medRxiv : serveur de preprint pour les sciences médicales

Background Outcomes of community antiretroviral therapy (ART) distribution (CAD), in which provider-led ART teams deliver integrated HIV services at health posts in communities, have been mixed in sub-Saharan African countries. CAD outcomes and costs relative to facility-based care have not been reported from Malawi. Methods We performed a retrospective cohort study in two Malawian districts (Lilongwe and Chikwawa districts), comparing CAD with facility-based ART care. We selected an equal number of clients in CAD and facility-based care who were aged >13 years, had an undetectable viral load (VL) result in the last year and were stable on first-line ART for [≥]1 year. We compared retention in care (alive and no period of [≥]60 days without ART) using Kaplan-Meier survival analysis and Cox regression and maintenance of VL suppression (<1,000 copies/mL) during follow-up using logistic regression. We also compared costs (in US$) from the health system and client perspectives for the two models of care. Data were collected in October and November 2020. Results 700 ART clients (350 CAD, 350 facility-based) were included. The median age was 43 years (IQR 36-51), median duration on ART was 7 years (IQR 4-9), and 75% were female. Retention in care did not differ significantly between clients in CAD (89.4% retained) and facility-based care (89.3%), p=0.95. No significant difference in maintenance of VL suppression were observed between CAD and facility-based care (aOR: 1.24, 95% CI: 0.47-3.20, p=0.70). CAD resulted in slightly higher health system costs than facility-based care: $118/year vs. $108/year per person accessing care; and $133/year vs. $122/year per person retained in care. CAD decreased individual client costs compared to facility-based care: $3.20/year vs. $11.40/year per person accessing care; and $3.60/year vs. $12.90/year per person retained in care. Conclusion Clients in provider-led CAD care in Malawi had very good retention in care and VL suppression outcomes, similar to clients receiving facility-based care. While health system costs were somewhat higher with CAD, costs for clients were reduced substantially. More research is needed to understand the impact of other differentiated service delivery models on costs for the health system and clients.

Introduction: Digital twins derived from 3D scanning data were developed to measure soft tissue deformation in head and neck surgery by an artificial intelligence approach. This framework was applied suggesting feasibility of soft tissue shift detection as a hitherto unsolved problem. Methods: In a pig head cadaver model 104 soft tissue resection had been performed. The surface of the removed soft tissue (RTP) and the corresponding resection cavity (RC) was scanned (N=416) to train an artificial intelligence (AI) with two different 3D object detectors (HoloLens 2; ArtecEva). An artificial tissue shift (TS) was created by changing the tissue temperature from 7,91{+/-}4,1{degrees}C to 36,37{+/-}1,28{degrees}C. Results: Digital twins of RTP and RC in cold and warm conditions had been generated and volumes were calculated based on 3D surface meshes. Significant differences in number of vertices created by the different 3D scanners (HoloLens2 51313 vs. ArtecEva 21694, p<0.0001) hence result in differences in volume measurement of the RTC (p=0.0015). A significant TS could be induced by changing the temperature of the tissue of RC (p=0.0027) and RTP (p=<0.0001). RC showed more correlation in TS by heating than RTP with a volume increase of 3.1 l or 9.09% (p=0.449). Conclusions: Cadaver models are suitable for training a machine learning model for deformable registration through creation of a digital twin. Despite different point cloud densities, HoloLens and ArtecEva provide only slightly different estimates of volume. This means that both devices can be used for the task.TS can be simulated and measured by temperature change, in which RC and RTP react differently. This corresponds to the clinical behaviour of tumour and resection cavity during surgeries, which could be used for frozen section management and a range of other clinical applications.

Background: Vascular cognitive impairment (VCI) is one of the most common diseases among the elderly. However, few effective drugs have been approved for VCI. Traditional Chinese medicine (TCM) has been used in dementia for thousands of years. Currently, there is limited high-quality evidence for the efficacy of TCM, and the specific characteristics of its effects and the appropriate patient populations for TCM therapies remain unclear. Herein, we aim to explore the effectiveness and safety of TCM by conducting a longitudinal, patient-centered study. Methods: REgistry for Vascular cognitive Impairment trEatment With Traditional Chinese Medicine (REVIEW-TCM) is a prospective, observational disease registry study. 1000 VCI patients at the Hunan Hospital of Integrated Traditional Chinese and Western Medicine will be recruited based on the following criteria: aged 18 years or older, Montreal Cognitive Assessment (MoCA) score <26, and Hachinski Ischemic Score (HIS)[≥]7. There is no strict limit on the intervention, and different TCM formulas will be focused. Cognition, activity of daily living, quality of life, mental, psychology, ZHENG of TCM, and burden of caregiver will be evaluated at admission, and 6, 12, 18, and 24 months. Meanwhile, biological tests and neuroimaging examination will be applied to further explore the mechanism of TCM. Especially, a mixed-methods embedded design will be applied by adopting quantitative and qualitative studies to explore patients-reported outcomes of TCM. Finally, propensity score matching will be adopted to analyze the effectiveness of TCM. Discussion: To the best of our knowledge, the REVIEW-TCM study is the first comprehensive, prospective, mixed-methods, registry-based study to evaluate TCM treatment in VCI, which will analyze the effectiveness and safety of TCM in the real world and explore population characteristics and subtypes of VCI suitable for TCM. Study registration: This study was registered on www.chictr.org.cn (ChiCTR2200064756).

Background: ICI (immune checkpoint inhibitor) therapy is one of the most promising treatments for melanoma. ICI response however varies among patients, emphasizing the importance of identifying genomic biomarkers to predict likely therapeutic efficacy in advance of treatment. We hypothesised that a lncRNA based immunogencity (lnc-IM) score could be used to predict individual response to ICI treatment, and that this could complement the existing criterion for ICI selection based on tumor mutation burden (TMB). Methodology: The TCGA-SKCM (n=101) and the ICI treated UCLA (n=25), MSKCC (n=16) and DFCI (n=40) melanoma cohorts were used in this study, involving both clinical and transcriptomic data. Each patient was assigned an lnc-IM score based on the number of lncRNA sORF derived peptides predicted to be presented by their tumor's MHC-I genotype. For the ICI treated cohorts, a combined antigen score was defined as a sum of neo-antigen load (derived from TMB) and lnc-IM score. A logistic regression-based classifier was used to predict ICI responses based on these combined antigen scores. Results: Survival analysis showed improved overall survival among patients with low lnc-IM scores (HR= 0.39, p=0.009) in the TCGA-SKCM cohort. We also observed a negative association between tumor immune cell concentration and lnc-IM scores, with low lnc-IM groups showing higher anti-tumor immune cell concentrations . Using the ICI treated cohorts, we demonstrated that a classifier based on combined antigen scoring improved the prediction of immunotherapy outcomes as compared to using TMB alone, yielding an area under the curve (AUC) of 0.71 with an accuracy of 0.54 and recall of 1. We also demonstrated a reduced rate of false negatives (14%) by using a combined antigen score as compared to the use of TMB alone (33%) in ICI treated cohorts. Conclusion: Our findings suggest that the use of combined antigen scores (using lnc-IM scores along with TMB derived neoantigen load) have potential in improving immunotherapy efficacy predictions. Prospective validation in larger cohort sizes is warranted.

Abstract Amyotrophic lateral sclerosis (ALS) is an always fatal neurodegenerative disease characterised by a gradual death of motor neurons in the spinal cord and brain. The cause of ALS is unknown. There appears to be both a genetic and an environmental aspect of ALS disease causation. Multiple occupations are associated with an elevated ALS risk. Interestingly, many of these occupations involve exposure to lead, indicating a possible connection between ALS and lead exposure. Examples include mechanical workers, military service occupations, leather workers and tanners. Gasoline station forecourt attendants, also known as petrol station assistants, show a severely increased ALS risk, and our measurements on forecourt attendants in South Africa show that they display significantly elevated blood lead concentrations. Thus, forecourt attendants can be added to the list of occupations with overlapping risks for lead exposure and ALS incidence. These occupational risks for forecourt attendants are expected to decrease in the future, as leaded gasoline is being phased out worldwide. Nonetheless, the overlapping risks for ALS incidence and lead exposure among forecourt attendants and several other occupations strengthen the hypothesis that lead exposure contributes to ALS.

Objectives: The aim of this study was to investigate how a specifically designed teaching program for adolescents on the subject of pain medication affects their knowledge and attitudes regarding pain medication. Methods This prospective interventional study used both quantitative and qualitative methods. The teaching intervention was codeveloped with end users. Adolescents completed a questionnaire at 3 time points: 1) at baseline before the teaching intervention, 2) immediately after the intervention and 3) at follow-up after approximately 1 or 2 months, depending on the schools' availability. A qualitative component with interviews on a subsample of participants was carried out between baseline and the 1 or 2 months follow-up. Results Nine classes, corresponding to 181 adolescents with a median age of 14 were exposed to the teaching intervention. 22% used pain medication at least once a week at baseline. Their baseline knowledge regarding the mechanism of action, side effects, dosage, and alternative methods to treat pain was low. Their levels of knowledge improved after the teaching intervention and we observed higher levels of knowledge and less uncertainty. However, despite the immediate positive effect, the retainment of knowledge was slightly reduced at the 1 or 2 month follow-up. Discussion Our intervention increased the overall knowledge on pain medication and reduced the adolescents' uncertainty. However, the retainment of knowledge was reduced after 1 or 2 month follow-up. Future interventions carried out on a longer time-span and with the inclusion of parents and delivery of online material might be designed to improve retainment of knowledge.

Introduction: To provide evidence to improve cervical screening for women living with HIV (WLHIV), we assessed the accuracy of screening tests that can be used in low-resource settings and give results at the same visit. Methods: We conducted a paired, prospective study among consecutive eligible WLHIV, aged 18 to 65 years, receiving cervical cancer screening at one hospital in Lusaka, Zambia. The histopathological reference standard was multiple biopsies taken at two time points. The target condition was high-grade cervical intraepithelial neoplasia (CIN2+). The index tests were high-risk human papillomavirus detection (hrHPV, Xpert HPV, Cepheid), portable colposcopy (Gynocular, Gynius), and visual inspection with acetic acid (VIA). Accuracy of stand-alone and test combinations were calculated as the point estimate with 95% confidence intervals. A sensitivity analysis considered disease when only visible lesions were biopsied. Results: Among 371 participants with histopathological results, 27% (101/371) women had CIN2+ and 23% (23/101) was not detected by any index test. Sensitivity and specificity for stand-alone tests were: hrHPV, 67.3% (95% CI: 57.7 to 75.7) and 65.3% (59.4 to 70.7); Gynocular 51.5% (41.9 to 61.0) and 80.0% (74.8 to 84.3); and VIA 22.8% (15.7 to 31.9) and 92.6% (88.8 to 95.2), respectively. The combination of hrHPV testing followed by Gynocular had the best balance of sensitivity (42.6% [33.4 to 52.3]) and specificity (89.6% [85.3 to 92.7]). All test accuracies improved in sensitivity analysis. Conclusion: The low accuracy of screening tests assessed might be explained by our reference standard, which reduced verification and misclassification biases. Better screening strategies for WLHIV in low-resource settings are urgently needed.

Wastewater, which contains everything from pathogens to pollutants, is a geospatially- and temporally-linked microbial fingerprint of a given population. As a result, it can be leveraged for monitoring multiple dimensions of public health across locales and time. Here, we integrate targeted and bulk RNA sequencing (n=1,419 samples) to track the viral, bacterial, and functional content over geospatially distinct areas within Miami Dade County from 2020-2022. First, we used targeted amplicon sequencing (n=966) to track diverse SARS-CoV-2 variants across space and time, and we found a tight correspondence with clinical caseloads from University students (N = 1,503) and Miami-Dade County hospital patients (N = 3,939 patients), as well as an 8-day earlier detection of the Delta variant in wastewater vs. in patients. Additionally, in 453 metatranscriptomic samples, we demonstrate that different wastewater sampling locations have clinically and public-health-relevant microbiota that vary as a function of the size of the human population they represent. Through assembly, alignment-based, and phylogenetic approaches, we also detect multiple clinically important viruses (e.g., norovirus) and describe geospatial and temporal variation in microbial functional genes that indicate the presence of pollutants. Moreover, we found distinct profiles of antimicrobial resistance (AMR) genes and virulence factors across campus buildings, dorms, and hospitals, with hospital wastewater containing a significant increase in AMR abundance. Overall, this effort lays the groundwork for systematic characterization of wastewater to improve public health decision making and a broad platform to detect emerging pathogens.

Classic psychedelics, lysergic acid diethylamide, psilocybin, mescaline and N,N-dimethyltryptamine, are potent psychoactive substances that have been studied for their physiological and psychological effects. However, our understanding of the potential interactions and outcomes when these substances are used in combination with other psychoactive drugs is limited. This systematic review aims to provide a comprehensive overview of the current research on drug-drug interactions between classic psychedelics and other psychoactive drugs in humans. We conducted a thorough literature search using multiple databases, including PubMed, PsycINFO, Web of Science and other sources to supplement our search for relevant studies. A total of 8,487 records published before April 20, 2023, were screened, and studies involving human data describing potential interactions (as well as the lack thereof) between classic psychedelics and other psychoactive drugs were included. In total, we identified 34 reports from 50 studies, encompassing 31 studies on LSD, 11 on psilocybin, 4 on mescaline, 3 on DMT and 1 on ayahuasca. These studies provide insights into the interactions between classic psychedelics and a range of drugs, including antidepressants, antipsychotics, anxiolytics, mood stabilisers, recreational drugs and others. The findings reveal various effects when psychedelics are combined with other drugs, including both attenuated and potentiated effects, as well as instances where no changes were observed. With the exception of a few case reports, no significant adverse drug reactions were discovered in the studies included. In-depth discussions of the results are presented, along with an exploration of potential molecular pathways that underlie the observed effects.

Drug-target Mendelian randomization (MR) is a popular approach for exploring the effects of pharmacological targets. Cis-MR designs select variants within the gene region that code for a protein of interest to mimic pharmacological perturbation. An alternative uses variants associated with behavioral proxies of target perturbation, such as drug usage. Both have been employed to investigate the effects of caffeine but have drawn different conclusions. We use the effects of caffeine on body mass index (BMI) as a case study to highlight two potential flaws of the latter strategy in drug-target MR: misidentifying the exposure and using invalid instruments. Some variants associate with caffeine consumption because of their role in caffeine metabolism. Since people with these variants require less caffeine for the same physiological effect, the direction of the caffeine-BMI association is flipped depending on whether estimates are scaled by caffeine consumption or plasma caffeine levels. Other variants seem to associate with caffeine consumption via behavioral pathways. Using multivariable-MR, we demonstrate that caffeine consumption behavior influences BMI independently of plasma caffeine. This implies the existence of behaviorally mediated exclusion restriction violations. Our results support the superiority of cis-MR study designs in pharmacoepidemiology over the use of behavioral proxies of drug targets.

Background Reducing Ebola virus transmission relies on the ability to limit contact with infected bodily fluids through biosecurity, safe sex practices, safe burial and vaccination. However, armed conflicts can complicate outbreak interventions due to the widespread disruption they cause to the government and the population. Guinea and the Democratic Republic of Congo have historically reported the largest and the most recent Ebola virus outbreaks, and understanding if conflict has played a role in these outbreaks may help in identifying key risks factors which can be used to improve disease control. Methods We used data from a range of publicly available data sources for both Ebola virus cases and conflict events for 2018 to 2021 in Guinea and the DRC. We fitted these data to conditional logistic regression models using the Self-Controlled Case Series methodology, to quantify the magnitude in which conflict increased the risk of reported Ebola virus cases in terms of incidence rate ratio. We re-ran our analysis sub-nationally, by conflict sub-event type and tested the effect of lag. Results Conflict was significantly associated with an increased risk of reported EVD cases in both the DRC and Guinea in recent outbreaks. The effect was of a similar magnitude at 1.88 and 1.98 for the DRC and Guinea, respectively. The greatest effects (often higher than the national values) were found in the most conflict prone areas and during protest/riot-related conflict events. Conflict was influential in terms of Ebola virus risk from 1 week following the event and remained important and, in some cases, more so by 10 weeks. Conclusion Extra vigilance is needed following protests and riot-related conflict events in terms of Ebola virus transmission. These events are highly disruptive in nature, in terms of access to transportation and healthcare and are often in dense urban areas, with high population densities. Additional public health messaging around these types of conflict events, relating to the risks and clinical symptoms may be helpful in reducing transmission. Future work should aim to further understand and quantify conflict severity and intensity, to evaluate a dose-response relationships in terms of disease risk.

Large-scale prevention of respiratory syncytial virus (RSV) infection may have ecological consequences for co-circulating pathogens, including influenza. We assessed if and for how long RSV infection alters the risk for subsequent influenza infection. We analysed a prospective longitudinal cohort study conducted in South Africa between 2016 and 2018. For participating households, nasopharyngeal samples were taken twice weekly, irrespective of symptoms, across three respiratory virus seasons, and real-time polymerase chain reaction (PCR) was used to identify infection with RSV and/or influenza. We fitted an individual-level hidden markov transmission model in order to estimate RSV and influenza infection rates and their interdependence. Of a total of 122113 samples collected, 1265 (1.0%) were positive for influenza and 1002 (0.8%) positive for RSV, with 15 (0.01%) samples from 12 individuals positive for both influenza and RSV. We observed 2.25-fold higher incidence of co-infection than expected if assuming infections were unrelated. We estimated that infection with influenza is 2.13 (95% CI 0.97 - 4.69) times more likely when already infected with, and for a week following, RSV infection, adjusted for age. This equates to 1.4% of influenza infections that may be attributable to RSV in this population. Due to the local seasonality (RSV season precedes the influenza season), we were unable to estimate changes in RSV infection risk following influenza infection. RSV infection was associated with an increased risk for influenza infection for a short period after infection. However, the impact on population-level transmission dynamics of this individual-level synergistic effect was not measurable in this setting.

Abstract Objectives- To study the Echocardiographic manifestations of covid 19 illness among patients admitted in our facility, Correlate MAPSE, TAPSE, PASP, CRP levels and CTSI among covid 19 patients with their 28 day outcome as survivors and non survivors and to look for evidence of residual RV dysfunction and Pulmonary hypertension using TTE after 1 year of follow-up. Study design- Prospective observational study at various medical wards and ICUs in SMS medical college and associated hospitals. Methods- 258 patients with a Covid-19 RT-PCR positive report from a throat or a nasal swab within 72 hours of admission were included in the study. Each patient underwent a complete clinical assessment and routine blood investigations including CRP levels were done. A complete transthoracic echocardiogram was done within 48 hours of admission. Patients also underwent a HRCT chest and CTSI scores were estimated. All patients were followed for a period of 28 days. The MAPSE, TAPSE, PASP, CTSI and CRP levels were then correlated with the outcome of the patient. The survivors again underwent a TTE at 1 year after their recovery from covid-19 illness to look for residual RV dysfunction by TAPSE and the development of pulmonary hypertension as measured by PASP using Bernoulli?s equation. Results-Amongst patient of covid 19 illness the MAPSE, TAPSE, PASP, CTSI and CRP levels all correlated well with outcome of patients. While most covid-19 survivors recovered from their illness yet some patients showed evidence of persistent RV dysfunction and pulmonary hypertension even after 1 year of follow up.

Chronic kidney disease (CKD) and chronic kidney disease of uncertain etiology (CKDu) are chronic renal diseases that pose a significant health burden in Sri Lanka. Leptospirosis is a bacterial zoonosis that primarily damages kidney tissues via colonization of Leptospira spp. in the renal tubules and is a suspected etiological agent of CKDu. Since Sri Lanka is a leptospirosis endemic country where outbreaks of the disease have been reported, this study aimed to find the association between leptospirosis and chronic renal disease in two geographically distinct regions of Sri Lanka, Kandy (CKDu non-endemic) and Badulla (CKDu endemic) districts. Forty-nine chronic renal disease patients and 135 controls from Kandy and 89 chronic renal disease patients and 149 controls from Badulla were examined serologically by microscopic agglutination test with a panel of 11 Leptospira serogroups. A seroprevalence of 35.3% and 32.5% for leptospirosis was observed in the Girandurukotte CKDu group and Girandurukotte control group respectively, while a seroprevalence of 36.7% and 31.9% were observed in the Kandy CKDu group and Kandy control group respectively. No statistically significant differences in leptospirosis seropositivity were observed between the chronic renal disease and control groups in both districts. However, further longitudinal studies assessing renal colonization among chronic renal patients and healthy individuals are required to conclusively state whether leptospirosis plays an important role in chronic renal disease development and/or progression in Sri Lanka.

In a fraction of SARS-CoV-2 infected individuals treated with the oral antiviral Paxlovid, the virus rebounds following treatment. The mechanism driving rebound is not understood. Here, we show that viral dynamic models based on the hypothesis that Paxlovid treatment near the time of symptom onset halts the depletion of target cells, but may not fully eliminate the virus, which can lead to viral rebound. We also show that the occurrence of viral rebound is sensitive to model parameters, and the time treatment is initiated, which may explain why only a fraction of individuals develop viral rebound. Finally, the models are used to test the therapeutic effects of two alternative treatment schemes. These findings also provide a possible explanation for rebounds following other antiviral treatments for SARS-CoV-2.

This study examined health information technology-related (HIT) incidents to identify patient details-related issues, their association with contributing factors, and outcomes. Sources of information comprised retrospectively collected incident reports (n=95) using two sampling methods, i.e., purposive and snowball sampling. The reports were collected in two formats - interviews (written and telephone) and/or a set of already existing reports from the local database. The incident reports were analyzed using both the inductive method (thematic analysis) and the deductive approach using an existing framework, i.e., the International Classification for Patient Safety. The studies identified 90 incidents with 120 patient details-related issues--categorized as either information-related (48%) or documentation-related (52%) problems; around two-thirds of the 120 issues were characterized by human factors. Of the total sample, 87 contributing factors were identified, of which "medical device/system" (45%) and "documentation" (20%) were the most common contributing factors. Of 90 incidents, more than half (59%) comprised patient-related outcomes--patient inconvenience (47%) and patient harm (12%) and the remaining 41% (n=37) included staff or organization-related outcomes. The study confirms that patient details-related problems with HIT systems were more likely to affect patient care delivery - more than half of the incidents resulted in patient-related outcomes, namely patient inconvenience and patient harm, including disease risks, severe health deterioration, injury, and even patient death. Incidents associated with patient details can cause deleterious effects; therefore, characterizing them should be a routine part of clinical practice to improve the constantly changing healthcare system.

The adoption of electronic health records (EHRs) has created opportunities to analyze historical data for predicting clinical outcomes and improving patient care. However, non-standardized data representations and anomalies pose major challenges to the use of EHRs in digital health research. To address these challenges, we have developed EHR-QC, a tool comprising two modules: the data standardization module and the preprocessing module. The data standardization module migrates source EHR data to a standard format using advanced concept mapping techniques, surpassing expert curation in benchmarking analysis. The preprocessing module includes several functions designed specifically to handle healthcare data subtleties. We provide automated detection of data anomalies and solutions to handle those anomalies. We believe that the development and adoption of tools like EHR-QC is critical for advancing digital health. Our ultimate goal is to accelerate clinical research by enabling rapid experimentation with data-driven observational research to generate robust, generalisable biomedical knowledge.

Many areas of medicine would benefit from deeper, more accurate phenotyping, but there are limited approaches for phenotyping using clinical notes without substantial annotated data. Large language models (LLMs) have demonstrated immense potential to adapt to novel tasks with no additional training using in-context learning. We investigated the performance of a publicly available LLM, Flan-T5, in phenotyping patients with postpartum hemorrhage (PPH) using discharge notes from electronic health records (n=271,081). The language model achieved strong performance in extracting 24 granular concepts associated with PPH. Identifying these granular concepts accurately allowed the development of interpretable, complex phenotypes and subtypes. The Flan-T5 model achieved high fidelity in phenotyping PPH (positive predictive value of 0.95), identifying 47% more patients with this complication compared to the current standard of using claims codes. This LLM pipeline can be used reliably for subtyping PPH and outperformed a claims-based approach on the three most common PPH subtypes associated with uterine atony, abnormal placentation, and obstetric trauma. The advantage of this approach to subtyping is its interpretability, as each concept contributing to the subtype determination can be evaluated. Moreover, as definitions may change over time due to new guidelines, using granular concepts to create complex phenotypes enables prompt and efficient updating of the algorithm. Using this language modelling approach enables rapid phenotyping without the need for any manually annotated training data across multiple clinical use cases.

Introduction: AI interventions for health care are on the rise. Decisions about coverage and reimbursement are often informed by Health Technology Assessment (HTA) bodies, who rely on Health Economic Evaluations (HEEs) to estimate the value for money (cost effectiveness) of interventions. Transparent reporting of HEEs ensures they can be used for decision making. Reporting guidance exists to support this, such as the Consolidated Health Economic Reporting Standards (CHEERS) checklist. We aim to identify consensus about specific items should be reported by HEEs that evaluate AI interventions and, if such items are identified, to develop them into an extension to CHEERS: CHEERS-AI Methods and analysis: The project will have 4 phases: - Phase 1 is a literature review to help identify potential AI-related reporting items. - Phase 2 commences a Delphi process, with a series of surveys to elicit the importance of the potential AI-related reporting items. - Phase 3 is a consensus-generation meeting to agree on the final extension items. - Phase 4 is dissemination of the project outputs.

Background: The best curative treatment for hepatocellular carcinoma (HCC) is liver transplant (LT), but the limited number of organs available for LT dictates strict eligibility criteria. Despite this patient selection stringency, current criteria often fail in pinpointing patients at risk of HCC relapse and in identifying good prognosis patients that could benefit from a LT. HepatoPredict kit was developed and clinically validated to forecast the benefit of LT in patients diagnosed with HCC. By combining clinical variables and a gene expression signature in an ensemble of machine learning algorithms, HepatoPredict stratifies HCC patients according to their risk of relapse after LT. Methods: Aiming at the characterization of the analytical performance of HepatoPredict kit in terms of sensitivity, specificity and robustness, several variables were tested which included reproducibility between operators and between RNA extractions and RT-qPCR runs, interference of input RNA levels or varying reagent levels. The described methodologies, included in the HepatoPredict kit, were tested according to analytical validation criteria of multi-target genomic assays described in guidelines such as ISO201395-2019, MIQE, CLSI-MM16, CLSI-MM17, and CLSI-EP17-A. Furthermore, a new retrained version of the HepatoPredict algorithms is also presented and tested. Results: The results of the analytical performance demonstrated that the HepatoPredict kit performed within the required levels of robustness (p > 0.05), analytical specificity (inclusivity [≥] 95 %), and sensitivity (LoB, LoD, linear range, and amplification efficiency between 90 and 110 %). The introduced operator, equipment, input RNA and reagents into the assay had no significant impact on HepatoPredict classifier results. As demonstrated in a previous clinical validation, a new retrained version of the HepatoPredict algorithm still outperformed current clinical criteria, in the accurate identification of HCC patients that more likely will benefit from a LT. Conclusions: Despite the variations in the molecular and clinical variables, the prognostic information obtained with HepatoPredict kit and does not change and can accurately identify HCC patients more likely to benefit from a LT. HepatoPredict performance robustness also validates its easy integration into standard diagnostic laboratories.