification and cleanup, library concentrations were quantified using the Quant-it dsDNA assay (Life Technologies, Carlsbad, CA, USA). Final libraries were normalized and pooled based on Agilent 2100 Bioanalyzer results (Agilent Technologies, Santa Clara, CA, USA) and size selected using a Pippin Prep (Sage Science, Beverly, MA, USA). Pooled libraries were diluted to a final concentration of 2–3 nM for sequencing on a HiSeq 4000, to a read depth of 30 million base pairs. Samples were multiplexed and sequenced on a HiSeq 4000. Lane-level sequencing reads were base quality checked using the FASTX-toolkit and FastQC, and aligned to hg19 with a reference transcriptome Ensembl v67, using TopHat2 suite followed by matefixing, as described previously .
Genes were filtered, to remove non-protein coding genes and low expressing genes, using the EdgeR “filterbyExpr” function, with a cutoff of 20 cpm . Linear models were constructed to identify differentially expressed genes (DEGs) for each phenotype, compared to the controls, using the limma-voom pipeline . The primary model adjusted for patient sex, proband age, cell culture time, cohort, and sample city of origin. Surrogate variable analysis (SVA) was performed and included in the final model, to account for unknown confounding and cellular heterogeneity of samples . Models stratified by patient sex were constructed using the same variables as the primary model, as well as model-specific surrogate variables. Adjustment for multiple comparisons was performed using the Benjamini–Hochberg method, and genes were considered significant when the false discovery rate (FDR) was < 0.05.
Pathway analysis was performed using the rotational self-contained gene set testing method, FRY, for each differential expression model described above . FRY accounts for inter-gene correlations, by evaluating if the T statistic for a given gene is different than expected under the null hypothesis, and determines if an overall pathway is increased or decreased . The Kyoto Encyclopedia of Genes and Genomes (KEGG) was used as the pathway database, and disease pathways were removed . Pathways with FDR < 0.05 were considered significant.
Transcription factor enrichment analysis was performed on the lists of DEGs from the primary and sex-stratified phenotype models using Enrichr, with the ENCODE/ChEA consensus TF library [ , , ]. TFs were considered significantly enriched when the FDR was < 0.05.
Overall, this study identified unique transcriptomic changes in primary cells derived from the calvaria of patients affected by four phenotypes of craniosynostosis. We also found sex-specific gene expression differences within the four individual phenotypes. This is particularly important, as metopic and sagittal phenotypes show higher occurrence in males, with coronal craniosynostosis occurring more frequently in females. Sixteen homeobox gene family members were associated with craniosynostosis phenotypes in our primary or male models. Differentially expressed genes for each phenotype were enriched for binding sites of TFs including AR, SUZ12, or EZH2 in primary and male models, highlighting a potential role for transcriptional regulation in eliciting altered gene expression. While many of the genes described in this paper have previous associations with craniosynostosis or osteoblast differentiation, many of these findings are novel, and will require additional experimental work to understand their role in craniosynostosis etiology. As the largest transcriptome-wide analysis of craniosynostosis to date, we anticipate that the results of this analysis will be utilized to generate new hypotheses on the molecular mechanisms underpinning craniosynostosis diagnoses.
|
Practices and Perceptions of Family-Centered Care: A Cross-Sectional Survey of Secondary School Athletic Trainers | 1bcca3d0-5c66-4858-b8d8-7bef93513f10 | 10049324 | Patient-Centered Care[mh] | Patient-centered care, which includes the healthcare provider being responsive to and respectful of the patient’s values and needs in their clinical care, has been widely studied across healthcare professions, specifically in athletic training . Athletic training is a healthcare profession that provides medical services to individuals involved in sports, physical activity, and work environments. The practice of athletic training involves five domains, including (1) risk reduction, wellness, and health literacy; (2) assessment, evaluation, and diagnosis; (3) critical incident management; (4) therapeutic intervention; and (5) healthcare administration and professional responsibility . Athletic trainers are educated at the postbaccalaureate level and must pass a board exam for national certification to practice clinically. The role of the athletic trainer is to work collaboratively to ensure patient-centered medical services . In a patient-centered environment, an individual should expect their provider to invite them to ask questions and make sure they are in full collaboration with their treatment plan decisions . Previous research specific to athletic training identified collegiate student-athletes who believed empathy was important and that with patient-centered communication, there would be a solid relationship and result in better outcomes . Moreover, most collegiate student-athletes viewed their athletic trainers to be patient-centered in terms of individualized care, the patient being a priority to them, and they felt they received the best options possible . However, goal setting continues to emerge as a limitation of communication strategies deployed by the athletic trainer . While most encounters involve patient-centered care through the lens of working with adults, the situation faced by athletic trainers in secondary schools is unique. In most patient encounters, secondary school athletic trainers typically care for patients under the age of 18 in a school-based health system and are tasked with collaborating with other healthcare professionals, but most importantly, are responsible for maintaining constant communication with the parents, guardians, or caregivers of minors . Due to the age group of these patients, there must be strong communication with the family on the care and treatment these student-athletes are receiving. It is the responsibility of the athletic trainer in secondary schools caring for adolescent athletes to be familiar with and comply with state laws and regulations specific to the consent authority. While it is required by law to communicate with a parent/guardian when treating minors, parent engagement with the healthcare provider may differ. The literature supports that parents value patient-centered care, also referred to as child-centered care in this case, and the ability to provide input when dealing with their children’s health . The inclusion of strong communication and connection to the families when talking about adolescent healthcare is known as family-centered care (FCC) . When talking about this form of care, it can be described as including “information sharing, partnering, respect, and negotiation” that prioritizes the engagement of the patient’s family in the shared decision-making process . Families are typically important support systems for children. The family wants to be told information on their child’s injury or illness, the course of action and treatment, and the prognosis. Without communication on these things, families could be left out of key decisions and possibly lose trust in the healthcare provider. The family not only serves as a patient advocate in the process but also provides financial assistance and emotional support outside of the healthcare facility . Previous research identified parents were aware of the training and knowledge of athletic trainers and felt having a provider on site was a key to safety . Engaging with parents, siblings, and other family members in the healthcare process for an adolescent student-athlete allows for the athletic trainer to have insight into their patient’s specific social determinants of health, preferences, and cultural considerations necessary to individualize the care process . However, the concept of FCC, whereby the family takes an active role in the planning, implementation, and evaluation of the care and has a say in the care of the child as much as a health professional, has not been examined in athletic trainers. The aim of FCC is to maintain the connections between the underage patient and their family, to prioritize the effectiveness in the care of the child, and to prevent or minimize the negative effects of illness, injury, or hospitalization. Research indicates the universal components of FCC include collaboration between family members and healthcare providers, the creation of flexible policies and procedures, and the need for patient and family education . The skills and behaviors align with the expected contributions of an athletic trainer providing care in middle and high school settings. Therefore, the purpose of this study was to describe the perceptions and practices of FCC care across secondary school athletic trainers.
We used a cross-sectional survey design to explore secondary school athletic trainers’ practices and perceptions of FCC. We used the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) Statement checklist for cross-sectional studies to guide the presentation of this manuscript. This study was conducted according to the guidelines of the Declaration of Helsinki and deemed exempt from approval by the Institutional Review Board of the University of South Carolina (Submission: Pro00121032; Approval Date: 23 May 2022). 2.1. Instrument To explore FCC, the research team used the content-validated and reliable Family-Centered Care Questionnaire-Revised (FCCQ-R) . In addition, the survey contained five demographic questions about the participant’s age, gender, years of credentialed experience as an athletic trainer, the highest level of education, and the type of secondary school they work in (e.g., private school, public school, or other). The FCCQ-R contains 45 items surrounding one’s beliefs about what FCC is or is not. Each item was assessed by the participant on the two FCCQ-R scales, which include the current practiced (CP) scale and the perceived necessary (PN) scale. The two FCCQ-R scales are further comprised of nine subscales, including (1) family is the constant, (2) parent/professional collaboration, (3) recognizing family individuality, (4) sharing information with parents, (5) developmental needs, (6) parent-to-parent support, (7) emotional and financial support, (8) design of healthcare system, and (9) emotional support for staff. Each item was assessed on both scales using a 5-point Likert scale from 1 = strongly disagree to 5 = strongly agree, with a higher number indicating a higher perception or practice of the item specific to FCC. In addition, the survey included an open-ended question where participants were able to write additional comments or suggestions of what they believe is needed to enhance family-centered care. Due to the nature of the FCCQ-R items being written about hospital care with words such as admission and facilities, the research team for this study engaged in a minor revision process before data collection. In total, 13 of the 45 items in the tool required grammatical changes. A panel of 3 athletic trainers with experience working in the secondary school setting, as well as doctoral-level training in research, was convened to content validate the changes. The primary investigator made the initial edits to the FCCQ-R items and sent the revised tool via e-mail to the expert panel for feedback. After feedback was collected, edits were made to the wording only on the changes (for example, hospital to athletic training facility). Following the edits, a content validation index (CVI) was performed for the clarity and relevance of the proposed changes . This resulted in a CVI relevancy average score of 0.961/1. In addition, a clarity average score of 0.955/1 was also achieved. With both CVI scores greater than or equal to 0.9, the tool modifications were considered to be content validated . We identified an overall total Cronbach’s alpha for the CP: α = 0.945 and PN: α = 0.946 scales demonstrating high reliability. provides the full-scale and subscale reliability assessment. 2.2. Participants This study aimed to explore FCC with athletic trainers working in the secondary school setting across the United States. We used G*Power software (version 3.1.9.4; Franz Faul, Universität Kiel, Kiel, Germany) to explore an appropriate sample size for the study. Using an α = 0.05 and a moderate effect size (0.3), the power calculation indicated a sample size of 132 participants was needed for an estimated power of 0.95. To do so, we recruited the entire sample (n = 7695) of eligible individuals from this job setting that were also members of the National Athletic Trainers’ Association (NATA) and opted into the research participation database. A total of 504 athletic trainers accessed the survey (response rate = 6.5%), and 305 individuals started the survey. Overall, 205 athletic trainers in the secondary school setting (men: n = 91, women: n = 113, preferred not to say: n = 1; age range: 18 to over 65 years) completed the survey yielding a 67.2% (n = 205/305) completion rate. The data from complete responses were used for the analysis. Full demographic information for the participants is presented in . 2.3. Procedures The recruitment e-mail was sent by the NATA research database describing the study and providing directions for completing the survey, as well as a direct link to the survey via a secure, web-based system (Qualtrics, Provo, UT, USA). Data collection began in May 2022 with reminder e-mails sent every week for four weeks (June 2022) to unfinished respondents. Due to the timing of the study (May–June), which aligned with the end of the traditional secondary school academic calendar, we opted to complete an additional 4 weeks of data collection using the same sample of unfinished respondents with reminder e-mails beginning again in September 2022 for four weeks ending in October 2022. After opening the survey link, participants were provided with an invitation to participate and the ability to indicate their willingness to engage in the study. Individuals that did not consent to participate (n = 20) and/or indicated they were not currently employed as an athletic trainer in the secondary school setting (n = 18) were excluded from the study. For those that met inclusion criteria and after consenting to participate, the participants were directed to the survey items, answering only the questions they wanted to, and could close the browser at any time. Participants completed the 45-item FCCQ-R instrument, one open-ended item about enhancing FCC, and five demographic questions. Specific directions and definitions were provided to the participant on the FCCQ-R. Participants were instructed to consider the extent they include this aspect of care in their everyday work at their secondary school (CP scale) and whether they believe this aspect of care is necessary for FCC to be provided in athletic training (PN scale). The questions on the tool also identified staff and family, which was each operationally defined in the survey with staff meaning “any health professional who provides direct care for children and their families, for example: school nurse, athletic trainer, social worker, clergy, psychologist, physician, etc.” and “family refers to parents, guardians, grandparents, and siblings, and any family friend who is significantly involved with the child’s everyday life or care” . 2.4. Data Analysis The data were collected and stored on the web-based platform. The completed survey responses were downloaded to SPSS (IBM Corp., IBM SPSS Statistics for Windows Version 28, Armonk, NY, USA) for statistical analyses. The data were first scored by adding scores for each item on the scales for each of the practice and perception to calculate a mean score. A total CP and PN scale score was also calculated by summing the mean for each of the 9 subscales, which were used to compare perception and practice. Scales were included if all or all but 1 item was answered for the questions within that scale. Means were calculated for the 2 scales and 9 subscales. The data were analyzed using paired samples t -tests to explore differences between the CP and PN scales. Finally, of the 205 participants, 44 secondary school athletic trainers responded to the one open-ended question in the survey in response to “please write any additional comments or suggestions of what is needed to enhance family-centered care.” Qualitative data were analyzed using an inductive process to examine trends in participant responses. Trustworthiness was achieved through researcher triangulation, where two members (Z.K.W., J.M.M.) of the research team coded the data and identified main themes separately before comparing findings and coming to a consensus. A strength of this design is that qualitative data can be used to provide richness and complement the quantitative findings. The data are presented as extracted quotes that best characterize each theme.
To explore FCC, the research team used the content-validated and reliable Family-Centered Care Questionnaire-Revised (FCCQ-R) . In addition, the survey contained five demographic questions about the participant’s age, gender, years of credentialed experience as an athletic trainer, the highest level of education, and the type of secondary school they work in (e.g., private school, public school, or other). The FCCQ-R contains 45 items surrounding one’s beliefs about what FCC is or is not. Each item was assessed by the participant on the two FCCQ-R scales, which include the current practiced (CP) scale and the perceived necessary (PN) scale. The two FCCQ-R scales are further comprised of nine subscales, including (1) family is the constant, (2) parent/professional collaboration, (3) recognizing family individuality, (4) sharing information with parents, (5) developmental needs, (6) parent-to-parent support, (7) emotional and financial support, (8) design of healthcare system, and (9) emotional support for staff. Each item was assessed on both scales using a 5-point Likert scale from 1 = strongly disagree to 5 = strongly agree, with a higher number indicating a higher perception or practice of the item specific to FCC. In addition, the survey included an open-ended question where participants were able to write additional comments or suggestions of what they believe is needed to enhance family-centered care. Due to the nature of the FCCQ-R items being written about hospital care with words such as admission and facilities, the research team for this study engaged in a minor revision process before data collection. In total, 13 of the 45 items in the tool required grammatical changes. A panel of 3 athletic trainers with experience working in the secondary school setting, as well as doctoral-level training in research, was convened to content validate the changes. The primary investigator made the initial edits to the FCCQ-R items and sent the revised tool via e-mail to the expert panel for feedback. After feedback was collected, edits were made to the wording only on the changes (for example, hospital to athletic training facility). Following the edits, a content validation index (CVI) was performed for the clarity and relevance of the proposed changes . This resulted in a CVI relevancy average score of 0.961/1. In addition, a clarity average score of 0.955/1 was also achieved. With both CVI scores greater than or equal to 0.9, the tool modifications were considered to be content validated . We identified an overall total Cronbach’s alpha for the CP: α = 0.945 and PN: α = 0.946 scales demonstrating high reliability. provides the full-scale and subscale reliability assessment.
This study aimed to explore FCC with athletic trainers working in the secondary school setting across the United States. We used G*Power software (version 3.1.9.4; Franz Faul, Universität Kiel, Kiel, Germany) to explore an appropriate sample size for the study. Using an α = 0.05 and a moderate effect size (0.3), the power calculation indicated a sample size of 132 participants was needed for an estimated power of 0.95. To do so, we recruited the entire sample (n = 7695) of eligible individuals from this job setting that were also members of the National Athletic Trainers’ Association (NATA) and opted into the research participation database. A total of 504 athletic trainers accessed the survey (response rate = 6.5%), and 305 individuals started the survey. Overall, 205 athletic trainers in the secondary school setting (men: n = 91, women: n = 113, preferred not to say: n = 1; age range: 18 to over 65 years) completed the survey yielding a 67.2% (n = 205/305) completion rate. The data from complete responses were used for the analysis. Full demographic information for the participants is presented in .
The recruitment e-mail was sent by the NATA research database describing the study and providing directions for completing the survey, as well as a direct link to the survey via a secure, web-based system (Qualtrics, Provo, UT, USA). Data collection began in May 2022 with reminder e-mails sent every week for four weeks (June 2022) to unfinished respondents. Due to the timing of the study (May–June), which aligned with the end of the traditional secondary school academic calendar, we opted to complete an additional 4 weeks of data collection using the same sample of unfinished respondents with reminder e-mails beginning again in September 2022 for four weeks ending in October 2022. After opening the survey link, participants were provided with an invitation to participate and the ability to indicate their willingness to engage in the study. Individuals that did not consent to participate (n = 20) and/or indicated they were not currently employed as an athletic trainer in the secondary school setting (n = 18) were excluded from the study. For those that met inclusion criteria and after consenting to participate, the participants were directed to the survey items, answering only the questions they wanted to, and could close the browser at any time. Participants completed the 45-item FCCQ-R instrument, one open-ended item about enhancing FCC, and five demographic questions. Specific directions and definitions were provided to the participant on the FCCQ-R. Participants were instructed to consider the extent they include this aspect of care in their everyday work at their secondary school (CP scale) and whether they believe this aspect of care is necessary for FCC to be provided in athletic training (PN scale). The questions on the tool also identified staff and family, which was each operationally defined in the survey with staff meaning “any health professional who provides direct care for children and their families, for example: school nurse, athletic trainer, social worker, clergy, psychologist, physician, etc.” and “family refers to parents, guardians, grandparents, and siblings, and any family friend who is significantly involved with the child’s everyday life or care” .
The data were collected and stored on the web-based platform. The completed survey responses were downloaded to SPSS (IBM Corp., IBM SPSS Statistics for Windows Version 28, Armonk, NY, USA) for statistical analyses. The data were first scored by adding scores for each item on the scales for each of the practice and perception to calculate a mean score. A total CP and PN scale score was also calculated by summing the mean for each of the 9 subscales, which were used to compare perception and practice. Scales were included if all or all but 1 item was answered for the questions within that scale. Means were calculated for the 2 scales and 9 subscales. The data were analyzed using paired samples t -tests to explore differences between the CP and PN scales. Finally, of the 205 participants, 44 secondary school athletic trainers responded to the one open-ended question in the survey in response to “please write any additional comments or suggestions of what is needed to enhance family-centered care.” Qualitative data were analyzed using an inductive process to examine trends in participant responses. Trustworthiness was achieved through researcher triangulation, where two members (Z.K.W., J.M.M.) of the research team coded the data and identified main themes separately before comparing findings and coming to a consensus. A strength of this design is that qualitative data can be used to provide richness and complement the quantitative findings. The data are presented as extracted quotes that best characterize each theme.
3.1. Practices and Perceptions Mean scores and paired samples t -tests on the practice and perception scales of FCC are presented in . The total mean score for the CP scale (mean = 26.83 ± 4.36) was significantly lower ( p ≤ 0.01) than the PN scale (mean = 35.33 ± 4.17). The parent-to-parent support collaboration subscale had both the most practiced (mean = 3.93 ± 0.85) and highest perceived need (mean = 4.61 ± 0.73). In addition, the developmental needs subscale had both the lowest score for CP (mean = 2.62 ± 0.79) and was considered by secondary school athletic trainers as the least necessary element of FCC (mean = 3.23 ± 0.76). The smallest subscale discrepancy between CP and PN was family is the constant meaning their perceptions and practice closely align. All FCC subscales compared between CP and PN were significantly different ( p ≤ 0.01), with each being of higher importance than CP in athletic training. 3.2. Open-Ended Responses Data analysis revealed four themes related to enhancing FCC in secondary schools: limited education and resources, staffing and space concerns, non-technical skills, and social determinants of health. Athletic trainers consistently described not having the appropriate education or resources to integrate FCC appropriately into their setting. They also highlighted staffing and space concerns that would not allow them to effectively communicate and connect to families when they have so many athletes and so few staff members. The third theme suggested that to adopt FCC correctly, athletic trainers need to have or acquire non-technical skills such as communication, establishing trust, and developing relationships. The final theme from the data indicated social determinants of health play an important role and must be considered when integrating FCC into secondary schools. provides extracted quotes from the data that best describe the associated theme.
Mean scores and paired samples t -tests on the practice and perception scales of FCC are presented in . The total mean score for the CP scale (mean = 26.83 ± 4.36) was significantly lower ( p ≤ 0.01) than the PN scale (mean = 35.33 ± 4.17). The parent-to-parent support collaboration subscale had both the most practiced (mean = 3.93 ± 0.85) and highest perceived need (mean = 4.61 ± 0.73). In addition, the developmental needs subscale had both the lowest score for CP (mean = 2.62 ± 0.79) and was considered by secondary school athletic trainers as the least necessary element of FCC (mean = 3.23 ± 0.76). The smallest subscale discrepancy between CP and PN was family is the constant meaning their perceptions and practice closely align. All FCC subscales compared between CP and PN were significantly different ( p ≤ 0.01), with each being of higher importance than CP in athletic training.
Data analysis revealed four themes related to enhancing FCC in secondary schools: limited education and resources, staffing and space concerns, non-technical skills, and social determinants of health. Athletic trainers consistently described not having the appropriate education or resources to integrate FCC appropriately into their setting. They also highlighted staffing and space concerns that would not allow them to effectively communicate and connect to families when they have so many athletes and so few staff members. The third theme suggested that to adopt FCC correctly, athletic trainers need to have or acquire non-technical skills such as communication, establishing trust, and developing relationships. The final theme from the data indicated social determinants of health play an important role and must be considered when integrating FCC into secondary schools. provides extracted quotes from the data that best describe the associated theme.
The present study aimed to describe the perceptions and practices of FCC across secondary athletic trainers, making this the first study to explore this concept in this profession. In athletic training, the Board of Certification Standards of Professional Practice states athletic trainers must inform “the parent/guardian of a minor patient of any risks involved in the treatment plan” and “develops and maintains a relationship of trust and confidence with the parent/guardian of a minor patient” . The data from our study suggest athletic trainers feel they are practicing FCC, with the exception of “developmental needs.” However, there were significant differences between their practices and perceptions of FCC. These findings are consistent with previous studies conducted across other healthcare settings, such as nursing and neonatal intensive care units [ , , ]. In addition, research has assessed the attitudes toward FCC among physical therapists, professionals with similar skill sets as athletic trainers, in the United States . The highest scores observed in that study included the respectful and supportive care subscale, which can be described as making parents feel respected as individuals, equals, and experts . Overall, there are several well-documented benefits of FCC in other peer healthcare professions, and we believe this study could help to bridge the gap between the FCC principles in sports medicine. 4.1. Parent-to-Parent Support At the time of injury, secondary school athletes reflected that they worried, as well as their family and friends, about the severity of their injury . In addition, previous research has identified families are at risk of developing mental health challenges, such as post-traumatic stress disorder, after their child sustains an injury . This sense of worry can be reduced or eliminated through parent-to-parent support groups. As originally proposed by the Association for the Care of Children’s Health , parent-to-parent support means assistance provided by a parent who may be facing similar challenges and who can understand what one is going through. Parents have expressed this as a key element of FCC . Parents feel the power of shared experiences allows them to empathize and be validated when facing difficult challenges with the health of their children. Furthermore, dating back to 1986, Minna Nathanson outlined three important functions of parent-to-parent support, which include (1) mutual support and friendship, (2) information gathering and sharing, and (3) improving the system . Overall, parents want the opportunity to help each other identify information to navigate the hardships associated with caring for their children and develop support systems along the way. Athletic trainers reported parent-to-parent support as the highest perceived and most frequently practiced element of FCC. Interestingly, this element is not ranked highly across other healthcare professions for both perception and practice [ , , , , , , , ]. Athletic trainers reported practicing parent-to-parent support (mean = 3.93 ± 0.85) more frequently as compared to other professions, with mean scores ranging from 2.38 to 3.55 [ , , , , , ]. Similarly, athletic trainers perceived this as a necessary element more than other healthcare professionals. To our knowledge, the studies conducted by Franck et al. and Petersen et al. are the only studies conducted in the U.S. to date that utilized the FCCQ-R . Participants in these studies reported similar perceptions as athletic trainers on the parent-to-parent scale, with scores of 4.23 and 4.04. It can be possible that healthcare professionals in the U.S. understand and value the shared experiences that parents have with regard to the care of their children and may promote it more than in other countries. It is of future interest to explore how athletic trainers coordinate and provide parent-to-parent support in the secondary school setting. 4.2. Developmental Needs Across athletic trainers, the least practiced and lowest perceived necessary element of FCC was developmental needs. In its original form, the incorporation of the developmental needs of children, adolescents, and their families into the healthcare systems was primarily intended to serve individuals with disabilities . The literature exploring FCC has predominately been conducted across nurses in pediatric or neonatal hospital units, where patients may face significant challenges with regard to growing up with disabilities and how to educate patients to be self-sufficient and independent [ , , ]. Findings indicate that most healthcare professionals practice this element of FCC and perceive it as necessary, with mean scores as high as 4.44 . In comparison, athletic trainers’ perception of developmental needs was 3.23. Secondary school athletic trainers do not typically face these developmental challenges, as they work with a population who is highly active and engaged in intense physical activity, and a possible explanation to why they did not perceive developmental needs as highly as other elements of FCC. Following an injury or illness, a patient may experience emotional and social developmental challenges that affect their interest and motivation. While the patients’ athletic trainers typically care for primarily able-bodied children, a focus needs to be placed on the emotional and social components when the child experiences an injury. Practicing and incorporating this element of FCC may be beneficial when caring for patients instead of a compartmentalized approach that focuses on the injury separate from the emotional and social aspects. To do so, we recommend athletic trainers approach patient care using the World Health Organization International Classification of Functioning, Disability, and Health (ICF) disablement model framework . The ICF model allows the clinician to explore other areas outside of the health condition, which could highlight developmental needs such as body functions, activity limitations, participation restrictions, environmental factors, and personal factors . 4.3. Family Is the Constant Athletic trainers provide services in secondary school after obtaining parental consent at the onset of the school year rather than at each patient encounter . This could be detrimental to the relationship and promoting FCC as it limits the connectivity between the parties, which reinforces the lack of knowledge and limited experiences for parents to meet and understand the role of the athletic trainer . Athletic trainers recognize that a relationship with parents is important and could serve as a barrier to providing FCC if not present. To implement FCC, it is imperative athletic trainers acknowledge that family is the constant construct. This means that as the adolescent discontinues sport participation, recovers from their injury or illness, and/or graduates from secondary school, their support system (i.e., parents, guardians, siblings) will continue while the athletic trainer’s role is temporary. While not the highest practiced element of FCC, the construct that the family is the constant had the lowest discrepancy between the CP and PN scales. Across the literature, this theme does not seem to be uncommon, with other health professions stating the same sentiment. In fact, the family is the constant scale is commonly the highest practiced FCC [ , , , , ]. Similarly, the implementation gap between perceptions and practice of this element of FCC has had the smallest differences across other studies [ , , , ]. Altogether, the “family as a constant” is a theme that health professionals, including athletic trainers, perceive as important when practicing FCC. Furthermore, it can be assumed that athletic trainers, in general, believe proper communication with the family and a good trusting relationship is vital for ensuring quality FCC. Unfortunately, athletic trainers in the secondary school setting cited parents as both the most common job-related stressor and source of conflict . There is a disconnect between the clinician-centered care approach from the athletic trainer resulting in an uninformed parent, a confused child, and a result-oriented athletic trainer. We recommend athletic trainers change their perspective on the role of the parent/family. The theory of social capital suggests leveraging the family as a resource to build relationships focused on trust, safety, and social cohesion to influence the sharing of information, increase health outcomes, and accomplish shared goals . In the case of a patient–provider relationship, the athletic trainer can build trusting relationships with the families to improve the minor’s health. A key strategy to incorporate is creating multiple communication options that meet the needs of the parents in terms of time and accessibility and avoiding medical jargon. The integration of FCC into clinical practice may increase the short-term burden of stress and conflict but ultimately should promote a common goal between the family and the athletic trainer. However, the “best interest” of the child held by the family and the athletic trainer may differ, resulting in conflict, power imbalances, and expert tension . A natural first step in changing the culture of healthcare for athletic trainers in the secondary school setting would be to explore the perception and reputation held by the student-athletes, family members, and personnel. The data should explore the current perceived training, skills, opportunities, and areas for necessary improvement. The commitment to engaging and embracing the family throughout the process could have positive, long-term benefits for addressing the healthcare of the adolescent. In addition, previous research has identified parents need greater education on issues that they can deem important for their children’s health . It is critical athletic trainers prioritize family education, as well as patient education, to encourage collaboration throughout the care plan. In doing so, the athletic trainer should explore the parent’s health literacy in the care decision-making process and provide education and options individualized to the family . Specifically, there is a need to encourage home care management facilitated by the parents or guardians. The active engagement of the family in the therapeutic intervention process could improve buy-in from both the patient and family and alleviate the perceived stressors and conflicts. 4.4. Exploring the Qualitative Findings We identified four themes from the open-ended responses related to an athletic trainer’s ability to effectively implement an FCC approach in secondary schools. These themes are consistently noted in athletic training and include limited education and resources, staffing and space concerns, non-technical skills, and social determinants of health. Staffing concerns for secondary school athletic trainers are not a new issue and can potentially affect the overall healthcare of athletes. McGuine et al. categorized over 2400 athletes based on athletic trainer availability and found that those with more access to an athletic trainer were positively influenced to report a sports-related concussion . In addition, subsequent post-concussion management was also enhanced by the presence of an athletic trainer. The problem identified in our research ( ) is that there is a shortage of athletic trainers employed in the secondary school setting for a variety of different reasons. Research in secondary schools suggests athletic directors and principals would like to employ athletic trainers in their schools, but they indicate several barriers . Athletic directors identify budget concerns, rural locations not close in proximity to hospitals/clinics, and misconceptions about the role of an athletic trainer as factors that contribute to their inability to hire athletic trainers . A recent study indicated 66% of secondary schools in the U.S. have access to athletic training services . The lack of adequate staffing in secondary schools is problematic for athletic trainers in providing quality healthcare, including FCC and other essential healthcare practices. Even when secondary schools are able to hire athletic trainers, our findings suggest they are inadequately trained or provided with the education and skills to implement FCC within their clinical practice. Athletic trainers see the need for continuing education to advance their practice , but evidence suggests there are barriers such as time, cost, and associated travel which ultimately deter the completion of continuing professional development . Professional athletic training education programs should incorporate the tenants of FCC instead of waiting for an individual to self-select continuing education opportunities on FCC. There is a need for professional education, as well as residency and fellowship experiences in athletic training, to align with the Accreditation Council for Graduate Medical Education (ACGME) standards for patient- and family-centered care . The Athletic Training Milestones have been established as a guiding assessment framework that mirrors the ACGME standards . While our study focused on the specific FCC scales, the ACGME Milestones describe general interpersonal and communication skills through the competency framework. We recommend future graduates of professional athletic training programs and all current practicing athletic trainers be able to (1) establish and maintain a therapeutic relationship using effective communication behaviors in challenging patient encounters and (2) identify complex barriers to effective communication, including personal bias . Findings from this study suggest athletic trainers need more formal training and education on FCC, as well as a better understanding of non-technical skills to integrate concepts of FCC into clinical practice. Participants from this study consistently spoke about the need for athletic trainers to use non-technical skills such as building trust, establishing relationships, and communicating appropriately with families to truly integrate FCC. A parent’s perception of an athletic trainer’s skillset is vital to the success of FCC. Previous research in secondary schools indicated that 55% of parents did not always trust the athletic trainer’s opinion . Weitzel et al. demonstrated that parents do value athletic trainers as important members of the secondary school healthcare team but also found that parents have varying perceptions of athletic trainers’ skill levels based on previous experiences . The interactions between the athletic trainer and family are critical and ultimately determine if the athletic trainer will be able to successfully implement FCC. The final theme to emerge from the qualitative data was the impact social determinants of health might play in implementing FCC. Athletic trainers indicated that the social determinants of health must be considered when trying to use an FCC approach, specifically in the secondary school setting. Research in athletic training has highlighted that athletic trainers can have a positive outcome in reducing negative social determinants of health experiences . A qualitative research study by Hernandez et al. examined secondary school athletic trainers’ experiences with providing healthcare to low socioeconomic status patient populations . Results were consistent with participant comments from our study, which both suggested athletic trainers are in a good position to support and advocate for low socioeconomic populations. Participants’ comments from this study also corroborated other findings from the qualitative analysis whereby participants shared that continuing professional development was needed and suggested that they were ill-equipped to navigate socioeconomic status challenges as they delivered care to patients in secondary schools. If athletic trainers in secondary schools want to incorporate FCC into their practice, more resources, including education (formal and informal) and the number of staff, will most certainly need to be provided. Many of the non-technical skills needed to care for patients will most likely be learned over time through experiences and interactions with different patients; however, there is a need for continuing professional development related to inclusion and equity for complex family dynamics. Athletic trainers will also need to gain a better understanding of the social determinants of health as they use FCC in their practice, as well as consider screening for these social factors to create individualized care plans and resources specific to the family . Athletic training education programs should focus on exposing students, both clinically and didactically, to the challenges identified in this research study to incorporate FCC into their practice. 4.5. Limitations and Future Research This study has potential limitations. While over 66% of secondary schools employ an athletic trainer, a small portion participated in this study, making the primary limitation the generalization of these results. Survey research in athletic training historically has suffered from small sample sizes, but the data from our study aligned with the sample size power analysis and should be shared to encourage future efforts surrounding FCC. We suggest attention should be placed on developing interventions and providing resources for secondary school athletic trainers to collaboratively work with patients and families in their setting. While the main concern of staffing and space cannot be directly addressed from the findings, we suggest formal education on delivering FCC is a key first step for stakeholder buy-in. Finally, future research should explore parent and family perceptions of their athletic trainers regarding the delivery of child- and family-centered care.
At the time of injury, secondary school athletes reflected that they worried, as well as their family and friends, about the severity of their injury . In addition, previous research has identified families are at risk of developing mental health challenges, such as post-traumatic stress disorder, after their child sustains an injury . This sense of worry can be reduced or eliminated through parent-to-parent support groups. As originally proposed by the Association for the Care of Children’s Health , parent-to-parent support means assistance provided by a parent who may be facing similar challenges and who can understand what one is going through. Parents have expressed this as a key element of FCC . Parents feel the power of shared experiences allows them to empathize and be validated when facing difficult challenges with the health of their children. Furthermore, dating back to 1986, Minna Nathanson outlined three important functions of parent-to-parent support, which include (1) mutual support and friendship, (2) information gathering and sharing, and (3) improving the system . Overall, parents want the opportunity to help each other identify information to navigate the hardships associated with caring for their children and develop support systems along the way. Athletic trainers reported parent-to-parent support as the highest perceived and most frequently practiced element of FCC. Interestingly, this element is not ranked highly across other healthcare professions for both perception and practice [ , , , , , , , ]. Athletic trainers reported practicing parent-to-parent support (mean = 3.93 ± 0.85) more frequently as compared to other professions, with mean scores ranging from 2.38 to 3.55 [ , , , , , ]. Similarly, athletic trainers perceived this as a necessary element more than other healthcare professionals. To our knowledge, the studies conducted by Franck et al. and Petersen et al. are the only studies conducted in the U.S. to date that utilized the FCCQ-R . Participants in these studies reported similar perceptions as athletic trainers on the parent-to-parent scale, with scores of 4.23 and 4.04. It can be possible that healthcare professionals in the U.S. understand and value the shared experiences that parents have with regard to the care of their children and may promote it more than in other countries. It is of future interest to explore how athletic trainers coordinate and provide parent-to-parent support in the secondary school setting.
Across athletic trainers, the least practiced and lowest perceived necessary element of FCC was developmental needs. In its original form, the incorporation of the developmental needs of children, adolescents, and their families into the healthcare systems was primarily intended to serve individuals with disabilities . The literature exploring FCC has predominately been conducted across nurses in pediatric or neonatal hospital units, where patients may face significant challenges with regard to growing up with disabilities and how to educate patients to be self-sufficient and independent [ , , ]. Findings indicate that most healthcare professionals practice this element of FCC and perceive it as necessary, with mean scores as high as 4.44 . In comparison, athletic trainers’ perception of developmental needs was 3.23. Secondary school athletic trainers do not typically face these developmental challenges, as they work with a population who is highly active and engaged in intense physical activity, and a possible explanation to why they did not perceive developmental needs as highly as other elements of FCC. Following an injury or illness, a patient may experience emotional and social developmental challenges that affect their interest and motivation. While the patients’ athletic trainers typically care for primarily able-bodied children, a focus needs to be placed on the emotional and social components when the child experiences an injury. Practicing and incorporating this element of FCC may be beneficial when caring for patients instead of a compartmentalized approach that focuses on the injury separate from the emotional and social aspects. To do so, we recommend athletic trainers approach patient care using the World Health Organization International Classification of Functioning, Disability, and Health (ICF) disablement model framework . The ICF model allows the clinician to explore other areas outside of the health condition, which could highlight developmental needs such as body functions, activity limitations, participation restrictions, environmental factors, and personal factors .
Athletic trainers provide services in secondary school after obtaining parental consent at the onset of the school year rather than at each patient encounter . This could be detrimental to the relationship and promoting FCC as it limits the connectivity between the parties, which reinforces the lack of knowledge and limited experiences for parents to meet and understand the role of the athletic trainer . Athletic trainers recognize that a relationship with parents is important and could serve as a barrier to providing FCC if not present. To implement FCC, it is imperative athletic trainers acknowledge that family is the constant construct. This means that as the adolescent discontinues sport participation, recovers from their injury or illness, and/or graduates from secondary school, their support system (i.e., parents, guardians, siblings) will continue while the athletic trainer’s role is temporary. While not the highest practiced element of FCC, the construct that the family is the constant had the lowest discrepancy between the CP and PN scales. Across the literature, this theme does not seem to be uncommon, with other health professions stating the same sentiment. In fact, the family is the constant scale is commonly the highest practiced FCC [ , , , , ]. Similarly, the implementation gap between perceptions and practice of this element of FCC has had the smallest differences across other studies [ , , , ]. Altogether, the “family as a constant” is a theme that health professionals, including athletic trainers, perceive as important when practicing FCC. Furthermore, it can be assumed that athletic trainers, in general, believe proper communication with the family and a good trusting relationship is vital for ensuring quality FCC. Unfortunately, athletic trainers in the secondary school setting cited parents as both the most common job-related stressor and source of conflict . There is a disconnect between the clinician-centered care approach from the athletic trainer resulting in an uninformed parent, a confused child, and a result-oriented athletic trainer. We recommend athletic trainers change their perspective on the role of the parent/family. The theory of social capital suggests leveraging the family as a resource to build relationships focused on trust, safety, and social cohesion to influence the sharing of information, increase health outcomes, and accomplish shared goals . In the case of a patient–provider relationship, the athletic trainer can build trusting relationships with the families to improve the minor’s health. A key strategy to incorporate is creating multiple communication options that meet the needs of the parents in terms of time and accessibility and avoiding medical jargon. The integration of FCC into clinical practice may increase the short-term burden of stress and conflict but ultimately should promote a common goal between the family and the athletic trainer. However, the “best interest” of the child held by the family and the athletic trainer may differ, resulting in conflict, power imbalances, and expert tension . A natural first step in changing the culture of healthcare for athletic trainers in the secondary school setting would be to explore the perception and reputation held by the student-athletes, family members, and personnel. The data should explore the current perceived training, skills, opportunities, and areas for necessary improvement. The commitment to engaging and embracing the family throughout the process could have positive, long-term benefits for addressing the healthcare of the adolescent. In addition, previous research has identified parents need greater education on issues that they can deem important for their children’s health . It is critical athletic trainers prioritize family education, as well as patient education, to encourage collaboration throughout the care plan. In doing so, the athletic trainer should explore the parent’s health literacy in the care decision-making process and provide education and options individualized to the family . Specifically, there is a need to encourage home care management facilitated by the parents or guardians. The active engagement of the family in the therapeutic intervention process could improve buy-in from both the patient and family and alleviate the perceived stressors and conflicts.
We identified four themes from the open-ended responses related to an athletic trainer’s ability to effectively implement an FCC approach in secondary schools. These themes are consistently noted in athletic training and include limited education and resources, staffing and space concerns, non-technical skills, and social determinants of health. Staffing concerns for secondary school athletic trainers are not a new issue and can potentially affect the overall healthcare of athletes. McGuine et al. categorized over 2400 athletes based on athletic trainer availability and found that those with more access to an athletic trainer were positively influenced to report a sports-related concussion . In addition, subsequent post-concussion management was also enhanced by the presence of an athletic trainer. The problem identified in our research ( ) is that there is a shortage of athletic trainers employed in the secondary school setting for a variety of different reasons. Research in secondary schools suggests athletic directors and principals would like to employ athletic trainers in their schools, but they indicate several barriers . Athletic directors identify budget concerns, rural locations not close in proximity to hospitals/clinics, and misconceptions about the role of an athletic trainer as factors that contribute to their inability to hire athletic trainers . A recent study indicated 66% of secondary schools in the U.S. have access to athletic training services . The lack of adequate staffing in secondary schools is problematic for athletic trainers in providing quality healthcare, including FCC and other essential healthcare practices. Even when secondary schools are able to hire athletic trainers, our findings suggest they are inadequately trained or provided with the education and skills to implement FCC within their clinical practice. Athletic trainers see the need for continuing education to advance their practice , but evidence suggests there are barriers such as time, cost, and associated travel which ultimately deter the completion of continuing professional development . Professional athletic training education programs should incorporate the tenants of FCC instead of waiting for an individual to self-select continuing education opportunities on FCC. There is a need for professional education, as well as residency and fellowship experiences in athletic training, to align with the Accreditation Council for Graduate Medical Education (ACGME) standards for patient- and family-centered care . The Athletic Training Milestones have been established as a guiding assessment framework that mirrors the ACGME standards . While our study focused on the specific FCC scales, the ACGME Milestones describe general interpersonal and communication skills through the competency framework. We recommend future graduates of professional athletic training programs and all current practicing athletic trainers be able to (1) establish and maintain a therapeutic relationship using effective communication behaviors in challenging patient encounters and (2) identify complex barriers to effective communication, including personal bias . Findings from this study suggest athletic trainers need more formal training and education on FCC, as well as a better understanding of non-technical skills to integrate concepts of FCC into clinical practice. Participants from this study consistently spoke about the need for athletic trainers to use non-technical skills such as building trust, establishing relationships, and communicating appropriately with families to truly integrate FCC. A parent’s perception of an athletic trainer’s skillset is vital to the success of FCC. Previous research in secondary schools indicated that 55% of parents did not always trust the athletic trainer’s opinion . Weitzel et al. demonstrated that parents do value athletic trainers as important members of the secondary school healthcare team but also found that parents have varying perceptions of athletic trainers’ skill levels based on previous experiences . The interactions between the athletic trainer and family are critical and ultimately determine if the athletic trainer will be able to successfully implement FCC. The final theme to emerge from the qualitative data was the impact social determinants of health might play in implementing FCC. Athletic trainers indicated that the social determinants of health must be considered when trying to use an FCC approach, specifically in the secondary school setting. Research in athletic training has highlighted that athletic trainers can have a positive outcome in reducing negative social determinants of health experiences . A qualitative research study by Hernandez et al. examined secondary school athletic trainers’ experiences with providing healthcare to low socioeconomic status patient populations . Results were consistent with participant comments from our study, which both suggested athletic trainers are in a good position to support and advocate for low socioeconomic populations. Participants’ comments from this study also corroborated other findings from the qualitative analysis whereby participants shared that continuing professional development was needed and suggested that they were ill-equipped to navigate socioeconomic status challenges as they delivered care to patients in secondary schools. If athletic trainers in secondary schools want to incorporate FCC into their practice, more resources, including education (formal and informal) and the number of staff, will most certainly need to be provided. Many of the non-technical skills needed to care for patients will most likely be learned over time through experiences and interactions with different patients; however, there is a need for continuing professional development related to inclusion and equity for complex family dynamics. Athletic trainers will also need to gain a better understanding of the social determinants of health as they use FCC in their practice, as well as consider screening for these social factors to create individualized care plans and resources specific to the family . Athletic training education programs should focus on exposing students, both clinically and didactically, to the challenges identified in this research study to incorporate FCC into their practice.
This study has potential limitations. While over 66% of secondary schools employ an athletic trainer, a small portion participated in this study, making the primary limitation the generalization of these results. Survey research in athletic training historically has suffered from small sample sizes, but the data from our study aligned with the sample size power analysis and should be shared to encourage future efforts surrounding FCC. We suggest attention should be placed on developing interventions and providing resources for secondary school athletic trainers to collaboratively work with patients and families in their setting. While the main concern of staffing and space cannot be directly addressed from the findings, we suggest formal education on delivering FCC is a key first step for stakeholder buy-in. Finally, future research should explore parent and family perceptions of their athletic trainers regarding the delivery of child- and family-centered care.
Athletic trainers in the secondary school setting are vital components of the healthcare team for middle and high school athletes in the US. Overall, most athletic trainers self-reported that they are performing all FCC aspects significantly less in their daily practice than they perceive necessary. The participants reported the FCC components were needed more in daily clinical practice than what was currently happening. In this setting, athletic trainers may feel compromised by time to educate, inform, and collaborate with the family to provide support. Participants noted barriers and concerns to FCC that may explain the significant differences in perception and practice.
|
Effects of Variation in | 328855c6-00b9-4625-a728-4948acd92069 | 10049481 | Microbiology[mh] | Floodplains, located in the transition zone between aquatic and terrestrial ecosystems, are characterized by high primary productivity and biological diversity . Due to the interference of climate change and human activities, floodplain ecosystem stability has been seriously disturbed, and this has resulted in soil erosion and land degradation . Aiming at preventing soil erosion, shrub vegetation restoration projects are more extensive and are an effective method to combat land degradation than herb and tree in floodplain ecosystems . Shrub plantation drives changes in soil physiochemical properties, soil organic matter (SOM) formation, and pyrolysis as well as vegetation structure composition and plant biomass allocation pattern [ , , ]. All the changes have profound impacts on soil microbial community composition and activity . Soil microorganisms play critical roles in nutrient cycling, carbon sequestration, and ecosystem stability in terrestrial ecosystems . Soil microbial biomass, community composition, and physiological activities are sensitive to aboveground vegetation and belowground conditions . Root-associated arbuscular mycorrhizal fungi acquired substantial amounts of nitrogen and phosphorous for plant uptake from decomposed organic matter . Bacteria are essential for affecting nutrient availability and transforming plant-derived and microbial-derived biomass into soil organic matter . Shrubs can maintain ecosystem functions by forming fertile islands beneath their canopies . Indeed, T. chinensis ( Tamarix chinensis Lour.) can grow in various shrub patch sizes, and soil properties and soil microbial community structure vary with shrub patch size . Variation in soil microbial structure is controlled by various biotic and abiotic factors . Shifts in the quantity and quality of carbon due to variations in plant communities affect the composition and function of microbial communities . Resource availability (e.g., water, soil organic carbon, and phosphorus) has a strong effect on the soil microbial community composition . In the nutrient fertilization experiment, spatial heterogeneity in the distribution of soil organic matter alleviates microbial carbon limitation . Moreover, shrub patch size has a significant effect on soil organic carbon, total nitrogen, and soil microbial biomass . The fungal abundance and biomass of Artemisia gmelinii are higher in inside-canopy soils than in outside-canopy soils due to the increased SOC input in semiarid land . The ability of shrubs to alter the accumulation of nutrients has been demonstrated in desert-wash shrub communities, shrub-dominated semi-desert ecosystems, semi-arid savanna, and alpine ecosystems [ , , , ]. However, the effects of patch size and sampling location around individual shrub plants on the relationship between soil microbial community structure and soil physicochemical properties have not been thoroughly examined. Tamarix chinensis Lour., northern Tamarisk, a shrub species with high coverage and high tolerance to soil erosion and land degradation, is a vital part of the ecological project of “Southern Mangrove and Northern Tamarisk” . This species is widely distributed in the middle Yellow River floodplain . The middle Yellow River floodplain is derived from the Loess Plateau and experiences severe soil erosion; nutrients in this region are thus being depleted from its fine-grained soils . Shrub patch size can affect the dynamics of soil physiochemical properties and alter soil microbial community composition [ , , ]. In turn, soil microorganisms such as bacteria and fungi can affect soil nutrient cycling processes by catalyzing decomposition of litter and SOM . Significant research has mainly focused on the soil properties under different shrub plantations [ , , , ], but the dynamics of microbial community composition under the T. chinensis plantations remain unclear. To a certain extent, the data deficiency has hindered the construction of T. chinensis shelter forest in the floodplain. In this paper, we investigated the effects of T. chinensis plantation of different shrub patch sizes and sampling locations on soil physicochemical properties and soil microbial community composition in the middle Yellow River floodplain. This study was motivated by two questions: (1) How do the variations of microbial community composition change through shrub patch size and sampling location following the T. chinensis plantation? (2) What were the underlying mechanisms influencing the responses of microbial community activities to T. chinensis plantation in the middle Yellow River floodplain?
2.1. Study Region Description This study was conducted at the Yellow River Floodplain Ecosystems Research Station (34°59′65″ N, 113°25′05″ E, 100 m a.s.l.), Xingyang County, Henan Province, China ( a). The region experiences a monsoon climate. The mean annual average temperature was 14.3 °C; the highest monthly mean temperature was 42.5 °C (in July), and the lowest monthly mean temperature was −9.6 °C (in January). The mean annual precipitation was 645.5 mm, and 65% of the precipitation fell between July and September. The soil at the study site was sandy, and the pH was approximately 8.10 . The vegetation is dominated by T. chinensis , Phragmites australis (Cav.) Trin., Cynodon dactylon (L.) Pers., Calamagrostis epigeios (L.) Roth., and Aster subulatus (Michx.) Nesom. . 2.2. Experimental Design and Sampling T. chinensis is an important component of the vegetation, and it grows in individual clumps of plants up to 300 cm across the canopy in the Middle Yellow River floodplain according to our survey data. This experiment used a randomized block design with four sites (20 m × 20 m) selected, which included the three sizes (small: <100 cm; medium: 100–200 cm, and large: >200 cm) in each site according to the canopy diameter, height, and number of stems of plants in late May 2021 ( b; ). We selected 12 T. chinensis plants for the three size categories, with four repeats for each size. Soil samples were collected along horizontal transects on two different radii surrounding the base of each T. chinensis stem ( c). The soil sampling locations were: (1) at the edge of middle radius of the canopy (inside canopy) and (2) at the edge of one and a half radius of the canopy (outside canopy). The mid-canopy samples were the most appropriate to represent soil physicochemical properties under canopies . The outside samples were at half a canopy radius beyond the drip line to avoid influences from neighboring shrubs . At a 0–15 cm soil depth, four soil samples in each radius were collected by a 4 cm inner diameter soil auger in East, South, West, and North, respectively, and then composited into one sample ( ) . There were, in total, 24 mixed soil samples which were passed through a 2 mm sieve and placed in sterile plastic pouches. Each sample was divided into two parts: one part of the sample was air-dried, ground, and used for the determination of soil physical and chemical properties, and the other part was stored in iceboxes for microbial analyses. 2.3. Analysis of Soil Properties Soil particle size analyses were conducted using the laser diffraction technique with a Longbench MasterSizer 2000 (Malvern Instruments, Malvern, England, UK) to calculate the percentage of clay, silt, fine sand, and coarse sand. The soil water content (SWC, g of water per 100 g of dry soil) was determined by oven drying (ZXRD_A7230, Zhicheng, Shanghai, China) samples at 105 °C to a constant weight and then taking weight measurements. Soil pH was measured using a pH meter (Sartorius PB-10, Göttingen, Germany) with a 2.5:1 ratio of deionized water/air-dried soil. The soil salt content (SSC) was determined using the gravimetric method . Soil organic carbon (SOC) was analyzed using dichromate-sulfuric acid oxidation with heating. SOC was converted to soil organic matter (SOM) by multiplying with the constant of 1.724 . Soil total nitrogen (TN) was measured using a Vario Max CNS elemental analyzer (Elementar Analysensysteme GmbH, Hanau, Germany). Available phosphorus (AP) was extracted with 0.5 M NaHCO3; measurements were then taken using the UV Spectrophotometer (Daojin UV-1900, Kyoto, Japan). 2.4. Phospholipid Fatty Acid Analysis Soil microbial community composition was characterized by analyzing PLFAs . Briefly, lipids were extracted from soil with a chloroform: methanol: 0.05 M sodium phosphate buffer mixture (1:2:0.8 ( v / v / v )); they were then separated into neutral lipids, glycolipids, and phospholipids using a pre-packed silica column. Phospholipids subjected to mild alkaline methanolysis and fatty acid methyl esters were identified using a gas chromatograph with a flame ionization detector (FID) (GC6890, Agilent Technologies, Bracknell, UK) with methyl nonadecanoate (19:0) as the internal standard. The abundance of individual fatty acid methylesters was expressed as nmol/g dry soil. The total microbial biomass was estimated as the sum of all the extracted PLFAs. Lipid markers associated with microbial functional groups were analyzed by summing their concentrations. The groups detected included bacteria (15:0, i15:0, i16:0, a16:0, 16:1 w 6c, a17:0, c17:0, i17:0, 17:1 w 8c, i17:1 w 9c, 18:1 w 5c, 18:1 w 7c, c19:0 w 8c, 20:1 w 9c) , fungi (16:1 w 5c, 18:1 w 9c, 18:2 w 6,9c) , Gram-positive bacteria (GP) (a15:0, i15:0, i16:0, a16:0, a17:0, i17:0, i17:1 w 9c), Gram-negative bacteria (GN) (16:1 w 6c, c17:0, 17:1 w 8c, 18:1 w 5c, 18:1 w 7c, c19:0 w 8c, 20:1 w 9c) , and arbuscular mycorrhizal fungi (16:1 w 5c) . 2.5. Statistical Analysis All data were expressed as mean ± standard error (SE) of the mean. All variables were transformed to the meet criteria for normality and homoscedasticity, but the results and figures are presented with untransformed values. Two-way analysis of variance was used to assess the effects of shrub patch size, sampling location, and their interaction on soil physicochemical properties and microbial communities. Changes in soil physicochemical properties under the three shrub patch sizes were evaluated using one-way ANOVA, and the differences among sampling locations were evaluated using paired t -tests. The least significant difference (LSD) test ( p < 0.05) was used to identify significant effects. The relationships between microbial communities and soil physicochemical properties were examined using Pearson correlation coefficients. The above statistical analyses were conducted in SPSS 20 (SPSS, Inc., Chicago, IL, USA). Redundancy analysis (RDA) in Canoco 5.0 was performed to determine the environmental factors that affected microbial community composition.
This study was conducted at the Yellow River Floodplain Ecosystems Research Station (34°59′65″ N, 113°25′05″ E, 100 m a.s.l.), Xingyang County, Henan Province, China ( a). The region experiences a monsoon climate. The mean annual average temperature was 14.3 °C; the highest monthly mean temperature was 42.5 °C (in July), and the lowest monthly mean temperature was −9.6 °C (in January). The mean annual precipitation was 645.5 mm, and 65% of the precipitation fell between July and September. The soil at the study site was sandy, and the pH was approximately 8.10 . The vegetation is dominated by T. chinensis , Phragmites australis (Cav.) Trin., Cynodon dactylon (L.) Pers., Calamagrostis epigeios (L.) Roth., and Aster subulatus (Michx.) Nesom. .
T. chinensis is an important component of the vegetation, and it grows in individual clumps of plants up to 300 cm across the canopy in the Middle Yellow River floodplain according to our survey data. This experiment used a randomized block design with four sites (20 m × 20 m) selected, which included the three sizes (small: <100 cm; medium: 100–200 cm, and large: >200 cm) in each site according to the canopy diameter, height, and number of stems of plants in late May 2021 ( b; ). We selected 12 T. chinensis plants for the three size categories, with four repeats for each size. Soil samples were collected along horizontal transects on two different radii surrounding the base of each T. chinensis stem ( c). The soil sampling locations were: (1) at the edge of middle radius of the canopy (inside canopy) and (2) at the edge of one and a half radius of the canopy (outside canopy). The mid-canopy samples were the most appropriate to represent soil physicochemical properties under canopies . The outside samples were at half a canopy radius beyond the drip line to avoid influences from neighboring shrubs . At a 0–15 cm soil depth, four soil samples in each radius were collected by a 4 cm inner diameter soil auger in East, South, West, and North, respectively, and then composited into one sample ( ) . There were, in total, 24 mixed soil samples which were passed through a 2 mm sieve and placed in sterile plastic pouches. Each sample was divided into two parts: one part of the sample was air-dried, ground, and used for the determination of soil physical and chemical properties, and the other part was stored in iceboxes for microbial analyses.
Soil particle size analyses were conducted using the laser diffraction technique with a Longbench MasterSizer 2000 (Malvern Instruments, Malvern, England, UK) to calculate the percentage of clay, silt, fine sand, and coarse sand. The soil water content (SWC, g of water per 100 g of dry soil) was determined by oven drying (ZXRD_A7230, Zhicheng, Shanghai, China) samples at 105 °C to a constant weight and then taking weight measurements. Soil pH was measured using a pH meter (Sartorius PB-10, Göttingen, Germany) with a 2.5:1 ratio of deionized water/air-dried soil. The soil salt content (SSC) was determined using the gravimetric method . Soil organic carbon (SOC) was analyzed using dichromate-sulfuric acid oxidation with heating. SOC was converted to soil organic matter (SOM) by multiplying with the constant of 1.724 . Soil total nitrogen (TN) was measured using a Vario Max CNS elemental analyzer (Elementar Analysensysteme GmbH, Hanau, Germany). Available phosphorus (AP) was extracted with 0.5 M NaHCO3; measurements were then taken using the UV Spectrophotometer (Daojin UV-1900, Kyoto, Japan).
Soil microbial community composition was characterized by analyzing PLFAs . Briefly, lipids were extracted from soil with a chloroform: methanol: 0.05 M sodium phosphate buffer mixture (1:2:0.8 ( v / v / v )); they were then separated into neutral lipids, glycolipids, and phospholipids using a pre-packed silica column. Phospholipids subjected to mild alkaline methanolysis and fatty acid methyl esters were identified using a gas chromatograph with a flame ionization detector (FID) (GC6890, Agilent Technologies, Bracknell, UK) with methyl nonadecanoate (19:0) as the internal standard. The abundance of individual fatty acid methylesters was expressed as nmol/g dry soil. The total microbial biomass was estimated as the sum of all the extracted PLFAs. Lipid markers associated with microbial functional groups were analyzed by summing their concentrations. The groups detected included bacteria (15:0, i15:0, i16:0, a16:0, 16:1 w 6c, a17:0, c17:0, i17:0, 17:1 w 8c, i17:1 w 9c, 18:1 w 5c, 18:1 w 7c, c19:0 w 8c, 20:1 w 9c) , fungi (16:1 w 5c, 18:1 w 9c, 18:2 w 6,9c) , Gram-positive bacteria (GP) (a15:0, i15:0, i16:0, a16:0, a17:0, i17:0, i17:1 w 9c), Gram-negative bacteria (GN) (16:1 w 6c, c17:0, 17:1 w 8c, 18:1 w 5c, 18:1 w 7c, c19:0 w 8c, 20:1 w 9c) , and arbuscular mycorrhizal fungi (16:1 w 5c) .
All data were expressed as mean ± standard error (SE) of the mean. All variables were transformed to the meet criteria for normality and homoscedasticity, but the results and figures are presented with untransformed values. Two-way analysis of variance was used to assess the effects of shrub patch size, sampling location, and their interaction on soil physicochemical properties and microbial communities. Changes in soil physicochemical properties under the three shrub patch sizes were evaluated using one-way ANOVA, and the differences among sampling locations were evaluated using paired t -tests. The least significant difference (LSD) test ( p < 0.05) was used to identify significant effects. The relationships between microbial communities and soil physicochemical properties were examined using Pearson correlation coefficients. The above statistical analyses were conducted in SPSS 20 (SPSS, Inc., Chicago, IL, USA). Redundancy analysis (RDA) in Canoco 5.0 was performed to determine the environmental factors that affected microbial community composition.
3.1. Shrub Growth Characters and Soil Physicochemical Properties The basal diameter, height, and canopy diameter of T. chinensis significantly increased with shrub patch size ( p < 0.001; ). The clay content from small to large shrub patch size changed from 5.54% to 7.82% in inside-canopy soils and from 4.79% to 7.36% in outside-canopy soils. The silt content from small to large shrub patch size altered from 63.65% to 83.59% in inside-canopy soils and from 57.18% to 85.68% in outside-canopy soils. The sand content from small to large shrub patch size varied from 8.58% to 31.76% in inside-canopy soils and from 6.96% to 39.10% in outside-canopy soils ( ). The SSC, SOM, TN, and AP content significantly varied with shrub patch size across the two sampling locations (all p < 0.05; ). Averaged across the three shrub patch sizes, the SSC, SOM, TN, and AP were larger in inside-canopy soils than in outside-canopy soils. The SSC, SOM, and TN of inside-canopy soils in the large shrub patch size were 155.47%, 80.33%, and 43.17% higher than the SSC, SOM, and TN of outside-canopy soils in the same shrub patch size, respectively, and these differences were significant. Available p was 20.23%, 24.50%, and 54.64% higher in the small, medium, and large shrub patch size in inside-canopy soils than in outside-canopy soils, respectively, and these differences were significant (all p < 0.05; ). The SSC, SOM, TN, and AP in inside-canopy soils increased by 86.51%, 75.57%, 38.12%, and 39.03% from the small to large shrub patch size, respectively (all p < 0.05; ). The SSC, SOM, and AP in outside-canopy soils increased by 6.15%, 17.42%, and 8.09% from the small to large shrub patch size, respectively, but the TN decreased by 1.19%. There was a strong interaction effect between shrub patch size and sampling location on SSC, SOM, TN, and AP ( ). The increases in SSC, SOM, TN, and AP from small to large shrub patch size across both sampling locations were 1.07 g·kg −1 , 2.62 g·kg −1 , 0.29 g·kg −1 , and 14.33 mg·kg −1 . The value of soil pH decreased with shrub patch size and was lowest under a large patch size in both sampling locations. Neither shrub patch size nor sampling location affected the soil water content. 3.2. Effects of Shrub Patch Size and Sampling Location on PLFAs Microbial community composition varied with shrub patch size and sampling location. Significant interactive effects of total PLFAs, as well as bacterial, fungal, GN bacterial, and AMF PLFAs with shrub patch size and sampling location were detected (all p < 0.05; ). In inside-canopy soils, the total PLFAs, as well as bacterial, GP bacterial, GN bacterial, fungal, and AMF PLFAs, were increased by 41.07%, 43.61%, 29.33%, 53.72%, 28.58%, and 127.66% from the small to large shrub patch size, respectively, and these increases were significant. In outside-canopy soils, the total PLFAs, as well as bacterial, GP bacterial, GN bacterial, fungal, and AMF PLFAs, were increased by 28.77%, 27.09%, 13.32%, 27.46%, 39.04%, and 132.43% from the small to large shrub patch size, respectively, and these increases were significant (all p < 0.05; and ; ). The total PLFAs, as well as bacterial, GP bacterial, and GN bacterial PLFAs, were significantly higher in inside-canopy soils than outside-canopy soils in the large shrub patch size ( p < 0.05; and ; ). The fungal and AMF PLFAs significantly differed between inside-canopy and outside-canopy soils in all three shrub patch sizes. The fungal PLFAs of inside-canopy soils were 33.42%, 10.69%, and 23.38% higher in the small, medium, and large shrub patch sizes than in outside-canopy soils, respectively. The AMF PLFAs of inside-canopy soils were 80.66%, 49.71%, and 76.94% higher in the small, medium, and large shrub patch sizes than in outside-canopy soils, respectively ( p < 0.05; ; ). There were significant interaction effects between shrub patch size and sampling location on the ratio of fungi to bacteria and GP to GN bacteria ( p < 0.05; ). The ratio of fungi to bacteria in the inside-canopy soil decreased with shrub patch size, but the ratio of fungi to bacteria in the outside-canopy soils increased with shrub patch size ( p < 0.05; d; ). The ratio of GP to GN bacteria was highest in the small shrub patch size and lowest in the medium shrub patch size in outside-canopy soils ( p < 0.05; c; ). 3.3. Correlations between Soil Microbial Community Composition and Physicochemical Properties Pearson correlation analysis showed that total PLFAs, bacterial, fungal, GP bacterial, GN bacterial, and AMF PLFAs were positively correlated with clay, silt, SSC, SOM, TN, and AP but negatively correlated with soil sand and pH ( p < 0.05). The ratio of GP to GN bacteria was positively related to soil sand ( p < 0.01), but negatively related to clay, silt, SSC, TN, and AP ( p < 0.05) in inside-canopy soils under all shrub patch sizes ( a). In outside-canopy soils, GP bacteria was positively associated with soil silt but negatively associated with soil sand ( p < 0.05; b). RDA showed that variation among microbial functional groups was associated with specific soil physicochemical properties ( ). All the physicochemical properties explained 76.10% (axis 1: 68.67%; axis 2: 4.69%) and 48.50% (axis 1: 34.78%; axis 2: 9.26%) of the variance in inside-canopy and outside-canopy soils, respectively. In the inside-canopy soils, the SOM was the most significant variable selected by the forward selection, and it explained 61.90% of the variation in the PLFA data, followed by the AP. The SOM was mainly related to the Gram-positive bacteria PLFA marker (i17:1 w 9c) and Gram-negative bacteria PLFA marker (17:1 w 8c, 10me–16:0), and AP was correlated with the AMF PLFA marker (16:1 w 9c) ( p < 0.05).
The basal diameter, height, and canopy diameter of T. chinensis significantly increased with shrub patch size ( p < 0.001; ). The clay content from small to large shrub patch size changed from 5.54% to 7.82% in inside-canopy soils and from 4.79% to 7.36% in outside-canopy soils. The silt content from small to large shrub patch size altered from 63.65% to 83.59% in inside-canopy soils and from 57.18% to 85.68% in outside-canopy soils. The sand content from small to large shrub patch size varied from 8.58% to 31.76% in inside-canopy soils and from 6.96% to 39.10% in outside-canopy soils ( ). The SSC, SOM, TN, and AP content significantly varied with shrub patch size across the two sampling locations (all p < 0.05; ). Averaged across the three shrub patch sizes, the SSC, SOM, TN, and AP were larger in inside-canopy soils than in outside-canopy soils. The SSC, SOM, and TN of inside-canopy soils in the large shrub patch size were 155.47%, 80.33%, and 43.17% higher than the SSC, SOM, and TN of outside-canopy soils in the same shrub patch size, respectively, and these differences were significant. Available p was 20.23%, 24.50%, and 54.64% higher in the small, medium, and large shrub patch size in inside-canopy soils than in outside-canopy soils, respectively, and these differences were significant (all p < 0.05; ). The SSC, SOM, TN, and AP in inside-canopy soils increased by 86.51%, 75.57%, 38.12%, and 39.03% from the small to large shrub patch size, respectively (all p < 0.05; ). The SSC, SOM, and AP in outside-canopy soils increased by 6.15%, 17.42%, and 8.09% from the small to large shrub patch size, respectively, but the TN decreased by 1.19%. There was a strong interaction effect between shrub patch size and sampling location on SSC, SOM, TN, and AP ( ). The increases in SSC, SOM, TN, and AP from small to large shrub patch size across both sampling locations were 1.07 g·kg −1 , 2.62 g·kg −1 , 0.29 g·kg −1 , and 14.33 mg·kg −1 . The value of soil pH decreased with shrub patch size and was lowest under a large patch size in both sampling locations. Neither shrub patch size nor sampling location affected the soil water content.
Microbial community composition varied with shrub patch size and sampling location. Significant interactive effects of total PLFAs, as well as bacterial, fungal, GN bacterial, and AMF PLFAs with shrub patch size and sampling location were detected (all p < 0.05; ). In inside-canopy soils, the total PLFAs, as well as bacterial, GP bacterial, GN bacterial, fungal, and AMF PLFAs, were increased by 41.07%, 43.61%, 29.33%, 53.72%, 28.58%, and 127.66% from the small to large shrub patch size, respectively, and these increases were significant. In outside-canopy soils, the total PLFAs, as well as bacterial, GP bacterial, GN bacterial, fungal, and AMF PLFAs, were increased by 28.77%, 27.09%, 13.32%, 27.46%, 39.04%, and 132.43% from the small to large shrub patch size, respectively, and these increases were significant (all p < 0.05; and ; ). The total PLFAs, as well as bacterial, GP bacterial, and GN bacterial PLFAs, were significantly higher in inside-canopy soils than outside-canopy soils in the large shrub patch size ( p < 0.05; and ; ). The fungal and AMF PLFAs significantly differed between inside-canopy and outside-canopy soils in all three shrub patch sizes. The fungal PLFAs of inside-canopy soils were 33.42%, 10.69%, and 23.38% higher in the small, medium, and large shrub patch sizes than in outside-canopy soils, respectively. The AMF PLFAs of inside-canopy soils were 80.66%, 49.71%, and 76.94% higher in the small, medium, and large shrub patch sizes than in outside-canopy soils, respectively ( p < 0.05; ; ). There were significant interaction effects between shrub patch size and sampling location on the ratio of fungi to bacteria and GP to GN bacteria ( p < 0.05; ). The ratio of fungi to bacteria in the inside-canopy soil decreased with shrub patch size, but the ratio of fungi to bacteria in the outside-canopy soils increased with shrub patch size ( p < 0.05; d; ). The ratio of GP to GN bacteria was highest in the small shrub patch size and lowest in the medium shrub patch size in outside-canopy soils ( p < 0.05; c; ).
Pearson correlation analysis showed that total PLFAs, bacterial, fungal, GP bacterial, GN bacterial, and AMF PLFAs were positively correlated with clay, silt, SSC, SOM, TN, and AP but negatively correlated with soil sand and pH ( p < 0.05). The ratio of GP to GN bacteria was positively related to soil sand ( p < 0.01), but negatively related to clay, silt, SSC, TN, and AP ( p < 0.05) in inside-canopy soils under all shrub patch sizes ( a). In outside-canopy soils, GP bacteria was positively associated with soil silt but negatively associated with soil sand ( p < 0.05; b). RDA showed that variation among microbial functional groups was associated with specific soil physicochemical properties ( ). All the physicochemical properties explained 76.10% (axis 1: 68.67%; axis 2: 4.69%) and 48.50% (axis 1: 34.78%; axis 2: 9.26%) of the variance in inside-canopy and outside-canopy soils, respectively. In the inside-canopy soils, the SOM was the most significant variable selected by the forward selection, and it explained 61.90% of the variation in the PLFA data, followed by the AP. The SOM was mainly related to the Gram-positive bacteria PLFA marker (i17:1 w 9c) and Gram-negative bacteria PLFA marker (17:1 w 8c, 10me–16:0), and AP was correlated with the AMF PLFA marker (16:1 w 9c) ( p < 0.05).
4.1. Effect of T. chinensis on Soil Properties Previous studies demonstrated that the fertile island of shrub enriched soil available nutrients under their canopy, such as SOM, N, and P, supplying more substrates for soil microbes [ , , , ]. Here, our results found shrub patch size had a positive effect on soil nutrient status. The SOM, TN, and AP were greater in the inside-canopy soils than in outside-canopy soils under all three shrub patch sizes. The increase in soil nutrients under a canopy may be induced by the microbial decomposition of fallen leaves and root turnover . Aboveground plant litter might be the sources of soil nitrogen, especially under deciduous shrubs, which may enhance nitrogen nutrition . Furthermore, dust particles rich in nitrogen and other mineral elements that accumulate in the leaves and branches could be transported to the soil through stemflow and throughfall . T. chinensis generated a “fertile island effect” that could promote the accumulation of soil nutrients, and it tended to increase in strength with shrub patch size . The soil organic matter was 20.61%, 56.43%, and 80.33% higher in the small, medium, and large patch size, respectively, in inside-canopy soils than in outside-canopy soils. The significant difference in soil nutrients between inside-canopy and outside-canopy soils in the three patch sizes indicates that nutrient accumulation occurs over long periods rather than short periods. Increases in soil nutrient conditions might stem from the litter enrichment of shrubs with a larger canopy, which is consistent with previous observations in an arid desert in northwest China . As shrub patch size increased, increases in nutrient availability, especially SOM, under large shrub patch sizes, were mainly attributed to gradual changes in biogeochemical cycles, such as litter decomposition, rhizosphere secretions, and root turnover . Furthermore, shifts in soil physicochemical properties among shrub patch sizes and sampling locations might reflect changes in soil microbial communities. 4.2. Effect of T. chinensis on Soil Microbial Communities Soil microorganisms, which have an intricate relationship with soil physicochemical properties, are sensitive to shrub patch size and sampling location [ , , ]. Previous studies have shown that the growth of vegetated patches has a neutral , promotion , or inhibition effect on soil microbial activities. The results of our study showed that the growth of vegetated patches has a positive effect on soil microbial communities. Indeed, the abundances of total PLFAs, bacteria, and fungi were significantly higher in inside-canopy soils than in outside-canopy soils and increased with shrub patch size. As shrub patch size increased, increases in GN bacteria were greater than increases in GP bacteria, which caused the ratio of GP to GN bacteria to decrease. High-nutrient environments favor the growth of r-strategy microbes, which are GN bacteria, and enhance nutrient cycling rates . Higher concentrations of soil organic matter and total nitrogen might be linked to variations in fungal abundances and microbial biomass [ , , ]. The main drivers of variation in soil microbial structure were changes in soil organic matter, total nitrogen, and available phosphorus, which is consistent with the results of previous studies in the middle Yellow River floodplain of China . AMF PLFAs increased with shrub patch size, which is consistent with the results of previous studies . Significant positive correlations between the soil salt content and AMF PLFAs suggest that islands of salinity provide substrates that promote the growth of AMF by the colonization of spores . The soil salt content increased with shrub patch size at the two sampling locations. Increases in AMF can result in the generation of a large network of mycelia as the mycorrhizal roots extend, and this might have a positive effect on the nutrient uptake and soil organic matter accumulation of T. chinensis plants . The growth of AMF can promote the formation and stabilization of soil aggregates and thus contribute to soil quality . 4.3. Implications Vegetation restoration can promote increases in soil nutrients and soil quality [ , , ]. In general, our study shows that T. chinensis plantations can enhance the content of soil nutrients and thus improve soil microbial community composition. The abundance of AMF increased with shrub patch size; increases in AMF promote plant growth and improve soil quality by altering the nutrient uptake capacity of plant roots . Therefore, T. chinensis plantations could be planted in the middle Yellow River to promote ecological restoration efforts and improve the quality of the soil. Due to the one-time sampling in this research, the results may just be revealed as a snapshot, and further future long-term investigations concerning the lasting influence of T. chinensis plantations are required in this area.
Previous studies demonstrated that the fertile island of shrub enriched soil available nutrients under their canopy, such as SOM, N, and P, supplying more substrates for soil microbes [ , , , ]. Here, our results found shrub patch size had a positive effect on soil nutrient status. The SOM, TN, and AP were greater in the inside-canopy soils than in outside-canopy soils under all three shrub patch sizes. The increase in soil nutrients under a canopy may be induced by the microbial decomposition of fallen leaves and root turnover . Aboveground plant litter might be the sources of soil nitrogen, especially under deciduous shrubs, which may enhance nitrogen nutrition . Furthermore, dust particles rich in nitrogen and other mineral elements that accumulate in the leaves and branches could be transported to the soil through stemflow and throughfall . T. chinensis generated a “fertile island effect” that could promote the accumulation of soil nutrients, and it tended to increase in strength with shrub patch size . The soil organic matter was 20.61%, 56.43%, and 80.33% higher in the small, medium, and large patch size, respectively, in inside-canopy soils than in outside-canopy soils. The significant difference in soil nutrients between inside-canopy and outside-canopy soils in the three patch sizes indicates that nutrient accumulation occurs over long periods rather than short periods. Increases in soil nutrient conditions might stem from the litter enrichment of shrubs with a larger canopy, which is consistent with previous observations in an arid desert in northwest China . As shrub patch size increased, increases in nutrient availability, especially SOM, under large shrub patch sizes, were mainly attributed to gradual changes in biogeochemical cycles, such as litter decomposition, rhizosphere secretions, and root turnover . Furthermore, shifts in soil physicochemical properties among shrub patch sizes and sampling locations might reflect changes in soil microbial communities.
Soil microorganisms, which have an intricate relationship with soil physicochemical properties, are sensitive to shrub patch size and sampling location [ , , ]. Previous studies have shown that the growth of vegetated patches has a neutral , promotion , or inhibition effect on soil microbial activities. The results of our study showed that the growth of vegetated patches has a positive effect on soil microbial communities. Indeed, the abundances of total PLFAs, bacteria, and fungi were significantly higher in inside-canopy soils than in outside-canopy soils and increased with shrub patch size. As shrub patch size increased, increases in GN bacteria were greater than increases in GP bacteria, which caused the ratio of GP to GN bacteria to decrease. High-nutrient environments favor the growth of r-strategy microbes, which are GN bacteria, and enhance nutrient cycling rates . Higher concentrations of soil organic matter and total nitrogen might be linked to variations in fungal abundances and microbial biomass [ , , ]. The main drivers of variation in soil microbial structure were changes in soil organic matter, total nitrogen, and available phosphorus, which is consistent with the results of previous studies in the middle Yellow River floodplain of China . AMF PLFAs increased with shrub patch size, which is consistent with the results of previous studies . Significant positive correlations between the soil salt content and AMF PLFAs suggest that islands of salinity provide substrates that promote the growth of AMF by the colonization of spores . The soil salt content increased with shrub patch size at the two sampling locations. Increases in AMF can result in the generation of a large network of mycelia as the mycorrhizal roots extend, and this might have a positive effect on the nutrient uptake and soil organic matter accumulation of T. chinensis plants . The growth of AMF can promote the formation and stabilization of soil aggregates and thus contribute to soil quality .
Vegetation restoration can promote increases in soil nutrients and soil quality [ , , ]. In general, our study shows that T. chinensis plantations can enhance the content of soil nutrients and thus improve soil microbial community composition. The abundance of AMF increased with shrub patch size; increases in AMF promote plant growth and improve soil quality by altering the nutrient uptake capacity of plant roots . Therefore, T. chinensis plantations could be planted in the middle Yellow River to promote ecological restoration efforts and improve the quality of the soil. Due to the one-time sampling in this research, the results may just be revealed as a snapshot, and further future long-term investigations concerning the lasting influence of T. chinensis plantations are required in this area.
The results of this study demonstrated that T. chinensis plants induced the enrichment of soil nutrients and microbial communities via the fertile island effect. Shifts in microbial community composition were mainly associated with changes in SOM, TN, and AP. The observed variation in soil microbial communities and soil nutrients suggests that T. chinensis could be used to promote ecological restoration efforts and alleviate the effects of land degradation in the middle Yellow River floodplain of China.
|
Patient Education Improves Pain and Health-Related Quality of Life in Patients with Established Spinal Osteoporosis in Primary Care—A Pilot Study of Short- and Long-Term Effects | 0b8d09be-3315-4f75-87a9-899fda958c82 | 10049553 | Patient Education as Topic[mh] | Osteoporosis is defined as a systemic skeletal disease characterized by low bone mass and micro-architectural deterioration of bone tissue, leading to enhanced bone fragility and a consequent increase in fracture risk . Fragility fractures, in specific vertebral and hip fractures, are associated with high morbidity, mortality and socioeconomic costs [ , , , , ], and are also associated with an increased risk of new fractures, often in the near future . Patients with a vertebral fracture (VF) often have lifelong and opioid-requiring pain as well as functional disabilities and reduced health-related quality of life (HRQoL) . Thus, it is important to assess these patients for osteoporosis and to initiate adequate pharmacological treatment as well as to work on preventable risk factors, such as the risk of a fall (balance, functional disabilities, and home environment), nutrition, and adherence to treatment. Though well known, these adjustable risk factors are not systematically worked upon in many health organizations handling patients with osteoporosis. In most organizations, the primary health care provider is responsible for their osteoporosis patients and organizing and securing this interdisciplinary teamwork. Osteoporosis schools, i.e., patient education with an interdisciplinary focus, are part of some health organizations. The schools’ content, both regarding theory parts and possible physical training, vary as well as the timeframe and different patient categories included with regard to fracture history . Theory content in the group’s education often focuses on knowledge of osteoporosis, medication, and nutrition . Furthermore, many osteoporosis schools include physical activity in various arrangements concerning the frequency and specific activity. Though the results are inconsistent, probably because of heterogenous study designs and groups, some studies have shown a positive effect of patient education and training on patients’ HRQoL, and patient empowerment . There is an obvious need for high-quality randomized controlled trials to evaluate the effectiveness of patient education combined with physical training in people with osteoporosis . In the present study including patients with established spinal osteoporosis, we aimed to investigate the short- and long-term effects of interdisciplinary patient education with or without physical training or mindfulness and medical yoga on (1) chronic pain; (2) HRQoL; (3) physical strength and balance performance; (4) fall risk and physical activity; (5) theoretical knowledge; and (6) patient enablement.
2.1. Study Design This pilot randomized study was called the School of Osteoporosis in Linköping (the SOL study) and included a 10-week intervention period with once-weekly theory education with or without additional physical training. Furthermore, a preceding observation period of 10 weeks, as well as a 1-year post-intervention follow-up, were included in the study design ( ). The study protocol has been registered at ClinicalTrials.gov, (ClinicalTrials.gov Identifier: NCT05227976) 12 January 2022. https://clinicaltrials.gov/ct2/show/NCT05227976 (assessed on 9 March 2023). 2.2. Participants To be included in the SOL study, participants had to meet four criteria: (1) diagnosed with established spinal osteoporosis (at least one vertebral fracture and osteoporosis); (2) >3 months had passed since the most recent VF; (3) age ≥60 years; and (4) the physical ability to walk without an indoor walker. Patients with an inability to understand the Swedish language or difficulty following the research protocol, or dementia were excluded. The patients fulfilling the criteria are mostly followed up by the primary health care in Sweden. 2.3. Study Procedures Participants were recruited by means of advertisements through the regional patient organization, local newspapers, primary health care centers and the osteoporosis unit (i.e., the unit performing dual-energy X-ray absorptiometry [DXA] scans for all patients in the region). A research nurse made the screening process by phone calls. Clinical testing and a questionnaire-based evaluation were performed at baseline (May 2018), post-observation (Aug./Sept. 2018), and post-intervention (Nov./Dec. 2018) ( ). The 1-year post-intervention follow-up (Nov./Dec. 2019) was based on a questionnaire, solely ( ). The questionnaires were sent by postal mail and were answered at home. At post-intervention, the participants and therapists were asked about their experiences with the study procedures and interventions. The randomization was done blind by research staff after the baseline tests, where the subjects’ hidden names were drawn consecutively to one of the three groups, i.e., (1) theory only (T group); (2) theory and physical exercise (TPh group); and (3) theory and mindfulness/medical yoga (TMMY group). 2.4. Observation During the four-month non-interventional observation period, participants were asked to live as usual. The data were analyzed as pooled data ( n = 21), as no intervention had been done prior to the observation. 2.5. Interventions All three intervention arms included the same theoretical lectures organized as a 1-h weekly theory session for 10 weeks. In addition, the TPh and TMMY groups had a 1-h training session scheduled in adjunction to the theory sessions, respectively. A coffee break was included as a social event for each meeting. The theory themes were (1) osteoporosis and physical activity (led by a physiotherapist); (2) diagnosis of osteoporosis and pharmacological treatment, lasting two sessions (led by a physician who specialized in endocrinology/osteoporosis); (3) mindfulness and medical yoga (led by a physiotherapist/MMY teacher); (4) orthopedic technician aspects of activating spinal orthosis and stable shoes (led by an orthopedic technician and a representative from an orthosis company); (5) nutritional aspects (led by a dietitian); (6) balance performance and balance training (led by a physiotherapist); (7) information from the regional patient association for osteoporosis (led by two representatives of the local patient organization); (8) ergonomic aspects concerning daily living activities and adequate technical support (led by an occupational therapist); and (9) physiology of pain (led by a member of a team working with interdisciplinary pain rehabilitation at Linköping University Hospital). The theory sessions were organized by a moderator and were conducted in a conference room at Linköping University for the T group, whereas the TPh and TMMY groups had theory lessons together in a conference room at the training center. An experienced physiotherapist, with knowledge of appropriate training for persons with osteoporotic vertebral fractures, supervised the TPh group for 45 min once a week prior to the theory sessions. The exercise program started with a warm-up phase for six minutes and was followed by circuit training (performed by standing or walking) at nine training stations focusing on muscle strength and balance exercises for forty-five seconds times three sets. The sessions ended with a 5 min cool-down and stretching. All participants in the TPh group also received a home training program. The TMMY sessions were equally scheduled prior to the theory sessions and started with thirty-minute modified medical yoga exercises and yoga meditations (sitting on comfortable chairs) and ended with one leg standing. The yoga poses for the back were modified to suit each participant’s individual needs and there were neither extreme positions, strenuous spinal flexion exercises, exercises with rotation nor exercises bending the trunk to an end-range position . Furthermore, the TMMY group practiced the mindfulness concept for another 30 min with mainly breathing and awareness exercises. Mindfulness theory as well as weekly training follow-ups were part of the concept, and the participants also received a CD with mindfulness awareness exercises on their first session for daily home practice. An experienced physiotherapist, who was also a mindfulness/medical yoga instructor, supervised the TMMY sessions . 2.6. Outcomes The patient-reported questionnaires consisted of two parts: (1) a self-constructed questionnaire including clinical background data, present medication, including pain killers, history of falls, self-estimated disease knowledge, and performed physical activity; and (2) validated instruments about pain, HRQoL, fall risk evaluation, physical activity, and patient enablement. 2.6.1. Pain Present pain, pain from last week, and worst pain were estimated using a numeric rating scale (NRS), ranging from 0 = no pain to 10 = worst possible pain at the clinical testing occasions. Furthermore, patients were asked about their usage of painkillers (including the type and regularity). Effects on pain were also assessed by the HRQoL instruments (below). 2.6.2. Health-Related Quality of Life The measurement of HRQoL included generic instruments (the European quality of life, EQ-5D-3L, and RAND-36), as well as a disease-specific (established spinal osteoporosis) instrument, the “quality of life questionnaire in the European Foundation for Osteoporosis-41” (Qualeffo-41) [ , , , ]. The EQ-5D-3L comprises five dimensions i.e., mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. These dimensions are combined with an EQ-5D-3L index normalized to a reference ranging from −0.594 to 1, where 1 indicates optimal health . The RAND-36 comprises 36 items organized into eight different health domains i.e., physical function (PF), role physical (RP), bodily pain (BP), general health (GH), vitality (VT), social function (SF), role emotional (RE), and mental health (MH) . The scores were transformed into a 0–100 scale (0 = worst possible HRQoL and 100 = the best possible). The Qualeffo-41 includes 41 questions summarized into seven domains: (1) pain (backache); (2) activities of daily living (ADL); (3) jobs (around the house); (4) mobility; (5) social function; (6) general health perception; and (7) mental function . In contrast to RAND-36, 0 indicates the best possible and 100 the worst possible HRQoL. 2.6.3. Physical Strength, Balance Performance, and Anthropometry Static and dynamic balance tests were performed, including the one-leg standing (right and left) tests that were done with the opposite foot positioned on the calf of the tested leg and the arms along the sides with the eyes open and then closed, respectively, and were limited to a maximum of 30 s. The dynamic balance tests were done by tandem walking heel-to-toe forwards, and toe-to-heel backwards on a line. The number of steps were counted and maximized to 15 correct steps. The balance tests were performed three times and the best trial was used as the final score . In the chair-stand test, the participants were asked to rise as many times as possible from a standard chair with knees bent at a 90° angle for 30 s without the assistance of the arms . Grip strength (kg) of the dominant and the non-dominant hand was measured by the standard Jamar dynamometer. Each test was performed three times and the best trial was used . The distance (cm) between the seventh cervical vertebra (C7) and the wall was measured by a folding ruler to estimate the back-straightening ability . Body height was measured with a stadiometer and body weight with a digital scale. All clinical tests were performed by a physiotherapist with extensive clinical and research experience and great familiarity with the tests. 2.6.4. Fall Risk and Physical Activity The Swedish Falls Efficacy Scale International (FES-I) comprises 16 items about the risk of falling (including a four-level scale, where 1 = not at all concerned and 4 = very concerned). The individual scores were summarized to a total score . The questionnaire also comprised questions about physical exercise, physical activity, and sedentary behavior . 2.6.5. Theoretical Knowledge Assessment To evaluate the effect of the patients’ education, their knowledge of osteoporosis was assessed by ten open-ended questions at baseline and post-intervention. The questions tested a basic knowledge of osteoporosis, medication, exercise and fall prevention, and were produced and evaluated by experienced university teachers. The total score was between 0 (worst knowledge) and 22 (best knowledge). 2.6.6. Patient Enablement and Overall Experiences of the SOL The Patient Enablement Instrument (PEI) was used to measure the extent to which a patient can understand and cope with his or her illness after an intervention . The PEI consisted of six questions starting with “As a result of the osteoporosis school, do you feel you are …” followed by the alternative answers, where score 2 was for much better/much more; score 1 was for better/more; and score 0 was for same, less, or not applicable. The total score (range 0–12) was calculated for those participants who answered at least five questions. A higher score indicated higher enablement . The Swedish version has been shown to have acceptable validity and reliability for patients with chronic pain . The participants’ overall experiences of the theoretical lectures and the physical training were scored on a six-level scale (where 5 was considered very satisfied and 0 was not at all satisfied) post-intervention.
This pilot randomized study was called the School of Osteoporosis in Linköping (the SOL study) and included a 10-week intervention period with once-weekly theory education with or without additional physical training. Furthermore, a preceding observation period of 10 weeks, as well as a 1-year post-intervention follow-up, were included in the study design ( ). The study protocol has been registered at ClinicalTrials.gov, (ClinicalTrials.gov Identifier: NCT05227976) 12 January 2022. https://clinicaltrials.gov/ct2/show/NCT05227976 (assessed on 9 March 2023).
To be included in the SOL study, participants had to meet four criteria: (1) diagnosed with established spinal osteoporosis (at least one vertebral fracture and osteoporosis); (2) >3 months had passed since the most recent VF; (3) age ≥60 years; and (4) the physical ability to walk without an indoor walker. Patients with an inability to understand the Swedish language or difficulty following the research protocol, or dementia were excluded. The patients fulfilling the criteria are mostly followed up by the primary health care in Sweden.
Participants were recruited by means of advertisements through the regional patient organization, local newspapers, primary health care centers and the osteoporosis unit (i.e., the unit performing dual-energy X-ray absorptiometry [DXA] scans for all patients in the region). A research nurse made the screening process by phone calls. Clinical testing and a questionnaire-based evaluation were performed at baseline (May 2018), post-observation (Aug./Sept. 2018), and post-intervention (Nov./Dec. 2018) ( ). The 1-year post-intervention follow-up (Nov./Dec. 2019) was based on a questionnaire, solely ( ). The questionnaires were sent by postal mail and were answered at home. At post-intervention, the participants and therapists were asked about their experiences with the study procedures and interventions. The randomization was done blind by research staff after the baseline tests, where the subjects’ hidden names were drawn consecutively to one of the three groups, i.e., (1) theory only (T group); (2) theory and physical exercise (TPh group); and (3) theory and mindfulness/medical yoga (TMMY group).
During the four-month non-interventional observation period, participants were asked to live as usual. The data were analyzed as pooled data ( n = 21), as no intervention had been done prior to the observation.
All three intervention arms included the same theoretical lectures organized as a 1-h weekly theory session for 10 weeks. In addition, the TPh and TMMY groups had a 1-h training session scheduled in adjunction to the theory sessions, respectively. A coffee break was included as a social event for each meeting. The theory themes were (1) osteoporosis and physical activity (led by a physiotherapist); (2) diagnosis of osteoporosis and pharmacological treatment, lasting two sessions (led by a physician who specialized in endocrinology/osteoporosis); (3) mindfulness and medical yoga (led by a physiotherapist/MMY teacher); (4) orthopedic technician aspects of activating spinal orthosis and stable shoes (led by an orthopedic technician and a representative from an orthosis company); (5) nutritional aspects (led by a dietitian); (6) balance performance and balance training (led by a physiotherapist); (7) information from the regional patient association for osteoporosis (led by two representatives of the local patient organization); (8) ergonomic aspects concerning daily living activities and adequate technical support (led by an occupational therapist); and (9) physiology of pain (led by a member of a team working with interdisciplinary pain rehabilitation at Linköping University Hospital). The theory sessions were organized by a moderator and were conducted in a conference room at Linköping University for the T group, whereas the TPh and TMMY groups had theory lessons together in a conference room at the training center. An experienced physiotherapist, with knowledge of appropriate training for persons with osteoporotic vertebral fractures, supervised the TPh group for 45 min once a week prior to the theory sessions. The exercise program started with a warm-up phase for six minutes and was followed by circuit training (performed by standing or walking) at nine training stations focusing on muscle strength and balance exercises for forty-five seconds times three sets. The sessions ended with a 5 min cool-down and stretching. All participants in the TPh group also received a home training program. The TMMY sessions were equally scheduled prior to the theory sessions and started with thirty-minute modified medical yoga exercises and yoga meditations (sitting on comfortable chairs) and ended with one leg standing. The yoga poses for the back were modified to suit each participant’s individual needs and there were neither extreme positions, strenuous spinal flexion exercises, exercises with rotation nor exercises bending the trunk to an end-range position . Furthermore, the TMMY group practiced the mindfulness concept for another 30 min with mainly breathing and awareness exercises. Mindfulness theory as well as weekly training follow-ups were part of the concept, and the participants also received a CD with mindfulness awareness exercises on their first session for daily home practice. An experienced physiotherapist, who was also a mindfulness/medical yoga instructor, supervised the TMMY sessions .
The patient-reported questionnaires consisted of two parts: (1) a self-constructed questionnaire including clinical background data, present medication, including pain killers, history of falls, self-estimated disease knowledge, and performed physical activity; and (2) validated instruments about pain, HRQoL, fall risk evaluation, physical activity, and patient enablement. 2.6.1. Pain Present pain, pain from last week, and worst pain were estimated using a numeric rating scale (NRS), ranging from 0 = no pain to 10 = worst possible pain at the clinical testing occasions. Furthermore, patients were asked about their usage of painkillers (including the type and regularity). Effects on pain were also assessed by the HRQoL instruments (below). 2.6.2. Health-Related Quality of Life The measurement of HRQoL included generic instruments (the European quality of life, EQ-5D-3L, and RAND-36), as well as a disease-specific (established spinal osteoporosis) instrument, the “quality of life questionnaire in the European Foundation for Osteoporosis-41” (Qualeffo-41) [ , , , ]. The EQ-5D-3L comprises five dimensions i.e., mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. These dimensions are combined with an EQ-5D-3L index normalized to a reference ranging from −0.594 to 1, where 1 indicates optimal health . The RAND-36 comprises 36 items organized into eight different health domains i.e., physical function (PF), role physical (RP), bodily pain (BP), general health (GH), vitality (VT), social function (SF), role emotional (RE), and mental health (MH) . The scores were transformed into a 0–100 scale (0 = worst possible HRQoL and 100 = the best possible). The Qualeffo-41 includes 41 questions summarized into seven domains: (1) pain (backache); (2) activities of daily living (ADL); (3) jobs (around the house); (4) mobility; (5) social function; (6) general health perception; and (7) mental function . In contrast to RAND-36, 0 indicates the best possible and 100 the worst possible HRQoL. 2.6.3. Physical Strength, Balance Performance, and Anthropometry Static and dynamic balance tests were performed, including the one-leg standing (right and left) tests that were done with the opposite foot positioned on the calf of the tested leg and the arms along the sides with the eyes open and then closed, respectively, and were limited to a maximum of 30 s. The dynamic balance tests were done by tandem walking heel-to-toe forwards, and toe-to-heel backwards on a line. The number of steps were counted and maximized to 15 correct steps. The balance tests were performed three times and the best trial was used as the final score . In the chair-stand test, the participants were asked to rise as many times as possible from a standard chair with knees bent at a 90° angle for 30 s without the assistance of the arms . Grip strength (kg) of the dominant and the non-dominant hand was measured by the standard Jamar dynamometer. Each test was performed three times and the best trial was used . The distance (cm) between the seventh cervical vertebra (C7) and the wall was measured by a folding ruler to estimate the back-straightening ability . Body height was measured with a stadiometer and body weight with a digital scale. All clinical tests were performed by a physiotherapist with extensive clinical and research experience and great familiarity with the tests. 2.6.4. Fall Risk and Physical Activity The Swedish Falls Efficacy Scale International (FES-I) comprises 16 items about the risk of falling (including a four-level scale, where 1 = not at all concerned and 4 = very concerned). The individual scores were summarized to a total score . The questionnaire also comprised questions about physical exercise, physical activity, and sedentary behavior . 2.6.5. Theoretical Knowledge Assessment To evaluate the effect of the patients’ education, their knowledge of osteoporosis was assessed by ten open-ended questions at baseline and post-intervention. The questions tested a basic knowledge of osteoporosis, medication, exercise and fall prevention, and were produced and evaluated by experienced university teachers. The total score was between 0 (worst knowledge) and 22 (best knowledge). 2.6.6. Patient Enablement and Overall Experiences of the SOL The Patient Enablement Instrument (PEI) was used to measure the extent to which a patient can understand and cope with his or her illness after an intervention . The PEI consisted of six questions starting with “As a result of the osteoporosis school, do you feel you are …” followed by the alternative answers, where score 2 was for much better/much more; score 1 was for better/more; and score 0 was for same, less, or not applicable. The total score (range 0–12) was calculated for those participants who answered at least five questions. A higher score indicated higher enablement . The Swedish version has been shown to have acceptable validity and reliability for patients with chronic pain . The participants’ overall experiences of the theoretical lectures and the physical training were scored on a six-level scale (where 5 was considered very satisfied and 0 was not at all satisfied) post-intervention.
Present pain, pain from last week, and worst pain were estimated using a numeric rating scale (NRS), ranging from 0 = no pain to 10 = worst possible pain at the clinical testing occasions. Furthermore, patients were asked about their usage of painkillers (including the type and regularity). Effects on pain were also assessed by the HRQoL instruments (below).
The measurement of HRQoL included generic instruments (the European quality of life, EQ-5D-3L, and RAND-36), as well as a disease-specific (established spinal osteoporosis) instrument, the “quality of life questionnaire in the European Foundation for Osteoporosis-41” (Qualeffo-41) [ , , , ]. The EQ-5D-3L comprises five dimensions i.e., mobility, self-care, usual activities, pain/discomfort, and anxiety/depression. These dimensions are combined with an EQ-5D-3L index normalized to a reference ranging from −0.594 to 1, where 1 indicates optimal health . The RAND-36 comprises 36 items organized into eight different health domains i.e., physical function (PF), role physical (RP), bodily pain (BP), general health (GH), vitality (VT), social function (SF), role emotional (RE), and mental health (MH) . The scores were transformed into a 0–100 scale (0 = worst possible HRQoL and 100 = the best possible). The Qualeffo-41 includes 41 questions summarized into seven domains: (1) pain (backache); (2) activities of daily living (ADL); (3) jobs (around the house); (4) mobility; (5) social function; (6) general health perception; and (7) mental function . In contrast to RAND-36, 0 indicates the best possible and 100 the worst possible HRQoL.
Static and dynamic balance tests were performed, including the one-leg standing (right and left) tests that were done with the opposite foot positioned on the calf of the tested leg and the arms along the sides with the eyes open and then closed, respectively, and were limited to a maximum of 30 s. The dynamic balance tests were done by tandem walking heel-to-toe forwards, and toe-to-heel backwards on a line. The number of steps were counted and maximized to 15 correct steps. The balance tests were performed three times and the best trial was used as the final score . In the chair-stand test, the participants were asked to rise as many times as possible from a standard chair with knees bent at a 90° angle for 30 s without the assistance of the arms . Grip strength (kg) of the dominant and the non-dominant hand was measured by the standard Jamar dynamometer. Each test was performed three times and the best trial was used . The distance (cm) between the seventh cervical vertebra (C7) and the wall was measured by a folding ruler to estimate the back-straightening ability . Body height was measured with a stadiometer and body weight with a digital scale. All clinical tests were performed by a physiotherapist with extensive clinical and research experience and great familiarity with the tests.
The Swedish Falls Efficacy Scale International (FES-I) comprises 16 items about the risk of falling (including a four-level scale, where 1 = not at all concerned and 4 = very concerned). The individual scores were summarized to a total score . The questionnaire also comprised questions about physical exercise, physical activity, and sedentary behavior .
To evaluate the effect of the patients’ education, their knowledge of osteoporosis was assessed by ten open-ended questions at baseline and post-intervention. The questions tested a basic knowledge of osteoporosis, medication, exercise and fall prevention, and were produced and evaluated by experienced university teachers. The total score was between 0 (worst knowledge) and 22 (best knowledge).
The Patient Enablement Instrument (PEI) was used to measure the extent to which a patient can understand and cope with his or her illness after an intervention . The PEI consisted of six questions starting with “As a result of the osteoporosis school, do you feel you are …” followed by the alternative answers, where score 2 was for much better/much more; score 1 was for better/more; and score 0 was for same, less, or not applicable. The total score (range 0–12) was calculated for those participants who answered at least five questions. A higher score indicated higher enablement . The Swedish version has been shown to have acceptable validity and reliability for patients with chronic pain . The participants’ overall experiences of the theoretical lectures and the physical training were scored on a six-level scale (where 5 was considered very satisfied and 0 was not at all satisfied) post-intervention.
Descriptive statistics at the different time points were reported with the median (Md) and interquartile range (IQR 25–75%), mean (M) and standard deviation (SD), and numbers and percentages. The Wilcoxon signed-rank test was chosen for comparing the change in each group over time. The relationships between the HRQoL measures RAND-36 GH, Qualeffo-41 GH and the total score, and the EQ-5D index, were investigated with Spearman’s rank correlation, using the following coefficients: 0–0.25 none to little; 0.25–0.50 fair; 0.50–0.75 moderate to good; >0.75 very good to excellent . All statistical tests were performed at the 5% significance level. For the statistical analyses, SPSS 25.0 software (IBM Statistics, New York, NY, USA) was used.
The inclusion period was set to four weeks. Sixty-two persons were interested in participating in the study and were contacted, but 50% ( n = 31) did not meet the inclusion criteria. The most common reason was that they had no vertebral fracture ( n = 23). An appropriate number of ten persons per group was set before the inclusion. However, as a total of thirty-one participants (two men and twenty-nine women) met the inclusion criteria during the enrollment period, all were included and randomized to either the T group (10 participants), TPh group (11 participants), or TMMY group (10 participants). Five people dropped out immediately after randomization, where three drop-outs were in the T group, one in the TPh group and one in the TMMY group. The reasons given for non-participation were the disappointment in being randomized to the T group, economic reasons, transportation to the education sessions, scheduled surgery, and serious illness of a family member. Another person in the TMMY group dropped out from the initiated intervention due to relocating to a new residence. Thus, the total number of study participants for the intervention period was 25. Several participants lived in other places than Linköping, at most 70 km from the SOL. There was a high attendance rate to the intervention program, with an average of nine out of ten (range 6–10) participants attending the SOL sessions. Twenty-one subjects (84%, 21/25) participated in the 1-year post-intervention follow-up. The four participants who dropped out of the one-year post-intervention follow-up were two participants from the T group; one from the TPh group; and one from the TMMY group. In the present study, only those 21 participants with complete data from baseline to the 1-year post-intervention follow-up were analyzed. 4.1. Background Characteristics The baseline median (range) age for all participants (twenty women and one man) was seventy-two years (60–82). The median age was 72 years (67–82) in the T group ( n = 5), 72 years (60–81) in the TPh group ( n = 9), and 72 years (63–81) in the TMMY group ( n = 7), without significant differences between the groups. Fifty-six percent of the participants had a history of fractures in addition to vertebral fractures. The most common fracture was the distal forearm reported by four participants. The majority of participants (70%) used regular painkillers (any type). Opioids were used regularly by 25% of participants. Most participants were clinically considered to have primary osteoporosis. Some participants had risk factors that might also affect the skeleton (secondary osteoporosis). This included six patients with thyroid disease and levothyroxine medication, one patient with chronic obstructive lung disease, and one person with previous breast cancer. In the final cohort ( n = 21), there were no participants using continuous glucocorticoids or with a known inherited bone metabolic disease. At the baseline, 62% had an ongoing antiresorptive treatment (alendronic acid [ n = 2], zoledronic acid [ n = 8], and denosumab [ n = 5]). No patients had an ongoing bone anabolic treatment. Calcium and/or vitamin D supplements were taken by 95% of participants. The present study focused on elderly community dwellers in primary care. No participants said that they had a community-based home help service. All participants were independent of indoor walking aids when included, as one of our exclusion criteria was a dependency on such aids. One person in each group was dependent on outdoor walking aids and some also used walking poles to support their balance during walks. 4.2. Observation Period During the non-interventional observation period, no changes were seen regarding pain from last week, worst pain, HRQoL (RAND-36, Qualeffo-41, and EQ-5D), fall risk (FES-I), and physical activity ( ). Significant improvements were seen regarding distance C7-wall (Md 6.5 cm vs. 6 cm, p = 0.025) and the chair-stand test (Md 9 vs. 11, p = 0.002), but not in the other clinical tests. 4.3. Intervention and Long-Term Effect 4.3.1. Pain The NRS score for “worst pain” improved from 7.8 (Md) at baseline to 6.3 post-intervention ( p = 0.013), and “pain last week” improved from 5 (Md) to 4.5 ( p = 0.042) when analyzing the pooled data of all participants ( ). When analyzing the separate intervention groups, the TPh group showed close to significant improvement regarding NRS scores for both variables ( p = 0.058 and p = 0.106). This pilot study was, however, not powered for subgroup analysis, which is why its significance should be interpreted cautiously, and its trends should be verified in larger cohorts. The percentage of all participants using opioids decreased from 25% (baseline) to 14% (post-intervention) and increased to 19% again at the 1-year post-intervention follow-up ( ). About two-thirds (70%) of the participants used painkillers regularly at baseline. After the intervention, this figure improved to 48% and remained at that level at the 1-year follow-up. 4.3.2. Health-Related Quality of Life The pooled data of all the participants showed a significantly improved scoring of social function ( p = 0.048), and close to significant scoring of physical function ( p = 0.060) and mental health ( p = 0.071) post-intervention compared to the baseline for RAND-36 ( ). No change was seen regarding RAND-36 dimensions at the 1-year post-intervention follow-up compared to post-intervention. When analyzing the separate intervention groups, the TPh group showed a significantly improved score regarding physical function ( p = 0.041) and close to significant for bodily pain ( p = 0.063). The TMMY group showed a significantly improved mental health score ( p = 0.040) and close to significant improvement in social function ( p = 0.066). No significant change was seen in the T group. These results should be interpreted cautiously, as the number of participants was very small. Regarding the Qualeffo-41, the social function domain was improved post-intervention ( p = 0.024) in the pooled data for all participants ( ). At the 1-year post-intervention follow-up, no change was seen in any Qualeffo-41 domain compared to post-intervention. For the separate intervention group analysis, the pain improved post-intervention in the TPh group ( p = 0.046), the social function improved in the TMMY group ( p = 0.043) and the mental health score almost significantly improved in the T group ( p = 0.068). No significant change was seen when analyzing the pooled data from EQ-5D neither at the baseline versus post-intervention nor post-intervention versus the 1-year post-intervention follow-up. Only the TPh group showed a trend of improved EQ-5D post-intervention ( p = 0.068). 4.3.3. Physical Strength, Balance Performance, and Anthropometry The pooled data from all participants showed improved values post-intervention regarding the chair-stand test ( p = 0.005), one-leg stand using the left leg with the eyes closed ( p = 0.030), and tandem walking backwards ( p = 0.027) ( ). The one-leg stand on the right leg with the eyes open showed worse results post-intervention when pooling data ( p = 0.019). However, there was no change when analyzing the separate intervention groups. The analysis of the intervention groups showed post-intervention balance improvements mostly in the TPh group, such as tandem walking forwards (close to significant p = 0.068), one-leg stance with eyes closed (right leg p = 0.040 and left leg p = 0.026) and close to significant improvements on the chair-stand test ( p = 0.075). The TMMY group showed improved chair-stand test results post-intervention ( p = 0.018). No change was seen regarding handgrip strength, C7-wall distance, body weight, or height. 4.3.4. Fall Risk and Physical Activity No change was seen in FES-I scoring neither post-intervention nor at the 1-year post-intervention follow-up ( ). There were five participants both at the baseline and post-intervention who reported physical activity of less than 150 min/week. At the 1-year post-intervention follow-up, six people reported insufficient physical activity (one person had recently suffered from a calcaneus fracture). There was no significant change in the reported total physical activity in the intervention groups between the baseline and post-intervention or between the post-intervention and 1-year post-intervention follow-up. However, a trend of more physical activity was seen in the T group ( p = 0.066). The sedentary inactive time (sitting/resting) did not change between the baseline and follow-ups. 4.3.5. Theoretical Knowledge The participants scored significantly better in the knowledge testing about osteoporosis post-intervention compared to the baseline. The median knowledge score increased from 15 at the baseline to 18 post-intervention (68% vs. 81% correct answers, p = 0.001), when analyzing the pooled data of all participants. The median score increased from 15 to 19 (68% vs. 86%, p = 0.043) in the T group, from 11 to 18 (50% vs. 81%, p = 0.011) in the TPh group, and from 17 to 20 (77% vs. 91%, p = 0.018) in the TMMY group. 4.3.6. Patient Enablement and Overall Experiences of the SOL The mean (SD) PEI total score post-intervention was 4.6 (2.8) (Md = 5, range 0–9). The mean (SD) values for the groups were 3.0 (2.9) for the T group, 3.6 (2.9) for the TPh group, and 5.8 (2.4) for the TMMY group. Improved patient enablement, i.e., “better” or “much better”, was reported by 40–92% of participants in each question, as shown in . The question with the highest percentage of participants reporting improved enablement was “(Q2)… able to understand your illness/disabilities” (92% of participants), and the question with the lowest improvement was “(Q5)… confident about your health” (40% of participants). The mean (SD) PEI total score at the 1-year follow-up was 4.1 (3.4) (Md = 4, range 0–11). The mean (SD) values for the groups were 3.8 (3.8) for the T group, 2.7 (2.6) for the TPh group, and 6.3 (3.1) for the TMMY group. The question with the highest percentage of participants reporting improved enablement was “(Q2)… able to understand your illness/disabilities” (76% of participants), and the question with the lowest improvement was “(Q5)… confident about your health” (33% of participants). The PEI total score did not change significantly between the post-intervention and 1-year post-intervention follow-up ( p = 0.390). The theoretical lectures were generally appreciated by the participants, scoring a mean of 4.4 (range 4.0–4.6) and a median of 5 (range 4.5–5). Similarly, the participants in the physical intervention groups scored highly regarding their satisfaction with the mindfulness/medical yoga activities (mean 4.7/Md 5), and also the physical training activities (mean 4.9/Md 5), respectively. 4.3.7. Adverse Events One participant in the TMMY group reported a rib fracture just before the start of the intervention. There was one participant in the TPh group who had a finger fracture during the group training activities. At the 1-year post-intervention follow-up, a vertebral fracture was reported in the T group, a calcaneus fracture (after falling from a ladder) was reported in the TPh group, and a rib fracture and a sacral fracture were reported in the TMMY group. At the baseline, one participant in the TPh group reported a fall in the past month, while one participant in the T group reported a fall at the end of the intervention period. No participants reported a low energy fall at the 1-year post-intervention follow-up.
The baseline median (range) age for all participants (twenty women and one man) was seventy-two years (60–82). The median age was 72 years (67–82) in the T group ( n = 5), 72 years (60–81) in the TPh group ( n = 9), and 72 years (63–81) in the TMMY group ( n = 7), without significant differences between the groups. Fifty-six percent of the participants had a history of fractures in addition to vertebral fractures. The most common fracture was the distal forearm reported by four participants. The majority of participants (70%) used regular painkillers (any type). Opioids were used regularly by 25% of participants. Most participants were clinically considered to have primary osteoporosis. Some participants had risk factors that might also affect the skeleton (secondary osteoporosis). This included six patients with thyroid disease and levothyroxine medication, one patient with chronic obstructive lung disease, and one person with previous breast cancer. In the final cohort ( n = 21), there were no participants using continuous glucocorticoids or with a known inherited bone metabolic disease. At the baseline, 62% had an ongoing antiresorptive treatment (alendronic acid [ n = 2], zoledronic acid [ n = 8], and denosumab [ n = 5]). No patients had an ongoing bone anabolic treatment. Calcium and/or vitamin D supplements were taken by 95% of participants. The present study focused on elderly community dwellers in primary care. No participants said that they had a community-based home help service. All participants were independent of indoor walking aids when included, as one of our exclusion criteria was a dependency on such aids. One person in each group was dependent on outdoor walking aids and some also used walking poles to support their balance during walks.
During the non-interventional observation period, no changes were seen regarding pain from last week, worst pain, HRQoL (RAND-36, Qualeffo-41, and EQ-5D), fall risk (FES-I), and physical activity ( ). Significant improvements were seen regarding distance C7-wall (Md 6.5 cm vs. 6 cm, p = 0.025) and the chair-stand test (Md 9 vs. 11, p = 0.002), but not in the other clinical tests.
4.3.1. Pain The NRS score for “worst pain” improved from 7.8 (Md) at baseline to 6.3 post-intervention ( p = 0.013), and “pain last week” improved from 5 (Md) to 4.5 ( p = 0.042) when analyzing the pooled data of all participants ( ). When analyzing the separate intervention groups, the TPh group showed close to significant improvement regarding NRS scores for both variables ( p = 0.058 and p = 0.106). This pilot study was, however, not powered for subgroup analysis, which is why its significance should be interpreted cautiously, and its trends should be verified in larger cohorts. The percentage of all participants using opioids decreased from 25% (baseline) to 14% (post-intervention) and increased to 19% again at the 1-year post-intervention follow-up ( ). About two-thirds (70%) of the participants used painkillers regularly at baseline. After the intervention, this figure improved to 48% and remained at that level at the 1-year follow-up. 4.3.2. Health-Related Quality of Life The pooled data of all the participants showed a significantly improved scoring of social function ( p = 0.048), and close to significant scoring of physical function ( p = 0.060) and mental health ( p = 0.071) post-intervention compared to the baseline for RAND-36 ( ). No change was seen regarding RAND-36 dimensions at the 1-year post-intervention follow-up compared to post-intervention. When analyzing the separate intervention groups, the TPh group showed a significantly improved score regarding physical function ( p = 0.041) and close to significant for bodily pain ( p = 0.063). The TMMY group showed a significantly improved mental health score ( p = 0.040) and close to significant improvement in social function ( p = 0.066). No significant change was seen in the T group. These results should be interpreted cautiously, as the number of participants was very small. Regarding the Qualeffo-41, the social function domain was improved post-intervention ( p = 0.024) in the pooled data for all participants ( ). At the 1-year post-intervention follow-up, no change was seen in any Qualeffo-41 domain compared to post-intervention. For the separate intervention group analysis, the pain improved post-intervention in the TPh group ( p = 0.046), the social function improved in the TMMY group ( p = 0.043) and the mental health score almost significantly improved in the T group ( p = 0.068). No significant change was seen when analyzing the pooled data from EQ-5D neither at the baseline versus post-intervention nor post-intervention versus the 1-year post-intervention follow-up. Only the TPh group showed a trend of improved EQ-5D post-intervention ( p = 0.068). 4.3.3. Physical Strength, Balance Performance, and Anthropometry The pooled data from all participants showed improved values post-intervention regarding the chair-stand test ( p = 0.005), one-leg stand using the left leg with the eyes closed ( p = 0.030), and tandem walking backwards ( p = 0.027) ( ). The one-leg stand on the right leg with the eyes open showed worse results post-intervention when pooling data ( p = 0.019). However, there was no change when analyzing the separate intervention groups. The analysis of the intervention groups showed post-intervention balance improvements mostly in the TPh group, such as tandem walking forwards (close to significant p = 0.068), one-leg stance with eyes closed (right leg p = 0.040 and left leg p = 0.026) and close to significant improvements on the chair-stand test ( p = 0.075). The TMMY group showed improved chair-stand test results post-intervention ( p = 0.018). No change was seen regarding handgrip strength, C7-wall distance, body weight, or height. 4.3.4. Fall Risk and Physical Activity No change was seen in FES-I scoring neither post-intervention nor at the 1-year post-intervention follow-up ( ). There were five participants both at the baseline and post-intervention who reported physical activity of less than 150 min/week. At the 1-year post-intervention follow-up, six people reported insufficient physical activity (one person had recently suffered from a calcaneus fracture). There was no significant change in the reported total physical activity in the intervention groups between the baseline and post-intervention or between the post-intervention and 1-year post-intervention follow-up. However, a trend of more physical activity was seen in the T group ( p = 0.066). The sedentary inactive time (sitting/resting) did not change between the baseline and follow-ups. 4.3.5. Theoretical Knowledge The participants scored significantly better in the knowledge testing about osteoporosis post-intervention compared to the baseline. The median knowledge score increased from 15 at the baseline to 18 post-intervention (68% vs. 81% correct answers, p = 0.001), when analyzing the pooled data of all participants. The median score increased from 15 to 19 (68% vs. 86%, p = 0.043) in the T group, from 11 to 18 (50% vs. 81%, p = 0.011) in the TPh group, and from 17 to 20 (77% vs. 91%, p = 0.018) in the TMMY group. 4.3.6. Patient Enablement and Overall Experiences of the SOL The mean (SD) PEI total score post-intervention was 4.6 (2.8) (Md = 5, range 0–9). The mean (SD) values for the groups were 3.0 (2.9) for the T group, 3.6 (2.9) for the TPh group, and 5.8 (2.4) for the TMMY group. Improved patient enablement, i.e., “better” or “much better”, was reported by 40–92% of participants in each question, as shown in . The question with the highest percentage of participants reporting improved enablement was “(Q2)… able to understand your illness/disabilities” (92% of participants), and the question with the lowest improvement was “(Q5)… confident about your health” (40% of participants). The mean (SD) PEI total score at the 1-year follow-up was 4.1 (3.4) (Md = 4, range 0–11). The mean (SD) values for the groups were 3.8 (3.8) for the T group, 2.7 (2.6) for the TPh group, and 6.3 (3.1) for the TMMY group. The question with the highest percentage of participants reporting improved enablement was “(Q2)… able to understand your illness/disabilities” (76% of participants), and the question with the lowest improvement was “(Q5)… confident about your health” (33% of participants). The PEI total score did not change significantly between the post-intervention and 1-year post-intervention follow-up ( p = 0.390). The theoretical lectures were generally appreciated by the participants, scoring a mean of 4.4 (range 4.0–4.6) and a median of 5 (range 4.5–5). Similarly, the participants in the physical intervention groups scored highly regarding their satisfaction with the mindfulness/medical yoga activities (mean 4.7/Md 5), and also the physical training activities (mean 4.9/Md 5), respectively. 4.3.7. Adverse Events One participant in the TMMY group reported a rib fracture just before the start of the intervention. There was one participant in the TPh group who had a finger fracture during the group training activities. At the 1-year post-intervention follow-up, a vertebral fracture was reported in the T group, a calcaneus fracture (after falling from a ladder) was reported in the TPh group, and a rib fracture and a sacral fracture were reported in the TMMY group. At the baseline, one participant in the TPh group reported a fall in the past month, while one participant in the T group reported a fall at the end of the intervention period. No participants reported a low energy fall at the 1-year post-intervention follow-up.
The NRS score for “worst pain” improved from 7.8 (Md) at baseline to 6.3 post-intervention ( p = 0.013), and “pain last week” improved from 5 (Md) to 4.5 ( p = 0.042) when analyzing the pooled data of all participants ( ). When analyzing the separate intervention groups, the TPh group showed close to significant improvement regarding NRS scores for both variables ( p = 0.058 and p = 0.106). This pilot study was, however, not powered for subgroup analysis, which is why its significance should be interpreted cautiously, and its trends should be verified in larger cohorts. The percentage of all participants using opioids decreased from 25% (baseline) to 14% (post-intervention) and increased to 19% again at the 1-year post-intervention follow-up ( ). About two-thirds (70%) of the participants used painkillers regularly at baseline. After the intervention, this figure improved to 48% and remained at that level at the 1-year follow-up.
The pooled data of all the participants showed a significantly improved scoring of social function ( p = 0.048), and close to significant scoring of physical function ( p = 0.060) and mental health ( p = 0.071) post-intervention compared to the baseline for RAND-36 ( ). No change was seen regarding RAND-36 dimensions at the 1-year post-intervention follow-up compared to post-intervention. When analyzing the separate intervention groups, the TPh group showed a significantly improved score regarding physical function ( p = 0.041) and close to significant for bodily pain ( p = 0.063). The TMMY group showed a significantly improved mental health score ( p = 0.040) and close to significant improvement in social function ( p = 0.066). No significant change was seen in the T group. These results should be interpreted cautiously, as the number of participants was very small. Regarding the Qualeffo-41, the social function domain was improved post-intervention ( p = 0.024) in the pooled data for all participants ( ). At the 1-year post-intervention follow-up, no change was seen in any Qualeffo-41 domain compared to post-intervention. For the separate intervention group analysis, the pain improved post-intervention in the TPh group ( p = 0.046), the social function improved in the TMMY group ( p = 0.043) and the mental health score almost significantly improved in the T group ( p = 0.068). No significant change was seen when analyzing the pooled data from EQ-5D neither at the baseline versus post-intervention nor post-intervention versus the 1-year post-intervention follow-up. Only the TPh group showed a trend of improved EQ-5D post-intervention ( p = 0.068).
The pooled data from all participants showed improved values post-intervention regarding the chair-stand test ( p = 0.005), one-leg stand using the left leg with the eyes closed ( p = 0.030), and tandem walking backwards ( p = 0.027) ( ). The one-leg stand on the right leg with the eyes open showed worse results post-intervention when pooling data ( p = 0.019). However, there was no change when analyzing the separate intervention groups. The analysis of the intervention groups showed post-intervention balance improvements mostly in the TPh group, such as tandem walking forwards (close to significant p = 0.068), one-leg stance with eyes closed (right leg p = 0.040 and left leg p = 0.026) and close to significant improvements on the chair-stand test ( p = 0.075). The TMMY group showed improved chair-stand test results post-intervention ( p = 0.018). No change was seen regarding handgrip strength, C7-wall distance, body weight, or height.
No change was seen in FES-I scoring neither post-intervention nor at the 1-year post-intervention follow-up ( ). There were five participants both at the baseline and post-intervention who reported physical activity of less than 150 min/week. At the 1-year post-intervention follow-up, six people reported insufficient physical activity (one person had recently suffered from a calcaneus fracture). There was no significant change in the reported total physical activity in the intervention groups between the baseline and post-intervention or between the post-intervention and 1-year post-intervention follow-up. However, a trend of more physical activity was seen in the T group ( p = 0.066). The sedentary inactive time (sitting/resting) did not change between the baseline and follow-ups.
The participants scored significantly better in the knowledge testing about osteoporosis post-intervention compared to the baseline. The median knowledge score increased from 15 at the baseline to 18 post-intervention (68% vs. 81% correct answers, p = 0.001), when analyzing the pooled data of all participants. The median score increased from 15 to 19 (68% vs. 86%, p = 0.043) in the T group, from 11 to 18 (50% vs. 81%, p = 0.011) in the TPh group, and from 17 to 20 (77% vs. 91%, p = 0.018) in the TMMY group.
The mean (SD) PEI total score post-intervention was 4.6 (2.8) (Md = 5, range 0–9). The mean (SD) values for the groups were 3.0 (2.9) for the T group, 3.6 (2.9) for the TPh group, and 5.8 (2.4) for the TMMY group. Improved patient enablement, i.e., “better” or “much better”, was reported by 40–92% of participants in each question, as shown in . The question with the highest percentage of participants reporting improved enablement was “(Q2)… able to understand your illness/disabilities” (92% of participants), and the question with the lowest improvement was “(Q5)… confident about your health” (40% of participants). The mean (SD) PEI total score at the 1-year follow-up was 4.1 (3.4) (Md = 4, range 0–11). The mean (SD) values for the groups were 3.8 (3.8) for the T group, 2.7 (2.6) for the TPh group, and 6.3 (3.1) for the TMMY group. The question with the highest percentage of participants reporting improved enablement was “(Q2)… able to understand your illness/disabilities” (76% of participants), and the question with the lowest improvement was “(Q5)… confident about your health” (33% of participants). The PEI total score did not change significantly between the post-intervention and 1-year post-intervention follow-up ( p = 0.390). The theoretical lectures were generally appreciated by the participants, scoring a mean of 4.4 (range 4.0–4.6) and a median of 5 (range 4.5–5). Similarly, the participants in the physical intervention groups scored highly regarding their satisfaction with the mindfulness/medical yoga activities (mean 4.7/Md 5), and also the physical training activities (mean 4.9/Md 5), respectively.
One participant in the TMMY group reported a rib fracture just before the start of the intervention. There was one participant in the TPh group who had a finger fracture during the group training activities. At the 1-year post-intervention follow-up, a vertebral fracture was reported in the T group, a calcaneus fracture (after falling from a ladder) was reported in the TPh group, and a rib fracture and a sacral fracture were reported in the TMMY group. At the baseline, one participant in the TPh group reported a fall in the past month, while one participant in the T group reported a fall at the end of the intervention period. No participants reported a low energy fall at the 1-year post-intervention follow-up.
The present study shows that patient education, including interdisciplinary themes, combined with supervised training is a well-accepted and safe intervention in a primary health care cohort of patients with established spinal osteoporosis. The intervention had positive effects on important health outcomes, including chronic pain, physical function, and HRQoL. Adherence to the once-weekly intervention program for ten weeks was high, with a mean attendance rate of 90%, even though several participants lived far away from where the SOL interventions took place. Most VFs are painful and cause persistent symptoms [ , , ]. In our study, one out of four participants reported the use of opioids at the baseline, and more than half of the group scored more than five on the NRS as the worst pain, thus highlighting the impact of the pain. After the intervention, regular usage of opioids as well as other painkillers decreased, and participants reported less pain (last week and worst). Physical training might decrease pain , which agrees with our finding that the TPh group showed improvements in the HRQoL pain dimensions after the intervention period. This strengthens the need for physical training as part of the education program for patients with established spinal osteoporosis . Coping with pain could also be influenced by increased knowledge of the disease and by shared experiences with other participants in a similar situation . Concerning the importance of observing and teaching the individuals within the groups, an upper limit of 8 to 10 participants is desirable . At the baseline, there were 10–11 randomized participants in each intervention group. However, the number decreased below ten because of the dropouts. A multi-component exercise program with progressive strength resistance and balance training, which was performed by the TPh group, is recommended for people with osteoporosis [ , , ]. The SOL was scheduled to be once a week for practical reasons, but 10 supervised 1-h group training sessions seemed not long enough to observe some changes, such as improved body posture. Indeed, resistance and balance training at least twice a week is recommended to achieve effects on falls and bone health . Most of the physical exercise training and the MMY training was safe. However, one finger fracture occurred in the TPh group, when exercising with a medicine ball. This specific exercise could be adjusted in the future to diminish the risk of injuries in this patient group. Previous studies on patient education in osteoporosis have shown inconsistent results on quality of life, maybe due to various interventions and study designs [ , , ]. Bergland et al. showed that a combined 3-month course of circle exercise (biweekly sessions) and a 3-h theory session had a positive effect on HRQoL in patients with vertebral fractures, which was still seen after 12 months . On the contrary, Kessenich et al. showed a minimal effect on HRQoL after an 8-week educational support group, along with weekly telephone calls, in women with established spinal osteoporosis. However, that study, which did not include any supervised group training activities (besides Tai Chi exercises introduced by a speaker), proposed a combination of rehabilitative approaches to improve the quality of life for the patients . In the present study of patient education (osteoporosis school) for established spinal osteoporosis, once-a-week sessions for 10 weeks seemed to be most effective when the theoretical lectures were followed by physical training. The importance of providing patient education to patients with osteoporosis is supported by recent guidelines from the Swedish National Board of Health and Welfare giving patient education a rather high priority . The one-leg stance with eyes closed was improved for all participants as a group but, when the subgroups were analyzed, only the TPh group had significant improvements in this test. However, there was no improvement in the one-leg stance with the eyes open. We experienced that the limit for the one-leg stance with the eyes open could be extended to 60 s, as many of the participants managed 30 s, which could make the outcome measure less sensitive. The chair-stand was improved in the TMMY group, maybe as a result of an actual improvement or a more confident patient after the supervised mindfulness/medical yoga training . Improved mental health and social function were also found in the TMMY group. To our knowledge, this is the first study using the PEI to measure patient enablement in patients with osteoporosis. The mean PEI total score with a range of 0-12 was 4.1 post-intervention, which is a similar improvement compared to another Swedish study investigating patient enablement in patients with chronic musculoskeletal pain after intervention . The power of this pilot study makes subgroup analysis uncertain, but some trends could be seen, including the additive effect when including physical training in the patients’ education. However, the pure T group was very small, with a risk of rejecting true improvements. With this perspective, it is still important to include a pure T group when designing RCTs. In addition, theory education could more easily be implemented through a digital osteoporosis school. A sample size calculation with “pain last week” as the primary outcome and a 20% reduction after an intervention show a need for at least 17 participants in each arm. In the study design, we also included a non-interventional observation period prior to the interventions. We initially intended to conduct the study with a control arm without patient education interventions, i.e., similar to how osteoporosis health care is arranged in many areas today. However, after some critical reviews from the ethical committee, we chose to let all participants have an active arm and added the preceding observation period. Analyses of these pooled data showed stable results regarding most variables strengthening the observations that the changes seen after intervention are effects of the SOL. In the observation period, the distance C7-wall was improved but did regress to the baseline values after the interventions. We believe that this was not a true change but a weakness of the test, i.e., too little precision from using a folding ruler to make the correct measurements. The general unaltered results during the observation period indicate that a non-interventional arm might be unnecessary for a future larger RCT. The informants in the TPh group, as well as participants in the other intervention groups, asked for supervised physical group training after the SOL, which was planned to begin in January 2020. Unfortunately, this training was impeded due to the lockdown caused by the COVID-19 pandemic. Study Strengths and Limitations The present SOL study is a pilot study with an RCT design, where patients were randomized to one of three different interventions, all with a common theoretical base. A limitation of our study is the small number of participants who completed both the intervention and the 1-year post-intervention follow-up ( n = 21).
The present SOL study is a pilot study with an RCT design, where patients were randomized to one of three different interventions, all with a common theoretical base. A limitation of our study is the small number of participants who completed both the intervention and the 1-year post-intervention follow-up ( n = 21).
The present pilot study showed that interdisciplinary patient group education, with or without supervised exercise or mindfulness/modified medical yoga, might improve short-term physical function and pain, and short- and long-term quality of life and patient enablement in patients with established spinal osteoporosis in primary care. More randomized controlled studies are needed with a higher number of participants to confirm our results.
|
Opportunities and challenges in early diagnosis of rheumatoid arthritis in general practice | e4b42a53-aa67-4c1b-966f-238f3dfbb8d5 | 10049595 | Family Medicine[mh] | Prompt diagnosis of rheumatoid arthritis (RA), the most common form of inflammatory arthritis, is crucial to optimise long-term patient outcomes through prevention of joint damage and disability. However, early disease can be challenging to identify in primary care, especially given that RA makes up a small proportion of the musculoskeletal conditions that account for one in seven GP appointments. Patients consult with GPs a mean of four times before being referred to rheumatology services. The non-specific nature of symptoms at the earliest stages of RA is a barrier to GPs identifying patients with newly presenting RA.
The rationale for identifying early disease is to initiate treatment using disease-modifying therapies (including biologics) in the reversible stage of the disease, referred to as the RA ‘therapeutic window of opportunity’ in the 3 months following the onset of clinical synovitis. This can significantly improve clinical outcomes and health-related quality of life, with earlier disease control reducing work-related disability. However, the discovery that circulating autoantibodies, including anticitrullinated protein antibodies (ACPA), precede the clinical onset of disease provides an opportunity to identify people with musculoskeletal symptoms who are at risk of developing RA. ACPA can be identified through an anti-cyclic citrullinated peptide (anti-CCP) test. A high positive anti-CCP result is more specific for joint pathology than rheumatoid factor, and is strongly associated with the development of RA . – The international rheumatology community has adopted the term ‘pre-RA’ to retrospectively describe a phase that an individual has progressed through once it is known that they have developed RA. It is during this period that patients may present in primary care with non-specific musculoskeletal symptoms. Secondary care models in autoantibody-positive patients have evolved to predict the early development of RA before synovitis is clinically apparent. However, the applicability of these models to primary care is unknown: non-specific musculoskeletal symptoms are common in the community, and the presence of RA-related autoantibodies (ACPA) may have important differences in natural history and prognosis when identified in those with non-specific musculoskeletal symptoms compared with disease that presents with clinical synovitis. New research from the Leeds anti-CCP cohort, analysing 6780 patients from 312 general practices throughout England, demonstrated that individuals with high anti-CCP levels and joint pain in their hands/feet (without synovitis) have an increased likelihood of developing RA, compared with those with low anti-CCP levels . Targeted anti-CCP testing in general practice could identify people at high risk of developing RA, enabling referral to rheumatology services prior to the development of synovitis to facilitate monitoring, diagnosis, and rapid initiation of treatment.
Identifying pre-RA is a ‘needle in the haystack’ in primary care due to the myriad of musculoskeletal presentations. Changing the diagnostic paradigm of RA to detection prior to the onset of classical clinical synovitis requires robust evidence regarding the appropriate selection of patients ‘at risk’ of RA in primary care, and that targeted anti-CCP testing results in overall benefit, minimises harms, and is cost-effective. Research is underway to develop criteria to identify people presenting to primary care with new-onset musculoskeletal symptoms who are likely to be anti-CCP positive. Economic modelling is also exploring the cost-effectiveness of such testing, considering the workload implications within general practice and rheumatology services, the resources needed to support interpretation of test results, and pathology costs of widespread anti-CCP testing.
Even if primary care prediction models perform adequately, evidence is required regarding the clinical- and cost-effectiveness and safety of ‘pre-RA’ intervention. The benefits of treating pre-RA may include reducing the risk of clinical outcomes associated with comorbidities, such as cardiovascular disease-related mortality in RA (relative risk 1.48 [95% confidence interval = 1.36 to 1.62]). New evidence is emerging to support an earlier therapeutic window, with disease-modifying treatments to halt the biological processes and prevent the onset of RA being tested within clinical trials. There are, however, substantial adverse effects of disease-modifying therapies, and it should not be assumed that evidence on the balance of benefits and harms found for patients with RA diagnosed following presentation with typical symptoms is generalisable to the pre-RA population. ’Pre-RA’ must, therefore, be recognised as a different entity from RA. Potential harms of a strategy that will label patients as having pre-RA must be considered, such as increased anxiety, reluctance to undertake usual levels of activity due to perceived disability, or wider social implications such as increased costs of insurance policies or restriction of occupational opportunities. The scale of such harms will depend on the extent of overdiagnosis that can be expected, that is, the proportion of individuals labelled at risk who would not have gone on to develop RA . While we understand the clinical risk factors for RA development in the at-risk population, there is still potential for a high rate of false positive anti-CCP tests and it is not yet understood how frequently we should monitor these clinical risk factors. The optimal primary and secondary care service models to monitor and support patients, and the associated workload and resource implications, also require further research. Potentially modifiable lifestyle risk factors such as raised body mass index and smoking are strongly associated with the development of RA. Our recent systematic review highlighted that individuals at risk of RA have a need for more knowledge about RA and their potentially modifiable risk factors, which in turn could support their engagement with preventive interventions. However, as yet there is no clear indication that modifying these lifestyle risk factors will prevent or delay the onset of disease. Further evidence is also needed to determine if disease-modifying therapies can prevent or delay the onset of RA. Accordingly, our team are currently recruiting those with musculoskeletal symptoms who have tested positive for anti-CCP antibodies and who are at moderate or high risk of developing RA using a risk-stratification prediction model (antibody concentration more than three times the upper limit of normal plus hands/feet tenderness, and/or ≥30 minutes early-morning stiffness) to participate in a therapeutic intervention study (48-week 2 mg daily oral dose of baricitinib) to determine if it reduces the incidence of RA.
Non-specific musculoskeletal symptoms constitute a large proportion of all consultations in primary care. Testing some of these patients using anti-CCP may provide a means to identify those at risk of RA and potentially delay or prevent its onset. Before these potential benefits can be adequately realised, further research is required to evaluate and mitigate countervailing harms and costs of such a strategy, and to understand how widespread testing can be integrated into routine primary care in a way that is acceptable to GPs and patients.
|
Eptifibatide, an Older Therapeutic Peptide with New Indications: From Clinical Pharmacology to Everyday Clinical Practice | 31412cb5-5974-4827-80df-21b92f49b87a | 10049647 | Pharmacology[mh] | Peptide-based therapies are an emerging class of medications . Following the rapid evolution of cutting-edge production, modification, and analytical technologies, peptide medication development has advanced significantly in recent years. These technological developments have helped to reduce the inherent limitations of peptides. As a result, a wide variety of natural and engineered peptides have been created, studied, and used for a range of medicinal purposes . In the last two decades, more than 25 peptide drugs have been approved for clinical use, and over 150 peptides are in active development today . Therapeutic peptides are notably utilized in the treatment of cardiovascular diseases, which are the leading cause of death and morbidity among non-communicable diseases . Several important peptide drugs have been discovered in this field, primarily targeting hypertension, vascular function, and manifestations of coronary artery disease, such as acute coronary syndromes (ACS) . In this narrative review, we focus on eptifibatide, an anti-aggregation peptide drug that targets the pathophysiological mechanism of platelet thrombus formation. First, the role of platelets in hemostasis and the function of the platelet glycoprotein IIb/IIIa (GPIIb/IIIa) receptor targeted by eptifibatide are presented. This is followed by a focused discussion on eptifibatide as a peptide drug, covering its pharmacological properties, clinical use, adverse effects, and interactions. Hemostasis is a physiological process involving the formation of a blood clot to halt bleeding. Adequate functioning of hemostasis is also responsible for ensuring blood flow . Hemostasis can be explained as a three-step process: Vasoconstriction of the damaged vessel immediately after the injury to reduce blood loss; Adhesion, activation, and aggregation of platelets that form a platelet plug (usually referred to as primary hemostasis); Coagulation of protein factors and formation of a dense fibrin mesh (usually referred to as secondary hemostasis). These processes are not necessarily consecutive steps but rather concurrent mechanisms that potentiate each other . Platelets are short-lived (8−10 days) cell fragments originating from megakaryocytes. They play an essential role in hemostasis, which is most prominent in but not limited to primary hemostasis. The latter consists of platelet adhesion that starts with endothelial dysfunction and subendothelial matrix exposure. Different intracellular, membrane, and extracellular receptors and proteins mediate slowing and adhesion at the site of the defect. This process facilitates specific membrane and intracellular platelet changes, known as platelet activation. Platelets change shape and release their cytoplasmic granules with various activation factors that further promote adhesion, activation, and aggregation. They adhere to each other and extracellular molecules, including the proteins involved in the coagulation cascade. Consequently, they form a thrombus that arrests bleeding. Nevertheless, dysfunction of hemostasis and thrombus formation in diseases can also unnecessarily occlude the vessel and cause tissue ischemia . Antiplatelet medications influence primary hemostasis through their effect on platelets . They prevent stent thrombosis, lessen arterial thrombosis following plaque degradation and rupture, and are used to treat and prevent myocardial infarction and ischemic stroke. There are numerous antiplatelet medications with various modes of action. Drugs with antiplatelet effects target the von Willebrand factor (VWF) interaction with glycoprotein Ib, glycoprotein VI (GPVI), the thrombin receptor PAR-1, the adenosine diphosphate (ADP) receptor P2Y12, cyclooxygenase, phosphodiesterase, or integrin receptor GPIIb/IIIa . In , we provide a comparison between eptifibatide and other antiplatelet drugs with similar indications . GPIIb/IIIa, also called integrin αIIbβ3, is a receptor expressed on platelets and megakaryocytes. It plays a crucial role in platelet function, being mainly involved in the processes of primary hemostasis. Biochemically, GPIIb/IIIa is an integrin molecule, a type of transmembrane glycoprotein that can convey information both into (outside-in pathway) and out of (inside-out pathway) the cell. Integrin is a protein formed by two noncovalently bound subunits (α and β) and contains an extracellular domain, a transmembrane domain, and a small cytoplasmic tail . Human platelets contain only two subgroups of integrins, namely integrins with subunits β1 and β3. However, there are more subgroups in the tissues of other mammals. Integrins in the platelets support their adhesion to the extracellular matrix molecules. GPIIb/IIIa is the primary integrin of platelets that can bind to ligands containing the arginine-glycine-aspartic acid sequence (such ligands are fibrinogen, fibrin, fibronectin, and the von Willebrand factor). As GPIIb/IIIa allows platelets to bind to its main ligand, fibrinogen, it cross-links them, enabling platelet aggregation. Loss of GPIIb/IIIa function is associated with a rare autosomal bleeding disorder, Glanzmann’s thrombasthenia, whereas its excessive activity is linked to arterial thrombosis . Signaling of GPIIb/IIIa also plays a vital role in releasing vascular endothelial growth factors from platelets and cancer progression in tumor cells . 4.1. Receptor GPIIb/IIIa Pathway As noted above, GPIIb/IIIa signal transmission is bidirectional. Accordingly, both the inside-out and outside-in pathways are discussed, as eptifibatide affects both when inhibiting GPIIb/IIIa. 4.1.1. Inside-Out Pathway The inside-out signaling pathway is initiated by the activation of G protein-coupled receptors by soluble agonists, such as ADP, epinephrine, thromboxane A2, and thrombin, or by the activation of other receptors by immobile agonists such as vWF and collagen. The inside-out pathway consists of several steps involving intracellular protein mediators such as talin and kindlin. It results in a conformational change of the extracellular domain that increases GPIIb/IIIa’s binding affinity for ligands, particularly fibrinogen . In addition to intracellular activators talin and kindlin, other proteins may be involved in the activation of inside-out activation of GPIIb/IIIa. On the other hand, several proteins, such as docking protein 1 (Dok1), tensin 1, and filamin, are hypothesized to inhibit GPIIb/IIIa activation . 4.1.2. Outside-in Pathway The other part of bidirectional signaling is the outside-in pathway, which begins with the binding of fibrinogen to the activated GPIIb/IIIa. The main results are activation of adhesion, reorganization of the cytoskeleton, spreading of cells, and irreversible aggregation of platelets, leading to thrombus growth. Many proteins participate in the outside-in pathway, including transmembrane proteins, Rho-family small GTPases, intracellular adaptor molecules, kinases, and phosphatases. Most of them are associated with Src kinase activation, and the pathway often culminates in novel protein-actin cytoskeleton linkages and activities that promote primary hemostasis . Both pathways and the effect of eptifibatide on them are presented in . As noted above, GPIIb/IIIa signal transmission is bidirectional. Accordingly, both the inside-out and outside-in pathways are discussed, as eptifibatide affects both when inhibiting GPIIb/IIIa. 4.1.1. Inside-Out Pathway The inside-out signaling pathway is initiated by the activation of G protein-coupled receptors by soluble agonists, such as ADP, epinephrine, thromboxane A2, and thrombin, or by the activation of other receptors by immobile agonists such as vWF and collagen. The inside-out pathway consists of several steps involving intracellular protein mediators such as talin and kindlin. It results in a conformational change of the extracellular domain that increases GPIIb/IIIa’s binding affinity for ligands, particularly fibrinogen . In addition to intracellular activators talin and kindlin, other proteins may be involved in the activation of inside-out activation of GPIIb/IIIa. On the other hand, several proteins, such as docking protein 1 (Dok1), tensin 1, and filamin, are hypothesized to inhibit GPIIb/IIIa activation . 4.1.2. Outside-in Pathway The other part of bidirectional signaling is the outside-in pathway, which begins with the binding of fibrinogen to the activated GPIIb/IIIa. The main results are activation of adhesion, reorganization of the cytoskeleton, spreading of cells, and irreversible aggregation of platelets, leading to thrombus growth. Many proteins participate in the outside-in pathway, including transmembrane proteins, Rho-family small GTPases, intracellular adaptor molecules, kinases, and phosphatases. Most of them are associated with Src kinase activation, and the pathway often culminates in novel protein-actin cytoskeleton linkages and activities that promote primary hemostasis . Both pathways and the effect of eptifibatide on them are presented in . The inside-out signaling pathway is initiated by the activation of G protein-coupled receptors by soluble agonists, such as ADP, epinephrine, thromboxane A2, and thrombin, or by the activation of other receptors by immobile agonists such as vWF and collagen. The inside-out pathway consists of several steps involving intracellular protein mediators such as talin and kindlin. It results in a conformational change of the extracellular domain that increases GPIIb/IIIa’s binding affinity for ligands, particularly fibrinogen . In addition to intracellular activators talin and kindlin, other proteins may be involved in the activation of inside-out activation of GPIIb/IIIa. On the other hand, several proteins, such as docking protein 1 (Dok1), tensin 1, and filamin, are hypothesized to inhibit GPIIb/IIIa activation . The other part of bidirectional signaling is the outside-in pathway, which begins with the binding of fibrinogen to the activated GPIIb/IIIa. The main results are activation of adhesion, reorganization of the cytoskeleton, spreading of cells, and irreversible aggregation of platelets, leading to thrombus growth. Many proteins participate in the outside-in pathway, including transmembrane proteins, Rho-family small GTPases, intracellular adaptor molecules, kinases, and phosphatases. Most of them are associated with Src kinase activation, and the pathway often culminates in novel protein-actin cytoskeleton linkages and activities that promote primary hemostasis . Both pathways and the effect of eptifibatide on them are presented in . Eptifibatide (also known as Integrilin, Intrifiban, SB-1, or Sch-60936; DrugBank accession number: DB00063) is a heptapeptide derived from a disintegrin protein in the rattlesnake venom. It reversibly inhibits the GPIIb/IIIa, preventing platelet aggregation and activation . 5.1. Associated Drug Classes Eptifibatide belongs functionally to the GPIIb/IIIa inhibitors (GPI) and, with respect to its structure and origin, to the family of disintegrin proteins . This section briefly presents the pharmacological basis of eptifibatide classification. 5.1.1. GPIIb/IIIa Inhibitors Several drugs affect GPIIb/IIIa, but three GPI are the most prominent, namely eptifibatide (Integrilin ® ), tirofiban (Aggrastat ® ), and abciximab (ReoPro ® ), which are all administered intravenously. In addition, some active oral peptide agents targeting GPIIb/IIIa, such as orbofiban, xemilofiban, sibrafiban, and roxifiban, have been tested. However, these have not shown encouraging outcomes and have been linked to an increase in mortality, warranting the termination of many clinical trials . 5.1.2. Snake Venom and Disintegrin Peptide Family With several derivative drugs in clinical or research use, snake venoms are an attractive natural source for drug discovery and development . They consist of a wide array of molecules, most of which are bioactive and have toxic effects on muscles, neurons, the heart, or other organ cells. Nevertheless, snake venom has been employed in medicine since ancient times, notably in traditional Chinese medicine. Moreover, in the 17th century, the Italian Felice Fontana demonstrated the effect of snake venom on human blood . Several toxins are now recognized as valuable therapeutic agents or diagnostic tools . The United States Food and Drug Administration (FDA) has already approved numerous snake venom-derived drugs, including Integrilin ® (eptifibatide), Captopril ® (enalapril), Aggrastat ® (tirofiban), Reptilase ® (batroxobin), and Exanta ® (ximelagatran). In addition, many medications are now at the preclinical and clinical stages of testing for therapeutic use . Pharmacologically, snake venom consists of various substances with numerous effects. Some of them are peptides that mainly target (in an enzymatic or non-enzymatic way) membrane receptors, enzymes, ion channels, and elements of the hemostatic system . Disintegrins are a class of small cysteine-rich peptides that target integrins and are found in different species of snake venom. They carry the KTS, MGD, RTS, VGD, KGD, WGD, or RGD amino acid motifs recognized by integrins. It is important to understand that proper cysteine bridges are essential for protein folding and conformational exposure of the binding motif, composed of three amino acids. As different motif exposure translates to different effects on different integrin protein types, these short peptides are involved in several processes. For example, they take part in the regulation of angiogenesis, platelet aggregation, apoptosis, cell migration, invasion, adhesion, and proliferation . Disintegrins may be classified into four groups : Approximately 41–51 amino acid long peptides with four cysteine bridges (echistatin and obtustatin); Approximately 70 amino acid long peptides with six cysteine bridges (barbourin, flavoviridin, and atrolysin); Approximately 84 amino acid long peptides with seven cysteine bridges (bitistatin); Macromolecular complexes of usually noncovalently bound homodimers or heterodimers, which are 67 amino acids long and have 10 cysteines incorporated into the structure. Eptifibatide is a disintegrin type of peptide that mimics a portion of barbourin, a toxic peptide found in the venom of the Southeastern pygmy rattlesnake ( Sistrurus miliarius barbouri ) . Barbourin is selective for GPIIb/IIIa, despite other disintegrins having nonspecific affinities for different integrins. This nonspecific binding of other disintegrins is mediated by the RGD motif, whereas barbourin derives its favorable selectivity from the unique KGD motif in which lysine is replaced by arginine. In eptifibatide, the KGD motif is preserved but modified . In the Protein Data Bank (PDB), there are currently three models of the GPIIb/IIIa complexes with eptifibatide (PDB ID: 7THO, 2VDN, and 7U60) . 5.2. Biochemical Structure Eptifibatide is a cyclic peptide derived from the disintegrin family protein, barbourin . Its molecular formula is C35H49N11O9S2, and its molecular mass is 832.0 g/mol . Both the structure and the amino acids sequence are presented in . It was derived by determining the minimum active sequence (MAS) of this component of snake venom. A minimum active sequence (MAS) is the shortest amino acid sequence derived from an endogenous peptide, still retaining its potency or binding affinity to its target . The process of truncation is used to remove biologically redundant amino acid residues from the protein. The endogenous peptides are “trimmed” in such a way as to reach a more economical protein size that can be widely synthesized without compromising its effect on biological targets. The cys-rich endogenous disintegrin protein from snake venom contains 73 amino acids (UniProt code: P22827), and a length of seven amino acids was achieved for eptifibatide by the process of truncation . The small size of the peptide explains its low immunogenicity, which is an essential factor in repetitive administration and the use of the drug on patients with an unknown history . Moreover, in developing eptifibatide, researchers had to overcome the drawbacks of peptide drugs, especially their low in vivo stability and membrane impermeability . Since the direct action of eptifibatide is thought to be limited only to the extracellular domain of GPIIb/IIIa on platelets, the drawback of membrane impermeability was eliminated. To increase the stability of eptifibatide, the peptide was cyclized using a disulfide bridge between the captopropionyl residue (des-amino-cysteinyl) and the cysteine. The cyclic structure increases the bioavailability of the drug and its resistance to plasma proteases . Besides these modifications, the peptide undergoes guanylation at the Lys side chain and deamination at the N-terminus (among other modifications), giving it a highly potent therapeutic value . 5.3. Pharmacodynamics Eptifibatide competes in a dose-dependent manner with fibrinogen for the GPIIb/IIIa. It is a specific inhibitor of the GPIIb/IIIa receptor, which limits the pharmacological effect of platelets and their precursors . The treatment objective is to achieve 80% inhibition of platelet aggregation depending on the dose and concentration of medication. This proposition has been demonstrated ex vivo with adenosine diphosphate (ADP) and other agonists that induce platelet aggregation. The immediate effect of eptifibatide can be observed after an intravenous injection; when a continuous infusion is subsequently administered, this treatment can successfully inhibit more than 80% of ADP-induced platelet aggregation ex vivo with normal calcium levels in the majority of patients . Furthermore, the eptifibatide effect can be quickly stopped, since the drug rapidly dissociates from GPIIb/IIIa, and after 4 h, platelet functions return to baseline and are swiftly cleared from plasma . 5.4. Pharmacokinetics Intravenous administration of therapeutic peptides has the advantage of avoiding pre-systemic metabolism by the liver and gastrointestinal enzymes, resulting in complete systemic availability. For bolus doses of 90 to 250 µg/kg and infusion rates of 0.5 to 3.0 µg/min, the pharmacokinetics of eptifibatide are linear and proportional to the dose. When infused at 2.0 µg/kg/min, the mean steady-state plasma concentration of eptifibatide in patients with coronary artery disease is 1.5 to 2.2 µg/mL. Plasma concentrations in this range can be attained rapidly if a bolus of 180 µg/kg is used prior to the infusion . The onset of action is rapid, with inhibition of platelet aggregation occurring 15 min after a bolus. The binding proportion of eptifibatide to human plasma proteins is approximately 25% . The pharmacokinetics of peptides are characterized by their typically short half-life in the bloodstream, which results from cleavage by proteases and peptidases. A short elimination half-life for endogenous peptides is desirable for regulating their concentrations and function. The eptifibatide plasma elimination half-life is 2.5 h . Peptides have a molecular weight between 1 and 10 kDa; therefore, the primary absorption process is diffusion-driven uptake into blood. On the other hand, eptifibatide has a molecular weight of 800 D. The volume of distribution for eptifibatide is 0.2 to 0.3 L/kg. Eptifibatide is not known to be metabolized by uridine-5-diphosphate glucuronosyltransferase enzymes or cytochrome P450 (CYP) but is deaminated by metabolic enzymes. Furthermore, kidney clearance accounts for approximately 50% of total body clearance; therefore, deaminated eptifibatide and polar metabolites are excreted in the urine . Hepatic metabolism is not the primary route of elimination for most peptides, but it can play an essential role in the metabolism of some peptide drugs . 5.5. Clinical Applications Eptifibatide is an antiplatelet agent; therefore, it is used in diseases in which thrombus formation is a critical part of pathogenesis or complications. As with any antithrombotic treatment, consideration should be given to the trade-off between the risk of ischemic injury and bleeding when administering eptifibatide . The FDA indicates its use for the treatment of ACS and in percutaneous coronary intervention (PCI) . However, research over the past decade has also sought to evaluate the role of eptifibatide in ischemic stroke, stenting of carotid and intracranial aneurysms, and septic shock . In we present summary of recently published meta-analyses. 5.5.1. Acute Coronary Syndromes: Angina Pectoris, STEMI, and NSTEMI ACS is a manifestation of coronary heart disease associated with an abrupt reduction in the blood supply to the heart. Underlying factors contributing to the disease are smoking, hyperlipidemia, obesity, diabetes, etc. The syndromes comprise different clinical presentations, including unstable angina pectoris, non-ST elevation myocardial infarction (NSTEMI), and ST-elevation myocardial infarction (STEMI). Despite different presentations, all syndromes usually present with chest discomfort at rest . In most cases, the pathophysiological basis of these diseases is the rupture of an atherosclerotic plaque in a cardiac vessel causing platelet aggregation and thrombus formation, which in turn restricts the blood flow to the heart tissue, resulting in cardiac ischemia . Angina Pectoris and non-ST Elevation Myocardial Infarction Eptifibatide is indicated for the prevention of myocardial infarction in unstable angina and NSTEMI in both drug-treated and PCI patients . The PURSUIT trial (1998) showed a reduction in endpoint mortality and a beneficial effect in preventing nonfatal myocardial infarction in these patients . In NSTEMI, a combination of loading dose by aspirin and maintenance treatment by eptifibatide could be used . Aspirin inhibits thromboxane A2 production and therefore prevents platelet aggregation. Although both drugs affect the platelets, their mechanisms of work are different, potentiating their effect. ST-Elevated Myocardial Infarction A meta-analysis by Karathanos et al. in 2019 showed that routine use of GPIs in patients with STEMI was associated with reduced mortality, which was probably the consequence of a reduction in recurrent ischemic events. Despite the promising result, it should be noted that these results are largely based on studies before dual antiplatelet therapy with prasugrel/ticagrelor was routinely used, as is the case today . Although less convincing, eptifibatide can also improve myocardial perfusion in STEMI, as shown in the TITAN-TIMI 34 trial . Nevertheless, more recent studies have shown that prehospital administration of GPI in STEMI has not shown benefits and even increases the bleeding risk compared to routine use in a catheterization laboratory . Although eptifibatide has not been tested in a randomized trial, the European Society of Cardiology guidelines (ESC) suggest it as bail-out therapy in high-risk patients (slow flow or no flow with occlusion of the stent, high thrombus burden, etc.) but not as a routine drug for primary PCI . A meta-analysis by Saleiro et al. in 2020 has shown that the use of GPIs as an adjunct to standard therapy may be beneficial in myocardial infarction that results in cardiogenic shock. In this study, the use of GPIs was associated with better outcomes, namely short-term and long-term survival. Moreover, it did not increase the risk of bleeding in the treated patients . Other newer studies have shown similar results . 5.5.2. Percutaneous Coronary Intervention PCI is a non-surgical but invasive procedure in which a catheter is used to insert a stent into narrowed or occluded coronary arteries, improving the blood supply. It is the preferred method of treatment for ACS . As the pretreatment in patients undergoing PCI, the ESC guidelines propose using a combination of eptifibatide and unfractionated heparin, as both anticoagulation and platelet inhibition are important in the pathogenesis-based therapy of NSTEMI . Guidelines from the American College of Cardiology (ACC) and the American Heart Association (AHA) similarly suggest the use of eptifibatide as an initial antiplatelet therapy in patients with high-risk features . Heparin dosages of 50–70 IU/kg i.v. should be used if administered in this combination . Although the FDA has approved using eptifibatide for patients undergoing PCI (including stenting), data for using eptifibatide in the periinterventional treatment of NSTEMI are limited and partially outdated. A major trial that has shown the benefits of eptifibatide use in PCI was the IMPACT-II trial, published in 1997 . Most research on the use of eptifibatide in PCI pretreatment predates routine dual antiplatelet treatment (DAPT) . In periinterventional antiplatelet treatment, oral P2Y12 receptor inhibitors were found to be as effective as GPI and are recommended for routine use . Therefore, little to no evidence exists to support the use of eptifibatide in patients who will undergo coronary angiography and are receiving DAPT . On the other hand, eptifibatide may be considered when facing high-risk PCI patients (slow flow or no-flow with the closure of the stent, high thrombus burden, etc.), patients who did not receive pretreatment with P2Y12 receptor inhibitors, and patients with thrombotic complications . The intracoronary administration of the drug is comparable to intravenous use . 5.5.3. Bridging Strategy for Patients Undergoing Surgery after Coronary Stent Insertion Postoperative bleeding prevention after cardiac surgery is crucial to decreasing morbidity and mortality. Since i.v. antiplatelet medications, such as eptifibatide, are quickly cleared from the system and their antiplatelet effect can be quickly reversed, they are used before cardiac and noncardiac surgery as a substitution for oral P2Y12 inhibitors . A 2022 meta-analysis by Wu et al. showed that eptifibatide might be safe and effective when used as a bridging strategy for patients undergoing coronary stent implantation requiring surgery. The GPI might be used without an increased bleeding risk when temporarily discontinuing DAPT. Nevertheless, further randomized studies are needed to substantiate this claim . In a 2019 study by Van Tuyl et al., eptifibatide was shown to be an effective choice for these patients and was even preferred over abciximab . Another drug that is often compared to eptifibatide in bridging strategies is cangrelor. Cangrelor is a reversible P2Y12 receptor inhibitor that prevents ADP-induced platelet aggregation and activation. It should be considered in patients with renal insufficiency, as clearance of eptifibatide is influenced by renal function . Moreover, a study by Yun et al. from 2019 showed that cangrelor and eptifibatide were similar in terms of overall bleeding events and major inpatient cardiac adverse events . 5.5.4. Ischemic Stroke and Carotid and Intracranial Aneurysm Stenting The CLEAR trial from 2008 showed that eptifibatide is beneficial in preventing intracerebral hemorrhage in patients with acute ischemic stroke if administered with TPA . Nevertheless, a 2022 meta-analysis by Liu et al. showed that adding eptifibatide to routine thrombolysis or thrombectomy treatment did not improve functional outcomes, favorable outcomes, or the National Institutes of Health Stroke Scale (NIHSS) score. Moreover, it might be associated with an increase in fatal ICH three months after AIS . Another meta-analysis by Zhu et al. in 2020 showed that eptifibatide might be promising when used at a reduced dose (a low dose was also used in the CLEAR study); thus, more randomized trials with different doses are needed to evaluate the role of eptifibatide in the treatment of acute ischemic stroke . A newer retrospective case-control study by Luo et al. (2022) compared routine therapy and treatment with an additional low dose of eptifibatide. Although the study reported no significant differences in NIHSS or adverse events, an analysis of the subgroups showed that eptifibatide is a safe and effective treatment when small artery occlusion is involved . Similarly, a trial by Rana et al. from 2022 showed that using eptifibatide during endovascular therapy in large vessel occlusion is associated with a higher rate of hemorrhages and no benefits to the NIHSS or 90-day mortality . On the other hand, a matched-control analysis by Ma et al. in 2022 showed that the use of eptifibatide was safe and effective in patients undergoing mechanical thrombectomy after ischemic stroke, because the rate of successful recanalization was significantly higher in the intervention group (91.3% versus 81.5%; p = 0.043) and the 3-month outcome on the modified Rankin Scale showed good results (53.1% versus 33.3%; p = 0.016) . As shown, the existing evidence in this research area is not yet conclusive, and further studies are needed to evaluate the use of eptifibatide in the treatment of ischemic stroke. Antiplatelet agents are administered to prevent one of the most critical complications of stenting, namely stent thrombosis. In a study by Osteraas et al., the use of eptifibatide (as a bolus followed by infusion for 24 h after stent placement) was shown to be associated with a lower risk of symptomatic intracranial hemorrhage after carotid stenting . Another study by Horev et al. from 2021 similarly reported a reduced number of complications when using eptifibatide immediately after carotid stenting . Eptifibatide has also been explored as a potential antiplatelet therapy in the stenting of intracranial aneurysms. A 2022 study by Aouni et al. compared three antiplatelet agents (ticagrelor, eptifibatide, and cangrelor) in the stent-assisted endovascular treatment of unruptured intracranial aneurysms and found no significant differences between them . 5.5.5. Septic Shock One of the main mechanisms in septic shock is the activation of the endothelium and platelets. This activation subsequently leads to generalized microvascular damage, microthrombi, capillary leaks, and coagulopathy caused by widespread consumption of coagulation factors . It is noteworthy that research on the use of eptifibatide in septic shock is extremely limited, as to our knowledge, only one study has been performed. The 2019 randomized and placebo-controlled double-blind trial by Berthelsen et al. showed the benefits in improving Sequential Organ Failure Assessment (SOFA) and reducing platelet consumption, fibrinolytic biomarkers, and endothelial damage when a combination of the synthetic analogs of prostacyclin, iloprost, and eptifibatide was used in patients with septic shock . Further research in this area is needed to elucidate the role of eptifibatide. 5.6. Contraindications Particular attention should be paid to patients who are hypersensitive to the active ingredient or any of the additives, as well as patients on any other parenteral GPI. Another group of contraindications that are important in the use of eptifibatide includes a history of bleeding diathesis within 30 days, current ongoing internal bleeding, and recent significant bleeding within the last six months, either in the gastrointestinal or genitourinary tract . Furthermore, eptifibatide is not recommended in hemodialysis patients due to an increased risk of bleeding. An alternative choice for treatment is abciximab, a mouse-human monoclonal antibody based on a murine analog . Similarly, severe, uncontrolled hypertension with a systolic blood pressure of >200 mmHg and a diastolic blood pressure of >110 mmHg in patients receiving antihypertensive treatment and a history of hemorrhagic stroke are also contraindications for the use of eptifibatide . Other conditions where eptifibatide is contraindicated include thrombocytopenia with a platelet count of <100,000 cells/mm 3 and a prothrombin time > >1.2 × higher from control values or international normalized ratio (INR) ≥ 2.0 . Thrombocytopenia is frequently observed in individuals with hepatic dysfunction; this is thought to be related to splenic sequestration due to portal hypertension or decreased production of thrombopoietin by damaged liver cells. It is also known that bone marrow suppression or folic acid deficiency may occur in people with alcoholic cirrhosis. Accordingly, eptifibatide is contraindicated in these categories of patients . We must be mindful of two other important conditions when initiating eptifibatide, namely recent major surgery or trauma within the past six weeks and a history of intracranial neoplasms, arteriovenous malformations, or aortic dissections. In addition, eptifibatide is a category B drug during pregnancy; however, it should be used in ACS because the benefits outweigh the risks. Nevertheless, special caution is needed in lactating mothers. Caution is also advised for women, the elderly, and people with a low body weight . 5.7. Administration and Dosages Eptifibatide is an effective periinterventional antiplatelet agent manufactured as a sterile solution in 10 mL single-dose vials containing 20 mg of eptifibatide and in 100 mL single-dose vials containing 75 mg of eptifibatide. It can be administered via IV and must be protected from direct light prior to administration . Before administration, baseline blood tests, such as complete blood count, prothrombin time (PT)/activated partial thromboplastin time (aPTT), serum creatinine, and activated clotting time, should be performed in patients undergoing PCI. During administration, diligent assessment and monitoring for possible arrhythmias should be performed. The dose regimen for NSTEMI is a double bolus of 180 μg/kg i.v. in 10 min intervals, followed by an infusion of 2.0 μg/kg/min for up to 18 h in patients with normal kidney function. Patients with a creatinine clearance of less than 50 mL/min should have a 50% reduction in their eptifibatide dose. In elderly patients, the hemostatic balance is shifted towards an increased tendency for clotting and reduced fibrinolysis. However, several other contributing factors exist in the elderly that can lead to an increased risk of bleeding. These include distinct pharmacokinetic and pharmacodynamic responses, polypharmacy, and increased comorbidities, all of which can interact with one another and increase the risk of bleeding. Nevertheless, while caution should be exercised in the elderly, advanced age is not a contraindication for the use of GPI . Eptifibatide may be administered concurrently with heparin and aspirin, as it was used in all clinical studies, and the dosage need not be altered. It is known that the concentration of eptifibatide in plasma increased rapidly and remained at a constant level of about 1640 ng/mL from half an hour to 24 h. After that, the concentration of the drug dropped rapidly in multiple stages . 5.8. Adverse Effects and Interactions 5.8.1. Adverse Effects Eptifibatide may cause a rare and significant adverse effect known as eptifibatide-induced thrombocytopenia (EIT), characterized by an acute, unexpected decrease in platelet count to less than <30,000 platelets/µL. However, patients must be exposed to eptifibatide for 5–7 days before they develop sensitization when receiving it for the first time. EIT was reported in several case studies and occurred in approximately 1 to 2% of patients . It is essential to monitor platelets frequently after starting treatment. The underlying causes of EIT are not fully understood; however, research indicates that it may involve the formation of IgG antibodies that target GPI only when the implicated drug is present. IgG can cause platelet aggregation and secretion, along with the activation of the tyrosine kinase cascade . Enzyme-linked immunosorbent assay testing using monoclonal antibodies can identify platelet glycoprotein targets, GPIIb/IIIa; however, such testing is not widely accessible . Treatment options for EIT include stopping the infusion, platelet transfusions (which may be inefficient because of the half-life of GPI), and fresh frozen plasma (used in major bleeding). Other treatment options include the use of corticosteroids (which are not effective) and intravenous immunoglobulin G (IVIG) . Bleeding is a more common and potentially serious adverse effect. Masood et al. showed that out of 28 reported complications of EIT, 57% involved bleeding events. These included 14% catheterization site bleeding, 14% petechial hemorrhage, 7% groin hematoma, 7% infusion site bleeding, 3.6% gastrointestinal bleeding, 3.6% hemoptysis, 3.6% epistaxis, and 3.6% hematuria . Diffuse alveolar hemorrhage (DAH) is another important but often underappreciated form of eptifibatide-related bleeding. The incidence rate of eptifibatide-induced DAH is 0.5% . In addition to the bleeding and thrombocytopenia mentioned above, side effects on the heart should not be forgotten. The PURSUIT study reported various arrhythmias, such as ventricular tachycardia, atrial fibrillation, ventricular fibrillation, and atrioventricular block, in addition to congestive heart failure and cardiac arrest. However, the side effects were attributed to the underlying disease . As for the other side effects, it is worth mentioning thrombosis, which occurred in 10.75% of patients and included deep vein thrombosis, pulmonary embolism, and in-stent thrombosis. However, the exact relationship between eptifibatide and thrombosis is unclear. The following side effect also occurred in some proportion, namely allergic reactions such as rigor, chills, angioedema, and hypotension. Positive eptifibatide antibodies were also found in 87.5% of cases . 5.8.2. Interactions There are 140 drugs known to interact with eptifibatide, of which 55 had major interactions. Considering the mechanism of action of eptifibatide, caution should be exercised with drugs inhibiting platelet aggregation. These drugs include clopidogrel, ticlopidine, thrombolytics, oral anticoagulants, adenosine, prostacyclin, sulfinpyrazone, nonsteroidal anti-inflammatory drugs (NSAIDs), dipyridamole, and dextran solution. On the other hand, if an oral anticoagulant such as warfarin was taken concomitantly, no further bleeding was observed . Eptifibatide belongs functionally to the GPIIb/IIIa inhibitors (GPI) and, with respect to its structure and origin, to the family of disintegrin proteins . This section briefly presents the pharmacological basis of eptifibatide classification. 5.1.1. GPIIb/IIIa Inhibitors Several drugs affect GPIIb/IIIa, but three GPI are the most prominent, namely eptifibatide (Integrilin ® ), tirofiban (Aggrastat ® ), and abciximab (ReoPro ® ), which are all administered intravenously. In addition, some active oral peptide agents targeting GPIIb/IIIa, such as orbofiban, xemilofiban, sibrafiban, and roxifiban, have been tested. However, these have not shown encouraging outcomes and have been linked to an increase in mortality, warranting the termination of many clinical trials . 5.1.2. Snake Venom and Disintegrin Peptide Family With several derivative drugs in clinical or research use, snake venoms are an attractive natural source for drug discovery and development . They consist of a wide array of molecules, most of which are bioactive and have toxic effects on muscles, neurons, the heart, or other organ cells. Nevertheless, snake venom has been employed in medicine since ancient times, notably in traditional Chinese medicine. Moreover, in the 17th century, the Italian Felice Fontana demonstrated the effect of snake venom on human blood . Several toxins are now recognized as valuable therapeutic agents or diagnostic tools . The United States Food and Drug Administration (FDA) has already approved numerous snake venom-derived drugs, including Integrilin ® (eptifibatide), Captopril ® (enalapril), Aggrastat ® (tirofiban), Reptilase ® (batroxobin), and Exanta ® (ximelagatran). In addition, many medications are now at the preclinical and clinical stages of testing for therapeutic use . Pharmacologically, snake venom consists of various substances with numerous effects. Some of them are peptides that mainly target (in an enzymatic or non-enzymatic way) membrane receptors, enzymes, ion channels, and elements of the hemostatic system . Disintegrins are a class of small cysteine-rich peptides that target integrins and are found in different species of snake venom. They carry the KTS, MGD, RTS, VGD, KGD, WGD, or RGD amino acid motifs recognized by integrins. It is important to understand that proper cysteine bridges are essential for protein folding and conformational exposure of the binding motif, composed of three amino acids. As different motif exposure translates to different effects on different integrin protein types, these short peptides are involved in several processes. For example, they take part in the regulation of angiogenesis, platelet aggregation, apoptosis, cell migration, invasion, adhesion, and proliferation . Disintegrins may be classified into four groups : Approximately 41–51 amino acid long peptides with four cysteine bridges (echistatin and obtustatin); Approximately 70 amino acid long peptides with six cysteine bridges (barbourin, flavoviridin, and atrolysin); Approximately 84 amino acid long peptides with seven cysteine bridges (bitistatin); Macromolecular complexes of usually noncovalently bound homodimers or heterodimers, which are 67 amino acids long and have 10 cysteines incorporated into the structure. Eptifibatide is a disintegrin type of peptide that mimics a portion of barbourin, a toxic peptide found in the venom of the Southeastern pygmy rattlesnake ( Sistrurus miliarius barbouri ) . Barbourin is selective for GPIIb/IIIa, despite other disintegrins having nonspecific affinities for different integrins. This nonspecific binding of other disintegrins is mediated by the RGD motif, whereas barbourin derives its favorable selectivity from the unique KGD motif in which lysine is replaced by arginine. In eptifibatide, the KGD motif is preserved but modified . In the Protein Data Bank (PDB), there are currently three models of the GPIIb/IIIa complexes with eptifibatide (PDB ID: 7THO, 2VDN, and 7U60) . Several drugs affect GPIIb/IIIa, but three GPI are the most prominent, namely eptifibatide (Integrilin ® ), tirofiban (Aggrastat ® ), and abciximab (ReoPro ® ), which are all administered intravenously. In addition, some active oral peptide agents targeting GPIIb/IIIa, such as orbofiban, xemilofiban, sibrafiban, and roxifiban, have been tested. However, these have not shown encouraging outcomes and have been linked to an increase in mortality, warranting the termination of many clinical trials . With several derivative drugs in clinical or research use, snake venoms are an attractive natural source for drug discovery and development . They consist of a wide array of molecules, most of which are bioactive and have toxic effects on muscles, neurons, the heart, or other organ cells. Nevertheless, snake venom has been employed in medicine since ancient times, notably in traditional Chinese medicine. Moreover, in the 17th century, the Italian Felice Fontana demonstrated the effect of snake venom on human blood . Several toxins are now recognized as valuable therapeutic agents or diagnostic tools . The United States Food and Drug Administration (FDA) has already approved numerous snake venom-derived drugs, including Integrilin ® (eptifibatide), Captopril ® (enalapril), Aggrastat ® (tirofiban), Reptilase ® (batroxobin), and Exanta ® (ximelagatran). In addition, many medications are now at the preclinical and clinical stages of testing for therapeutic use . Pharmacologically, snake venom consists of various substances with numerous effects. Some of them are peptides that mainly target (in an enzymatic or non-enzymatic way) membrane receptors, enzymes, ion channels, and elements of the hemostatic system . Disintegrins are a class of small cysteine-rich peptides that target integrins and are found in different species of snake venom. They carry the KTS, MGD, RTS, VGD, KGD, WGD, or RGD amino acid motifs recognized by integrins. It is important to understand that proper cysteine bridges are essential for protein folding and conformational exposure of the binding motif, composed of three amino acids. As different motif exposure translates to different effects on different integrin protein types, these short peptides are involved in several processes. For example, they take part in the regulation of angiogenesis, platelet aggregation, apoptosis, cell migration, invasion, adhesion, and proliferation . Disintegrins may be classified into four groups : Approximately 41–51 amino acid long peptides with four cysteine bridges (echistatin and obtustatin); Approximately 70 amino acid long peptides with six cysteine bridges (barbourin, flavoviridin, and atrolysin); Approximately 84 amino acid long peptides with seven cysteine bridges (bitistatin); Macromolecular complexes of usually noncovalently bound homodimers or heterodimers, which are 67 amino acids long and have 10 cysteines incorporated into the structure. Eptifibatide is a disintegrin type of peptide that mimics a portion of barbourin, a toxic peptide found in the venom of the Southeastern pygmy rattlesnake ( Sistrurus miliarius barbouri ) . Barbourin is selective for GPIIb/IIIa, despite other disintegrins having nonspecific affinities for different integrins. This nonspecific binding of other disintegrins is mediated by the RGD motif, whereas barbourin derives its favorable selectivity from the unique KGD motif in which lysine is replaced by arginine. In eptifibatide, the KGD motif is preserved but modified . In the Protein Data Bank (PDB), there are currently three models of the GPIIb/IIIa complexes with eptifibatide (PDB ID: 7THO, 2VDN, and 7U60) . Eptifibatide is a cyclic peptide derived from the disintegrin family protein, barbourin . Its molecular formula is C35H49N11O9S2, and its molecular mass is 832.0 g/mol . Both the structure and the amino acids sequence are presented in . It was derived by determining the minimum active sequence (MAS) of this component of snake venom. A minimum active sequence (MAS) is the shortest amino acid sequence derived from an endogenous peptide, still retaining its potency or binding affinity to its target . The process of truncation is used to remove biologically redundant amino acid residues from the protein. The endogenous peptides are “trimmed” in such a way as to reach a more economical protein size that can be widely synthesized without compromising its effect on biological targets. The cys-rich endogenous disintegrin protein from snake venom contains 73 amino acids (UniProt code: P22827), and a length of seven amino acids was achieved for eptifibatide by the process of truncation . The small size of the peptide explains its low immunogenicity, which is an essential factor in repetitive administration and the use of the drug on patients with an unknown history . Moreover, in developing eptifibatide, researchers had to overcome the drawbacks of peptide drugs, especially their low in vivo stability and membrane impermeability . Since the direct action of eptifibatide is thought to be limited only to the extracellular domain of GPIIb/IIIa on platelets, the drawback of membrane impermeability was eliminated. To increase the stability of eptifibatide, the peptide was cyclized using a disulfide bridge between the captopropionyl residue (des-amino-cysteinyl) and the cysteine. The cyclic structure increases the bioavailability of the drug and its resistance to plasma proteases . Besides these modifications, the peptide undergoes guanylation at the Lys side chain and deamination at the N-terminus (among other modifications), giving it a highly potent therapeutic value . Eptifibatide competes in a dose-dependent manner with fibrinogen for the GPIIb/IIIa. It is a specific inhibitor of the GPIIb/IIIa receptor, which limits the pharmacological effect of platelets and their precursors . The treatment objective is to achieve 80% inhibition of platelet aggregation depending on the dose and concentration of medication. This proposition has been demonstrated ex vivo with adenosine diphosphate (ADP) and other agonists that induce platelet aggregation. The immediate effect of eptifibatide can be observed after an intravenous injection; when a continuous infusion is subsequently administered, this treatment can successfully inhibit more than 80% of ADP-induced platelet aggregation ex vivo with normal calcium levels in the majority of patients . Furthermore, the eptifibatide effect can be quickly stopped, since the drug rapidly dissociates from GPIIb/IIIa, and after 4 h, platelet functions return to baseline and are swiftly cleared from plasma . Intravenous administration of therapeutic peptides has the advantage of avoiding pre-systemic metabolism by the liver and gastrointestinal enzymes, resulting in complete systemic availability. For bolus doses of 90 to 250 µg/kg and infusion rates of 0.5 to 3.0 µg/min, the pharmacokinetics of eptifibatide are linear and proportional to the dose. When infused at 2.0 µg/kg/min, the mean steady-state plasma concentration of eptifibatide in patients with coronary artery disease is 1.5 to 2.2 µg/mL. Plasma concentrations in this range can be attained rapidly if a bolus of 180 µg/kg is used prior to the infusion . The onset of action is rapid, with inhibition of platelet aggregation occurring 15 min after a bolus. The binding proportion of eptifibatide to human plasma proteins is approximately 25% . The pharmacokinetics of peptides are characterized by their typically short half-life in the bloodstream, which results from cleavage by proteases and peptidases. A short elimination half-life for endogenous peptides is desirable for regulating their concentrations and function. The eptifibatide plasma elimination half-life is 2.5 h . Peptides have a molecular weight between 1 and 10 kDa; therefore, the primary absorption process is diffusion-driven uptake into blood. On the other hand, eptifibatide has a molecular weight of 800 D. The volume of distribution for eptifibatide is 0.2 to 0.3 L/kg. Eptifibatide is not known to be metabolized by uridine-5-diphosphate glucuronosyltransferase enzymes or cytochrome P450 (CYP) but is deaminated by metabolic enzymes. Furthermore, kidney clearance accounts for approximately 50% of total body clearance; therefore, deaminated eptifibatide and polar metabolites are excreted in the urine . Hepatic metabolism is not the primary route of elimination for most peptides, but it can play an essential role in the metabolism of some peptide drugs . Eptifibatide is an antiplatelet agent; therefore, it is used in diseases in which thrombus formation is a critical part of pathogenesis or complications. As with any antithrombotic treatment, consideration should be given to the trade-off between the risk of ischemic injury and bleeding when administering eptifibatide . The FDA indicates its use for the treatment of ACS and in percutaneous coronary intervention (PCI) . However, research over the past decade has also sought to evaluate the role of eptifibatide in ischemic stroke, stenting of carotid and intracranial aneurysms, and septic shock . In we present summary of recently published meta-analyses. 5.5.1. Acute Coronary Syndromes: Angina Pectoris, STEMI, and NSTEMI ACS is a manifestation of coronary heart disease associated with an abrupt reduction in the blood supply to the heart. Underlying factors contributing to the disease are smoking, hyperlipidemia, obesity, diabetes, etc. The syndromes comprise different clinical presentations, including unstable angina pectoris, non-ST elevation myocardial infarction (NSTEMI), and ST-elevation myocardial infarction (STEMI). Despite different presentations, all syndromes usually present with chest discomfort at rest . In most cases, the pathophysiological basis of these diseases is the rupture of an atherosclerotic plaque in a cardiac vessel causing platelet aggregation and thrombus formation, which in turn restricts the blood flow to the heart tissue, resulting in cardiac ischemia . Angina Pectoris and non-ST Elevation Myocardial Infarction Eptifibatide is indicated for the prevention of myocardial infarction in unstable angina and NSTEMI in both drug-treated and PCI patients . The PURSUIT trial (1998) showed a reduction in endpoint mortality and a beneficial effect in preventing nonfatal myocardial infarction in these patients . In NSTEMI, a combination of loading dose by aspirin and maintenance treatment by eptifibatide could be used . Aspirin inhibits thromboxane A2 production and therefore prevents platelet aggregation. Although both drugs affect the platelets, their mechanisms of work are different, potentiating their effect. ST-Elevated Myocardial Infarction A meta-analysis by Karathanos et al. in 2019 showed that routine use of GPIs in patients with STEMI was associated with reduced mortality, which was probably the consequence of a reduction in recurrent ischemic events. Despite the promising result, it should be noted that these results are largely based on studies before dual antiplatelet therapy with prasugrel/ticagrelor was routinely used, as is the case today . Although less convincing, eptifibatide can also improve myocardial perfusion in STEMI, as shown in the TITAN-TIMI 34 trial . Nevertheless, more recent studies have shown that prehospital administration of GPI in STEMI has not shown benefits and even increases the bleeding risk compared to routine use in a catheterization laboratory . Although eptifibatide has not been tested in a randomized trial, the European Society of Cardiology guidelines (ESC) suggest it as bail-out therapy in high-risk patients (slow flow or no flow with occlusion of the stent, high thrombus burden, etc.) but not as a routine drug for primary PCI . A meta-analysis by Saleiro et al. in 2020 has shown that the use of GPIs as an adjunct to standard therapy may be beneficial in myocardial infarction that results in cardiogenic shock. In this study, the use of GPIs was associated with better outcomes, namely short-term and long-term survival. Moreover, it did not increase the risk of bleeding in the treated patients . Other newer studies have shown similar results . 5.5.2. Percutaneous Coronary Intervention PCI is a non-surgical but invasive procedure in which a catheter is used to insert a stent into narrowed or occluded coronary arteries, improving the blood supply. It is the preferred method of treatment for ACS . As the pretreatment in patients undergoing PCI, the ESC guidelines propose using a combination of eptifibatide and unfractionated heparin, as both anticoagulation and platelet inhibition are important in the pathogenesis-based therapy of NSTEMI . Guidelines from the American College of Cardiology (ACC) and the American Heart Association (AHA) similarly suggest the use of eptifibatide as an initial antiplatelet therapy in patients with high-risk features . Heparin dosages of 50–70 IU/kg i.v. should be used if administered in this combination . Although the FDA has approved using eptifibatide for patients undergoing PCI (including stenting), data for using eptifibatide in the periinterventional treatment of NSTEMI are limited and partially outdated. A major trial that has shown the benefits of eptifibatide use in PCI was the IMPACT-II trial, published in 1997 . Most research on the use of eptifibatide in PCI pretreatment predates routine dual antiplatelet treatment (DAPT) . In periinterventional antiplatelet treatment, oral P2Y12 receptor inhibitors were found to be as effective as GPI and are recommended for routine use . Therefore, little to no evidence exists to support the use of eptifibatide in patients who will undergo coronary angiography and are receiving DAPT . On the other hand, eptifibatide may be considered when facing high-risk PCI patients (slow flow or no-flow with the closure of the stent, high thrombus burden, etc.), patients who did not receive pretreatment with P2Y12 receptor inhibitors, and patients with thrombotic complications . The intracoronary administration of the drug is comparable to intravenous use . 5.5.3. Bridging Strategy for Patients Undergoing Surgery after Coronary Stent Insertion Postoperative bleeding prevention after cardiac surgery is crucial to decreasing morbidity and mortality. Since i.v. antiplatelet medications, such as eptifibatide, are quickly cleared from the system and their antiplatelet effect can be quickly reversed, they are used before cardiac and noncardiac surgery as a substitution for oral P2Y12 inhibitors . A 2022 meta-analysis by Wu et al. showed that eptifibatide might be safe and effective when used as a bridging strategy for patients undergoing coronary stent implantation requiring surgery. The GPI might be used without an increased bleeding risk when temporarily discontinuing DAPT. Nevertheless, further randomized studies are needed to substantiate this claim . In a 2019 study by Van Tuyl et al., eptifibatide was shown to be an effective choice for these patients and was even preferred over abciximab . Another drug that is often compared to eptifibatide in bridging strategies is cangrelor. Cangrelor is a reversible P2Y12 receptor inhibitor that prevents ADP-induced platelet aggregation and activation. It should be considered in patients with renal insufficiency, as clearance of eptifibatide is influenced by renal function . Moreover, a study by Yun et al. from 2019 showed that cangrelor and eptifibatide were similar in terms of overall bleeding events and major inpatient cardiac adverse events . 5.5.4. Ischemic Stroke and Carotid and Intracranial Aneurysm Stenting The CLEAR trial from 2008 showed that eptifibatide is beneficial in preventing intracerebral hemorrhage in patients with acute ischemic stroke if administered with TPA . Nevertheless, a 2022 meta-analysis by Liu et al. showed that adding eptifibatide to routine thrombolysis or thrombectomy treatment did not improve functional outcomes, favorable outcomes, or the National Institutes of Health Stroke Scale (NIHSS) score. Moreover, it might be associated with an increase in fatal ICH three months after AIS . Another meta-analysis by Zhu et al. in 2020 showed that eptifibatide might be promising when used at a reduced dose (a low dose was also used in the CLEAR study); thus, more randomized trials with different doses are needed to evaluate the role of eptifibatide in the treatment of acute ischemic stroke . A newer retrospective case-control study by Luo et al. (2022) compared routine therapy and treatment with an additional low dose of eptifibatide. Although the study reported no significant differences in NIHSS or adverse events, an analysis of the subgroups showed that eptifibatide is a safe and effective treatment when small artery occlusion is involved . Similarly, a trial by Rana et al. from 2022 showed that using eptifibatide during endovascular therapy in large vessel occlusion is associated with a higher rate of hemorrhages and no benefits to the NIHSS or 90-day mortality . On the other hand, a matched-control analysis by Ma et al. in 2022 showed that the use of eptifibatide was safe and effective in patients undergoing mechanical thrombectomy after ischemic stroke, because the rate of successful recanalization was significantly higher in the intervention group (91.3% versus 81.5%; p = 0.043) and the 3-month outcome on the modified Rankin Scale showed good results (53.1% versus 33.3%; p = 0.016) . As shown, the existing evidence in this research area is not yet conclusive, and further studies are needed to evaluate the use of eptifibatide in the treatment of ischemic stroke. Antiplatelet agents are administered to prevent one of the most critical complications of stenting, namely stent thrombosis. In a study by Osteraas et al., the use of eptifibatide (as a bolus followed by infusion for 24 h after stent placement) was shown to be associated with a lower risk of symptomatic intracranial hemorrhage after carotid stenting . Another study by Horev et al. from 2021 similarly reported a reduced number of complications when using eptifibatide immediately after carotid stenting . Eptifibatide has also been explored as a potential antiplatelet therapy in the stenting of intracranial aneurysms. A 2022 study by Aouni et al. compared three antiplatelet agents (ticagrelor, eptifibatide, and cangrelor) in the stent-assisted endovascular treatment of unruptured intracranial aneurysms and found no significant differences between them . 5.5.5. Septic Shock One of the main mechanisms in septic shock is the activation of the endothelium and platelets. This activation subsequently leads to generalized microvascular damage, microthrombi, capillary leaks, and coagulopathy caused by widespread consumption of coagulation factors . It is noteworthy that research on the use of eptifibatide in septic shock is extremely limited, as to our knowledge, only one study has been performed. The 2019 randomized and placebo-controlled double-blind trial by Berthelsen et al. showed the benefits in improving Sequential Organ Failure Assessment (SOFA) and reducing platelet consumption, fibrinolytic biomarkers, and endothelial damage when a combination of the synthetic analogs of prostacyclin, iloprost, and eptifibatide was used in patients with septic shock . Further research in this area is needed to elucidate the role of eptifibatide. ACS is a manifestation of coronary heart disease associated with an abrupt reduction in the blood supply to the heart. Underlying factors contributing to the disease are smoking, hyperlipidemia, obesity, diabetes, etc. The syndromes comprise different clinical presentations, including unstable angina pectoris, non-ST elevation myocardial infarction (NSTEMI), and ST-elevation myocardial infarction (STEMI). Despite different presentations, all syndromes usually present with chest discomfort at rest . In most cases, the pathophysiological basis of these diseases is the rupture of an atherosclerotic plaque in a cardiac vessel causing platelet aggregation and thrombus formation, which in turn restricts the blood flow to the heart tissue, resulting in cardiac ischemia . Angina Pectoris and non-ST Elevation Myocardial Infarction Eptifibatide is indicated for the prevention of myocardial infarction in unstable angina and NSTEMI in both drug-treated and PCI patients . The PURSUIT trial (1998) showed a reduction in endpoint mortality and a beneficial effect in preventing nonfatal myocardial infarction in these patients . In NSTEMI, a combination of loading dose by aspirin and maintenance treatment by eptifibatide could be used . Aspirin inhibits thromboxane A2 production and therefore prevents platelet aggregation. Although both drugs affect the platelets, their mechanisms of work are different, potentiating their effect. ST-Elevated Myocardial Infarction A meta-analysis by Karathanos et al. in 2019 showed that routine use of GPIs in patients with STEMI was associated with reduced mortality, which was probably the consequence of a reduction in recurrent ischemic events. Despite the promising result, it should be noted that these results are largely based on studies before dual antiplatelet therapy with prasugrel/ticagrelor was routinely used, as is the case today . Although less convincing, eptifibatide can also improve myocardial perfusion in STEMI, as shown in the TITAN-TIMI 34 trial . Nevertheless, more recent studies have shown that prehospital administration of GPI in STEMI has not shown benefits and even increases the bleeding risk compared to routine use in a catheterization laboratory . Although eptifibatide has not been tested in a randomized trial, the European Society of Cardiology guidelines (ESC) suggest it as bail-out therapy in high-risk patients (slow flow or no flow with occlusion of the stent, high thrombus burden, etc.) but not as a routine drug for primary PCI . A meta-analysis by Saleiro et al. in 2020 has shown that the use of GPIs as an adjunct to standard therapy may be beneficial in myocardial infarction that results in cardiogenic shock. In this study, the use of GPIs was associated with better outcomes, namely short-term and long-term survival. Moreover, it did not increase the risk of bleeding in the treated patients . Other newer studies have shown similar results . Eptifibatide is indicated for the prevention of myocardial infarction in unstable angina and NSTEMI in both drug-treated and PCI patients . The PURSUIT trial (1998) showed a reduction in endpoint mortality and a beneficial effect in preventing nonfatal myocardial infarction in these patients . In NSTEMI, a combination of loading dose by aspirin and maintenance treatment by eptifibatide could be used . Aspirin inhibits thromboxane A2 production and therefore prevents platelet aggregation. Although both drugs affect the platelets, their mechanisms of work are different, potentiating their effect. A meta-analysis by Karathanos et al. in 2019 showed that routine use of GPIs in patients with STEMI was associated with reduced mortality, which was probably the consequence of a reduction in recurrent ischemic events. Despite the promising result, it should be noted that these results are largely based on studies before dual antiplatelet therapy with prasugrel/ticagrelor was routinely used, as is the case today . Although less convincing, eptifibatide can also improve myocardial perfusion in STEMI, as shown in the TITAN-TIMI 34 trial . Nevertheless, more recent studies have shown that prehospital administration of GPI in STEMI has not shown benefits and even increases the bleeding risk compared to routine use in a catheterization laboratory . Although eptifibatide has not been tested in a randomized trial, the European Society of Cardiology guidelines (ESC) suggest it as bail-out therapy in high-risk patients (slow flow or no flow with occlusion of the stent, high thrombus burden, etc.) but not as a routine drug for primary PCI . A meta-analysis by Saleiro et al. in 2020 has shown that the use of GPIs as an adjunct to standard therapy may be beneficial in myocardial infarction that results in cardiogenic shock. In this study, the use of GPIs was associated with better outcomes, namely short-term and long-term survival. Moreover, it did not increase the risk of bleeding in the treated patients . Other newer studies have shown similar results . PCI is a non-surgical but invasive procedure in which a catheter is used to insert a stent into narrowed or occluded coronary arteries, improving the blood supply. It is the preferred method of treatment for ACS . As the pretreatment in patients undergoing PCI, the ESC guidelines propose using a combination of eptifibatide and unfractionated heparin, as both anticoagulation and platelet inhibition are important in the pathogenesis-based therapy of NSTEMI . Guidelines from the American College of Cardiology (ACC) and the American Heart Association (AHA) similarly suggest the use of eptifibatide as an initial antiplatelet therapy in patients with high-risk features . Heparin dosages of 50–70 IU/kg i.v. should be used if administered in this combination . Although the FDA has approved using eptifibatide for patients undergoing PCI (including stenting), data for using eptifibatide in the periinterventional treatment of NSTEMI are limited and partially outdated. A major trial that has shown the benefits of eptifibatide use in PCI was the IMPACT-II trial, published in 1997 . Most research on the use of eptifibatide in PCI pretreatment predates routine dual antiplatelet treatment (DAPT) . In periinterventional antiplatelet treatment, oral P2Y12 receptor inhibitors were found to be as effective as GPI and are recommended for routine use . Therefore, little to no evidence exists to support the use of eptifibatide in patients who will undergo coronary angiography and are receiving DAPT . On the other hand, eptifibatide may be considered when facing high-risk PCI patients (slow flow or no-flow with the closure of the stent, high thrombus burden, etc.), patients who did not receive pretreatment with P2Y12 receptor inhibitors, and patients with thrombotic complications . The intracoronary administration of the drug is comparable to intravenous use . Postoperative bleeding prevention after cardiac surgery is crucial to decreasing morbidity and mortality. Since i.v. antiplatelet medications, such as eptifibatide, are quickly cleared from the system and their antiplatelet effect can be quickly reversed, they are used before cardiac and noncardiac surgery as a substitution for oral P2Y12 inhibitors . A 2022 meta-analysis by Wu et al. showed that eptifibatide might be safe and effective when used as a bridging strategy for patients undergoing coronary stent implantation requiring surgery. The GPI might be used without an increased bleeding risk when temporarily discontinuing DAPT. Nevertheless, further randomized studies are needed to substantiate this claim . In a 2019 study by Van Tuyl et al., eptifibatide was shown to be an effective choice for these patients and was even preferred over abciximab . Another drug that is often compared to eptifibatide in bridging strategies is cangrelor. Cangrelor is a reversible P2Y12 receptor inhibitor that prevents ADP-induced platelet aggregation and activation. It should be considered in patients with renal insufficiency, as clearance of eptifibatide is influenced by renal function . Moreover, a study by Yun et al. from 2019 showed that cangrelor and eptifibatide were similar in terms of overall bleeding events and major inpatient cardiac adverse events . The CLEAR trial from 2008 showed that eptifibatide is beneficial in preventing intracerebral hemorrhage in patients with acute ischemic stroke if administered with TPA . Nevertheless, a 2022 meta-analysis by Liu et al. showed that adding eptifibatide to routine thrombolysis or thrombectomy treatment did not improve functional outcomes, favorable outcomes, or the National Institutes of Health Stroke Scale (NIHSS) score. Moreover, it might be associated with an increase in fatal ICH three months after AIS . Another meta-analysis by Zhu et al. in 2020 showed that eptifibatide might be promising when used at a reduced dose (a low dose was also used in the CLEAR study); thus, more randomized trials with different doses are needed to evaluate the role of eptifibatide in the treatment of acute ischemic stroke . A newer retrospective case-control study by Luo et al. (2022) compared routine therapy and treatment with an additional low dose of eptifibatide. Although the study reported no significant differences in NIHSS or adverse events, an analysis of the subgroups showed that eptifibatide is a safe and effective treatment when small artery occlusion is involved . Similarly, a trial by Rana et al. from 2022 showed that using eptifibatide during endovascular therapy in large vessel occlusion is associated with a higher rate of hemorrhages and no benefits to the NIHSS or 90-day mortality . On the other hand, a matched-control analysis by Ma et al. in 2022 showed that the use of eptifibatide was safe and effective in patients undergoing mechanical thrombectomy after ischemic stroke, because the rate of successful recanalization was significantly higher in the intervention group (91.3% versus 81.5%; p = 0.043) and the 3-month outcome on the modified Rankin Scale showed good results (53.1% versus 33.3%; p = 0.016) . As shown, the existing evidence in this research area is not yet conclusive, and further studies are needed to evaluate the use of eptifibatide in the treatment of ischemic stroke. Antiplatelet agents are administered to prevent one of the most critical complications of stenting, namely stent thrombosis. In a study by Osteraas et al., the use of eptifibatide (as a bolus followed by infusion for 24 h after stent placement) was shown to be associated with a lower risk of symptomatic intracranial hemorrhage after carotid stenting . Another study by Horev et al. from 2021 similarly reported a reduced number of complications when using eptifibatide immediately after carotid stenting . Eptifibatide has also been explored as a potential antiplatelet therapy in the stenting of intracranial aneurysms. A 2022 study by Aouni et al. compared three antiplatelet agents (ticagrelor, eptifibatide, and cangrelor) in the stent-assisted endovascular treatment of unruptured intracranial aneurysms and found no significant differences between them . One of the main mechanisms in septic shock is the activation of the endothelium and platelets. This activation subsequently leads to generalized microvascular damage, microthrombi, capillary leaks, and coagulopathy caused by widespread consumption of coagulation factors . It is noteworthy that research on the use of eptifibatide in septic shock is extremely limited, as to our knowledge, only one study has been performed. The 2019 randomized and placebo-controlled double-blind trial by Berthelsen et al. showed the benefits in improving Sequential Organ Failure Assessment (SOFA) and reducing platelet consumption, fibrinolytic biomarkers, and endothelial damage when a combination of the synthetic analogs of prostacyclin, iloprost, and eptifibatide was used in patients with septic shock . Further research in this area is needed to elucidate the role of eptifibatide. Particular attention should be paid to patients who are hypersensitive to the active ingredient or any of the additives, as well as patients on any other parenteral GPI. Another group of contraindications that are important in the use of eptifibatide includes a history of bleeding diathesis within 30 days, current ongoing internal bleeding, and recent significant bleeding within the last six months, either in the gastrointestinal or genitourinary tract . Furthermore, eptifibatide is not recommended in hemodialysis patients due to an increased risk of bleeding. An alternative choice for treatment is abciximab, a mouse-human monoclonal antibody based on a murine analog . Similarly, severe, uncontrolled hypertension with a systolic blood pressure of >200 mmHg and a diastolic blood pressure of >110 mmHg in patients receiving antihypertensive treatment and a history of hemorrhagic stroke are also contraindications for the use of eptifibatide . Other conditions where eptifibatide is contraindicated include thrombocytopenia with a platelet count of <100,000 cells/mm 3 and a prothrombin time > >1.2 × higher from control values or international normalized ratio (INR) ≥ 2.0 . Thrombocytopenia is frequently observed in individuals with hepatic dysfunction; this is thought to be related to splenic sequestration due to portal hypertension or decreased production of thrombopoietin by damaged liver cells. It is also known that bone marrow suppression or folic acid deficiency may occur in people with alcoholic cirrhosis. Accordingly, eptifibatide is contraindicated in these categories of patients . We must be mindful of two other important conditions when initiating eptifibatide, namely recent major surgery or trauma within the past six weeks and a history of intracranial neoplasms, arteriovenous malformations, or aortic dissections. In addition, eptifibatide is a category B drug during pregnancy; however, it should be used in ACS because the benefits outweigh the risks. Nevertheless, special caution is needed in lactating mothers. Caution is also advised for women, the elderly, and people with a low body weight . Eptifibatide is an effective periinterventional antiplatelet agent manufactured as a sterile solution in 10 mL single-dose vials containing 20 mg of eptifibatide and in 100 mL single-dose vials containing 75 mg of eptifibatide. It can be administered via IV and must be protected from direct light prior to administration . Before administration, baseline blood tests, such as complete blood count, prothrombin time (PT)/activated partial thromboplastin time (aPTT), serum creatinine, and activated clotting time, should be performed in patients undergoing PCI. During administration, diligent assessment and monitoring for possible arrhythmias should be performed. The dose regimen for NSTEMI is a double bolus of 180 μg/kg i.v. in 10 min intervals, followed by an infusion of 2.0 μg/kg/min for up to 18 h in patients with normal kidney function. Patients with a creatinine clearance of less than 50 mL/min should have a 50% reduction in their eptifibatide dose. In elderly patients, the hemostatic balance is shifted towards an increased tendency for clotting and reduced fibrinolysis. However, several other contributing factors exist in the elderly that can lead to an increased risk of bleeding. These include distinct pharmacokinetic and pharmacodynamic responses, polypharmacy, and increased comorbidities, all of which can interact with one another and increase the risk of bleeding. Nevertheless, while caution should be exercised in the elderly, advanced age is not a contraindication for the use of GPI . Eptifibatide may be administered concurrently with heparin and aspirin, as it was used in all clinical studies, and the dosage need not be altered. It is known that the concentration of eptifibatide in plasma increased rapidly and remained at a constant level of about 1640 ng/mL from half an hour to 24 h. After that, the concentration of the drug dropped rapidly in multiple stages . 5.8.1. Adverse Effects Eptifibatide may cause a rare and significant adverse effect known as eptifibatide-induced thrombocytopenia (EIT), characterized by an acute, unexpected decrease in platelet count to less than <30,000 platelets/µL. However, patients must be exposed to eptifibatide for 5–7 days before they develop sensitization when receiving it for the first time. EIT was reported in several case studies and occurred in approximately 1 to 2% of patients . It is essential to monitor platelets frequently after starting treatment. The underlying causes of EIT are not fully understood; however, research indicates that it may involve the formation of IgG antibodies that target GPI only when the implicated drug is present. IgG can cause platelet aggregation and secretion, along with the activation of the tyrosine kinase cascade . Enzyme-linked immunosorbent assay testing using monoclonal antibodies can identify platelet glycoprotein targets, GPIIb/IIIa; however, such testing is not widely accessible . Treatment options for EIT include stopping the infusion, platelet transfusions (which may be inefficient because of the half-life of GPI), and fresh frozen plasma (used in major bleeding). Other treatment options include the use of corticosteroids (which are not effective) and intravenous immunoglobulin G (IVIG) . Bleeding is a more common and potentially serious adverse effect. Masood et al. showed that out of 28 reported complications of EIT, 57% involved bleeding events. These included 14% catheterization site bleeding, 14% petechial hemorrhage, 7% groin hematoma, 7% infusion site bleeding, 3.6% gastrointestinal bleeding, 3.6% hemoptysis, 3.6% epistaxis, and 3.6% hematuria . Diffuse alveolar hemorrhage (DAH) is another important but often underappreciated form of eptifibatide-related bleeding. The incidence rate of eptifibatide-induced DAH is 0.5% . In addition to the bleeding and thrombocytopenia mentioned above, side effects on the heart should not be forgotten. The PURSUIT study reported various arrhythmias, such as ventricular tachycardia, atrial fibrillation, ventricular fibrillation, and atrioventricular block, in addition to congestive heart failure and cardiac arrest. However, the side effects were attributed to the underlying disease . As for the other side effects, it is worth mentioning thrombosis, which occurred in 10.75% of patients and included deep vein thrombosis, pulmonary embolism, and in-stent thrombosis. However, the exact relationship between eptifibatide and thrombosis is unclear. The following side effect also occurred in some proportion, namely allergic reactions such as rigor, chills, angioedema, and hypotension. Positive eptifibatide antibodies were also found in 87.5% of cases . 5.8.2. Interactions There are 140 drugs known to interact with eptifibatide, of which 55 had major interactions. Considering the mechanism of action of eptifibatide, caution should be exercised with drugs inhibiting platelet aggregation. These drugs include clopidogrel, ticlopidine, thrombolytics, oral anticoagulants, adenosine, prostacyclin, sulfinpyrazone, nonsteroidal anti-inflammatory drugs (NSAIDs), dipyridamole, and dextran solution. On the other hand, if an oral anticoagulant such as warfarin was taken concomitantly, no further bleeding was observed . Eptifibatide may cause a rare and significant adverse effect known as eptifibatide-induced thrombocytopenia (EIT), characterized by an acute, unexpected decrease in platelet count to less than <30,000 platelets/µL. However, patients must be exposed to eptifibatide for 5–7 days before they develop sensitization when receiving it for the first time. EIT was reported in several case studies and occurred in approximately 1 to 2% of patients . It is essential to monitor platelets frequently after starting treatment. The underlying causes of EIT are not fully understood; however, research indicates that it may involve the formation of IgG antibodies that target GPI only when the implicated drug is present. IgG can cause platelet aggregation and secretion, along with the activation of the tyrosine kinase cascade . Enzyme-linked immunosorbent assay testing using monoclonal antibodies can identify platelet glycoprotein targets, GPIIb/IIIa; however, such testing is not widely accessible . Treatment options for EIT include stopping the infusion, platelet transfusions (which may be inefficient because of the half-life of GPI), and fresh frozen plasma (used in major bleeding). Other treatment options include the use of corticosteroids (which are not effective) and intravenous immunoglobulin G (IVIG) . Bleeding is a more common and potentially serious adverse effect. Masood et al. showed that out of 28 reported complications of EIT, 57% involved bleeding events. These included 14% catheterization site bleeding, 14% petechial hemorrhage, 7% groin hematoma, 7% infusion site bleeding, 3.6% gastrointestinal bleeding, 3.6% hemoptysis, 3.6% epistaxis, and 3.6% hematuria . Diffuse alveolar hemorrhage (DAH) is another important but often underappreciated form of eptifibatide-related bleeding. The incidence rate of eptifibatide-induced DAH is 0.5% . In addition to the bleeding and thrombocytopenia mentioned above, side effects on the heart should not be forgotten. The PURSUIT study reported various arrhythmias, such as ventricular tachycardia, atrial fibrillation, ventricular fibrillation, and atrioventricular block, in addition to congestive heart failure and cardiac arrest. However, the side effects were attributed to the underlying disease . As for the other side effects, it is worth mentioning thrombosis, which occurred in 10.75% of patients and included deep vein thrombosis, pulmonary embolism, and in-stent thrombosis. However, the exact relationship between eptifibatide and thrombosis is unclear. The following side effect also occurred in some proportion, namely allergic reactions such as rigor, chills, angioedema, and hypotension. Positive eptifibatide antibodies were also found in 87.5% of cases . There are 140 drugs known to interact with eptifibatide, of which 55 had major interactions. Considering the mechanism of action of eptifibatide, caution should be exercised with drugs inhibiting platelet aggregation. These drugs include clopidogrel, ticlopidine, thrombolytics, oral anticoagulants, adenosine, prostacyclin, sulfinpyrazone, nonsteroidal anti-inflammatory drugs (NSAIDs), dipyridamole, and dextran solution. On the other hand, if an oral anticoagulant such as warfarin was taken concomitantly, no further bleeding was observed . In conclusion, eptifibatide is a useful cardiac peptide drug that is slowly gaining value in new indications such as the treatment of ischemic stroke, carotid stenting, stenting of intracranial aneurysms, and septic shock. However, more extensive randomized trials are needed to confirm its therapeutic value in these conditions. |
Subsets and Splits