Clinical Trials

Background: Statistical Analysis Plans (SAPs) are essential for trial transparency and credibility but are resource-intensive to produce. While Large Language Models (LLMs) have shown promise in drafting protocols, their ability to generate high-quality, protocol-compliant SAPs remains untested against current content guidance. This study developed and validated an LLM-based pipeline for drafting SAPs from clinical trial protocols. Methods: We developed a structured, section-by-section prompting pipeline aligned with standard SAP guidance. We applied this pipeline to nine clinical trial protocols using three leading LLMs: OpenAI GPT-5, Anthropic Claude Sonnet 4, and Google Gemini 2.5 Pro. The resulting 27 SAPs were evaluated against a 46-item quality checklist derived from the published SAP guidelines. Items were double-scored by independent trial statisticians on a 0 to 3 scale for accuracy. We compared performance across LLMs and between item types (descriptive vs. statistical reasoning) using mixed-effects logistic regression. Results: Across 9 trials, the models produced SAP drafts with high overall accuracy (77% to 78%), with no difference in performance between the three LLMs (p=0.79) but varied by content type (p < 0.001). All models performed well on descriptive items (e.g., administrative details, trial design), with lower accuracy for items requiring statistical reasoning (e.g., modelling strategies, sensitivity analyses). Accuracy for statistical items ranged from 67% to 72%, whereas descriptive items achieved 81% to 83% accuracy. Qualitatively, models were prone to specific failure modes in complex sections, such as omitting necessary details for secondary outcome models or hallucinating sensitivity analyses. Discussion: Current LLMs can effectively draft portions of SAPs, offering the potential for substantial time savings in trial documentation. However, a human-in-the-loop approach remains mandatory; while models demonstrate strong capability in producing descriptive content, their independent application to complex statistical methodology design still requires further methodological development and training. Future work should explore advanced prompt engineering, such as retrieval-augmented generation or agentic workflows, to improve reasoning capabilities.

ObjectiveAntimicrobial resistance (AMR) is an urgent global health threat, resulting in more than 5 million deaths globally in 2019. Timely and complete antimicrobial agent (AMA) clinical trial results reporting is essential to evaluate the safety and efficacy of investigational therapies. The Food and Drug Administration Amendments Act (FDAAA) of 2007 mandated results reporting for applicable clinical trials to ClinicalTrials.gov. After nearly ten years of underreporting, the HHS issued the Final Rule, requiring a designated responsible party to submit results to ClinicalTrials.gov and clarifying applicable clinical trial (ACT) criteria. ACTs and probable ACTs (pACTs) are interventional studies regulated by the FDA with at least one site based in the United States. However, pACTs were initiated prior to January 2017, when the Final Rule came into effect. This study investigates the compliance and timeliness of results reporting of ACTs and probable ACTs (pACTs) for AMAs. DesignWe extracted data from ClinicalTrials.gov for trials involving AMAs with primary completion dates between May 1, 2013, and May 1, 2023. We analyzed the time from primary completion to results reporting and estimated the hazard ratio to compare timeliness between ACTs and pACTs. Additionally, we assessed delays in reporting across different study types and funding sources. ResultsOur search resulted in 2629 NCT records. After exclusion of ineligible trials, we included 2525 trials. We found 1769 pACTs (70.1%; 95% CI, 69.3%-72.9%) and 756 ACTs (29.9%; 95% CI, 28.2%-31.8%). Among the 2525 eligible trials, 2249 trials (89.1%; 95% CI, 87.8%-90.2%) were reported on ClinicalTrials.gov or in journal publications. Overall, 81.3% (95% CI, 79.7%-82.3%) of trials were reported late or missing (75.0% of ACTs vs 83.6% of pACTs). ACTs were more likely to report results earlier than pACTs, with a hazard ratio of 1.4 (95% CI, 1.3-1.5). ConclusionsACTs demonstrated greater reporting compliance and shorter delays in the reporting of overdue results. While this analysis provides initial insights, limitations related to timeline and sample scope suggest that broader investigations are needed to fully evaluate the impact of the Final Rule.

BackgroundThe strength and transparency of clinical trial evidence supporting drug approvals has become increasingly scrutinized, particularly considering the increased use of regulatory flexibility and expedited pathways. While U.S. Food and Drug Administration (FDA) standards have been extensively analyzed, evidence standards at the European Medicines Agency (EMA) remain less well-characterized. Thus, this study aims to systematically assess the design, quality, and outcomes of pivotal efficacy trials supporting EMA drug approvals between 2020 and 2023. MethodsWe conducted a cross-sectional analysis of new medicines and biosimilars receiving positive opinions from the EMAs Committee for Medicinal Products for Human Use (CHMP) and subsequent approval by the European Commission between January 2020 and December 2023. Data were extracted from European Public Assessment Reports (EPARs) and EMA medicine databases. Key variables included trial design features, primary endpoint type and achievement status, and justification for approval in cases of failed efficacy endpoints. ResultsBetween 2020 and 2023, 232 drugs were approved by the EMA for 281 indications. Of these, 205 (88.4%) were new active substances and 65 (28.0%) were granted orphan designation. Forty-six products (19.8%) were approved via a special regulatory program, most commonly Conditional Approval (26 products; 11.2%). Cancer was the leading therapeutic area, accounting for 61 approvals (26.3%). Approvals were supported by 393 pivotal clinical trials. Of these, 327 (83.2%) were randomized controlled trials (RCTs) and 218 (66.6% of RCTs) had a superiority design. A total of 232/393 trials (59.0%) relied on surrogate endpoints. Overall, 22 approvals (9.5%) were supported by at least one pivotal trial in which at least one primary endpoint was not met; in seven of these cases (31.8%), the failed trial was the sole pivotal trial. The most common rationale for approval despite null primary results was reliance on the totality of evidence, secondary endpoints, or clinical judgment (9 products; 40.9%). ConclusionsOur findings reveal substantial variability in the design and evidentiary strength of pivotal trials supporting EMA approvals between 2020 and 2023. While the majority of studies were RCTs, reliance on surrogate endpoints was common. That 10% of approvals were based on pivotal trials with null primary endpoints highlights the nuanced role of regulatory judgment in therapeutic evaluation. These findings prompt reflection on evolving evidence standards in drug regulation and underscore the need for transparency and consistent justifications.

BackgroundIncomplete dissemination of clinical trial results remains an important challenge for research transparency and evidence synthesis. Although prior studies have quantified the overall extent of non-dissemination, less is known about whether trial characteristics observable at registration are associated with subsequent dissemination within sponsor portfolios. Methods and findingsWe conducted a retrospective cohort study of 17,537 completed interventional clinical trials registered on ClinicalTrials.gov between 2007 and 2024 across the 20 largest global pharmaceutical companies. We developed the Operational Complexity Index (OCI), a composite measure derived from planned enrollment, facility count, and geographic scope, and examined its association with trial dissemination using multivariable logistic regression and time-to-event analyses. Higher OCI was associated with greater odds of dissemination (adjusted odds ratio [aOR] = 2.40, 95% CI 2.23-2.60; p < 0.001), with dissemination increasing from 47% in the lowest OCI decile to 95% in the highest. Higher operational complexity was also associated with earlier dissemination; over a 1,095-day horizon, high-OCI trials were disseminated a mean of 310.88 days earlier than low-OCI trials (RMST difference, 310.88 days; 95% CI 300.59-320.96; p < 0.001). This pattern was observed across sponsors, clinical phases, and therapeutic areas. In predictive analyses using registration-time variables, the structural model achieved a cross- validated AUC of 0.816 and a holdout AUC of 0.814, whereas the full model, including sponsor identity, achieved a cross-validated AUC of 0.858 and a holdout AUC of 0.857. Using benchmark phase-based costing assumptions, the 5,019 non-disseminated trials corresponded to an estimated US$10.94-15.26 billion in sunk research investment. ConclusionsAmong trials conducted by the 20 largest pharmaceutical sponsors, greater operational complexity at registration was associated with a higher likelihood of dissemination and earlier dissemination. These findings suggest that aggregate sponsor-level transparency metrics may mask important heterogeneity within sponsor portfolios. Future work should assess whether registration-time trial characteristics can help identify trial subgroups at higher risk of non-dissemination. AUTHOR SUMMARYO_ST_ABSWhy was this study done?C_ST_ABSO_LIIncomplete dissemination of clinical trial results reduces the completeness of the medical evidence base and the public value of research participation. C_LIO_LIPrevious studies have described overall rates of trial non-dissemination, but less is known about whether dissemination varies systematically across different types of trials within sponsor portfolios. C_LIO_LIWe examined whether trial characteristics available at registration were associated with later dissemination of results among large pharmaceutical sponsors. C_LI What did the researchers do and find?O_LIWe analyzed 17,537 completed interventional clinical trials sponsored by the 20 largest pharmaceutical companies and registered on ClinicalTrials.gov between 2007 and 2024. C_LIO_LIWe developed an Operational Complexity Index (OCI) based on planned enrollment, number of facilities, and geographic scope to measure trial operational scale at registration. C_LIO_LIHigher OCI was associated with a greater likelihood of dissemination and earlier dissemination. Dissemination ranged from 47% in the lowest OCI decile to 95% in the highest. C_LIO_LIThis pattern was observed across sponsor portfolios, clinical phases, and therapeutic areas, with an average within-sponsor dissemination gap of 40 percentage points between lower- and higher-complexity trials. C_LIO_LIIn manual validation of 344 sampled trials, the automated dissemination-classification pipeline achieved 92.1% accuracy. C_LIO_LIUsing benchmark phase-based costing assumptions, the 5,019 non-disseminated trials corresponded to an estimated US$10.9-15.3 billion in sunk research investment. C_LI What do these findings mean?O_LIDissemination was not uniform across trial types within sponsor portfolios; trials with lower operational complexity were less likely to be disseminated than trials with higher operational complexity. C_LIO_LIAggregate sponsor-level transparency measures may therefore miss important differences within portfolios. C_LIO_LIRegistration-time trial characteristics showed predictive signal for non-dissemination, but whether such information could support monitoring strategies would require prospective validation. C_LIO_LIMore complete dissemination of trial results would strengthen the scientific record and improve the public value of clinical research. C_LI

Preregistration at public research registries is considered a promising solution to the credibility crisis in science, but empirical evidence of its actual benefit is limited. Guaranteeing research integrity is especially vital in clinical research, where human lives are at stake and investigators might suffer from financial pressure. This paper analyzes the distribution of p-values from pre-approval drug trials reported to ClinicalTrials.gov, the largest registry for research studies in human volunteers, conditional on the preregistration status. The z-score density of non-preregistered trials displays a significant upward discontinuity at the salient 5% threshold for statistical significance, indicative of p-hacking or selective reporting. The density of preregistered trials appears smooth at this threshold. With caliper tests, we establish that these differences between preregistered and non-preregistered trials are robust when conditioning on sponsor fixed effects and other design features commonly indicative of research integrity, such as blinding and data monitoring committees. Our results suggest that preregistration is a credible signal for the integrity of clinical trials, as far as it can be assessed with the currently available methods to detect p-hacking.

Pediatric trials are ethically and logistically difficult, so the U.S. FDA often extrapolates adult data to children when justified. Yet no public resource systematically documents these decisions. We present PED-X-Bench, the first dataset and benchmark that encodes FDA pediatric-extrapolation outcomes as a four-way classification task (Full, Partial, None, Unlabeled). PED-X-Bench contains 737 FDA drug-label sections ({approx} 1 M words of source text) for approvals issued 2007-2024 across all therapeutic areas. A two-stage o3-mini prompting pipeline mined full FDA label text; nine domain reviewers then adjudicated a stratified sample of 135 labels yielding an accuracy F1 of 0.74 and 0.63 respectively (inter-annotator {kappa} = 0.678) and spot-checking the remainder. For every drug we release the ground-truth label, concise efficacy and pharmacokinetic/safety summaries, and harmonized study metadata. To showcase utility we release two baseline models: (i) a logistic-regression classifier that uses structured metadata from FDAs pediatric-labeling dataset, and (ii) a fine-tuned BigBird BERT that ingests full label text. Both base-lines perform modestly, leaving ample headroom for future work. PED-X-Bench enables research on pediatric drug development, clinical NLP and drug safety; dataset card and code are made available here: github.com/tatonetti-lab/PedXBench huggingface.co/datasets/apoorvasrinivasan/Ped-X-Bench

Background Simon two-stage designs for binary endpoints and their time-to-event analogues, including the Kwak and Jung method, rely on a fixed null benchmark. Their Type I error control is valid only when that benchmark is correctly specified. In practice, historical benchmarks are often inconsistent due to small samples, population heterogeneity, changing eligibility criteria, and evolving standards of care. Even modest misspecifications can substantially inflate the Type I error rate, leading to costly advancement of ineffective treatments. Methods We propose the Interval-Null Robust (INR) two-stage design framework that accounts for uncertainty in the historical null benchmark. We define the null hypothesis as a plausible range of clinically uninteresting values: p[isin][p0L, p0U] for binary endpoints and {lambda}[isin][{lambda}0L, {lambda}0U] (or equivalent survival probabilities) for time-to-event endpoints. Type I error is controlled uniformly over the full null interval: sup{theta}[isin]{theta}0 Pr{theta}(Go) [&le;] . Under the monotonicity of the Go probability, the supremum occurs at the least favorable null configuration - p0U and {lambda}0L - but the design is not reduced to a point-null formulation. The interval defines the uncertainty set for error control and is used in selecting among feasible designs through robust criteria such as worst-case regret or minimal average expected sample size. Results Across representative planning scenarios for both endpoint types, classic designs calibrated to a single benchmark exhibit substantial Type I error inflation when the true null parameter exceeds the assumed planning value. INR designs maintain the nominal Type I error rate across the full null interval, directly addressing this vulnerability to benchmark misspecification. The robustness-efficiency trade-off can be managed through design constraints and robust optimization criteria while preserving uniform Type I error control. Conclusions INR two-stage designs offer a transparent framework for addressing historical control uncertainty in single-arm Phase II trials. By replacing reliance on a fixed benchmark assumption with a more realistic interval of clinically plausible null values, INR design reduces the risk of false-positive Go-decisions caused by benchmark misspecification. INR applies to both binary and time-to-event endpoints and is implemented in the open-source INRDesign R package and accompanying interactive Shiny app.

Objectives: To assess the availability of key clinical trial registration data and compliance with legal reporting requirements for all Phase 2-4 drug trials registered on the new European Clinical Trial Information System (CTIS) registry. This study is the first ever assessment of data quality and legal compliance with reporting requirements on CTIS. Design: Cross-sectional observational study of CTIS registry data combined with manual review of results documents. Setting: Cohort of all 7,547 Phase II-IV clinical trials registered on CTIS as of November 2025. Main outcome measures: Number and proportion of missing data points in CTIS registration data. Proportion of completed clinical trials that are compliant with regulatory reporting requirements. Results: Trial registration data quality was high overall with more than 99% of expected data present. Of 234 clinical trials legally required to report results, fewer than half (49.6%) fully reported results within the required timeframe, 20 trials (8.5%) fully reported results late, and 98 trials (41.9%) failed to fully report results. Legal compliance was similar for adult trials (79/158) and paediatric trials (37/76). Conclusions: Sponsor compliance with legal reporting requirements is weak. Current efforts by European regulators to monitor and enforce compliance appear to be insufficient. New results reporting functions currently being set up by trial registries worldwide will require quality assurance processes. Trial registration: Study protocol prospectively registered on OSF: https://osf.io/sn4j2/overview

Statistical calculations for clinical trials traditionally assume that if a treatment fails, it fails for mechanistic reasons--the drug itself is ineffective. However, patients may be treatment-resistant, rendering them unable to benefit from an otherwise effective treatment. This creates an identifiability problem: a null hypothesis that we fail to reject can indicate either an ineffective treatment, or an effective treatment tested in a population dominated by treatment-resistant subjects. However, the strategy to administer the drug should be different for these cases. Here we present a simple way to adjust the sample size of a randomized controlled trial to account for the anticipated level of treatment resistance to reach a certain statistical power. We show that the resistant-adjusted population size exponentially increases with the anticipated resistance prevalence, whereas power decreases almost linearly for a given population size as the resistance prevalence increases.

Target trial emulation (TTE) offers a principled way to estimate treatment effects using real-world observational data, but analyses of time-varying treatment strategies remain vulnerable to immortal time bias. The clone-censor-weight (CCW) approach is increasingly used to address this problem, yet key aspects of its causal interpretation and implementation remain unclear. In this work, we emulate a target trial using electronic health records (EHRs) to compare completion of a 3-dose 9-valent human papillomavirus vaccination (HPV) series within 12 months versus remaining partially vaccinated among vaccine initiators. We link CCW to the classic potential outcome framework in causal inference, evaluate the role of different weighting mechanisms, and account for within-subject correlation induced by cloning using cluster-robust variance estimation. Our study provides practical guidance for applying CCW in real-world comparative effectiveness studies to address immortal time bias and supports more rigorous and interpretable treatment effect estimation in TTE.

BackgroundNovel therapeutics should always be tested in a sample representative of the population in need of treatment. Initial efforts of drug development take place in early phase trials (phase-I and -II), setting the direction for late-stage studies (phase-III and -IV). However, study samples in early phase trials typically fail to recruit Black, Asian and minority ethnic groups, which might produce results which dont generalise to a broader population in later trials, and ultimately, clinical practice. Focusing on early phase clinical trials the present study (1) explored the barriers and incentives that determine participation of ethnic minorities in clinical research, and (2) proposes strategies that mitigate such barriers. MethodsA systematic literature review explored barriers affecting participation rates from individuals from diverse ethnic backgrounds. An exploratory phase involved two online surveys (researchers and general population) and focus groups (general population) analysed using thematic analysis. ResultsThe systematic review found little published evidence, with most studies undertaken in the USA and focused on specific clinical areas. The exploratory phase showed a discordance between researchers and general publics perspectives on both drivers and barriers to early phase trial participation. DiscussionThese findings were synthesised into a Clinical Trials Participatory Framework, which contextualises reasons for reduced trial participation, while providing mechanisms/strategies to increase uptake among minority ethnic participants. This may guide researchers when implementing strategies to aid under-representation in their samples. Further research should evaluate the framework by actively implementing, testing, and iterating upon the strategies.

BackgroundMultiple stakeholders need to locate results of registered clinical trials but frequently struggle to find them. Summary results of clinical trials are often not published in trial registries, and publications containing trial results are often not explicitly linked to their respective trial registrations. Finding these results is important to researchers, systematic reviewers, research funders, regulators, clinical practitioners, and patients. MethodsWe developed TrialScout, a computer program that uses a large language model to match clinical trials registered on ClinicalTrials.gov with corresponding result publications indexed in PubMed. TrialScouts performance was evaluated through comparison to human-coded matches from previous studies of results reporting rates. Subsequently, TrialScout was applied to a random sample of 9,600 completed or terminated trials. ResultsTrialScout had a sensitivity of 92.5% and a specificity of 81.2% compared to human coders. Manual review of 200 cases where TrialScout disagreed with human researchers showed that a majority (123/200, 61.5%, 95% CI, 54.4-68.3%) of disagreements were due to human errors. When used on 9,600 sampled trials in ClinicalTrials.gov, TrialScout found result publications for 6,110 (63.6%) of trials. DiscussionTrialScout reliably located results of completed clinical trials. The tool offers benefits in terms of speed and efficiency. Estimating TrialScouts accuracy is limited by the lack of a true gold standard. TrialScout can accelerate the process of locating trial results in the scientific literature and can assist in monitoring trial reporting practices.

BackgroundLarge language models (LLMs) are increasingly used in randomized clinical trial (RCT) screening, but their potential for sociodemographic bias remains unclear. ObjectiveTo determine whether LLM-based trial screening judgments vary with patient sociodemographic characteristics when clinical details and eligibility criteria are held constant. Design, Setting, and ParticipantsCross-sectional evaluation of Phase II-III RCT protocols from ClinicalTrials.gov (U.S. adult populations; 2023-2024). For each protocol, we created 15 physician-validated clinical vignettes rendered in 34 versions: one control (no identifiers) and 33 identity variants spanning gender, race/ethnicity, socioeconomic status, homelessness, unemployment, and sexual orientation. ExposuresIdentity labels applied to otherwise identical vignettes, evaluated by nine contemporary LLMs. Main Outcomes and MeasuresPrimary eligibility domain score (1-5 Likert scale) comparing identity variants versus control. Secondary: adherence, resources, risk-benefit, and trust/attitude domains. Mixed-effects models estimated adjusted mean differences with multiplicity-corrected P values; differences <.10 considered trivial. ResultsOf 69 protocols, 58 met inclusion criteria. Analysis of 5,324,400 model evaluations showed eligibility judgments were largely stable: most identity-related differences fell within {+/-}0.05 (transgender woman -.008 [95% CI -.04 to .02]; White male .036 [.01 to .07]). Only homelessness exceeded the trivial threshold (-.121 [-.15 to -.09], P<.001). Secondary domains revealed socioeconomic gradients, particularly for adherence (homeless -.595, P<.001) and resources (homeless -.715, P<.001), with smaller trust/attitude effects and negligible risk-benefit differences. Conclusions and RelevanceBias in LLM-assisted trial screening is conditional. Within fixed criteria, models reason consistently; outside them, they echo the inequities of their data. Responsible deployment in clinical research depends on preserving that boundary so that automation strengthens fairness in trial access rather than inheriting distortion.

One component of a systematic review is the quality assessment of studies to determine their inclusion or exclusion. Studies on e-cigarettes are conducted in the contentious atmosphere surrounding tobacco harm reduction, which has resulted at times in research bias. Therefore, the quality assessment of studies on e-cigarettes requires more scrutiny than what is provided by generic tools on study design. This topic-specific quality assessment must examine the tests, measurements, and analysis methods used for their adherence to research standards. Furthermore, the studies need to be carefully screened for bias. Because standard quality assessment tools do not provide this topic-specific guidance, we propose to develop quality assessment tools specifically for reviews on e-cigarettes, and for our living systematic reviews on e-cigarettes for tobacco harm reduction.

Randomized controlled trials (RCTs) provide high internal validity but often rely on restrictive eligibility criteria that limit generalizability and complicate real-world trial emulation. We propose AERO (AI Agent for Adaptive Eligibility Refinement and Optimization), an agentic framework that systematically adapts clinical trial eligibility criteria for application to electronic health record data. AERO integrates external clinical knowledge sources and large language model-based reasoning to classify criteria as strict inclusion, safety exclusion, confounder, or operational artifact. We evaluated AERO by emulating the WARCEF trial using Mayo Clinic Platform data restricted to the pre-trial completion period. Emulation with optimized criteria yielded a hazard ratio of 1.561 (p = 0.0605), consistent with the original neutral trial finding (HR = 1.01, p = 0.91). An ablation analysis demonstrated that eligibility handling decisions materially influence observed treatment effects. These results highlight the importance of systematic, knowledge-informed eligibility refinement in real-world evidence generation.

Clinical trials are a critical phase in the development of new drugs, and the recruitment of research participants is a pivotal prerequisite for the success of these trials. However, participants recruitment and retention often face numerous challenges, such as delayed recruitment progress, high rates of informed consent failure, and poor participants compliance, which significantly impact the timeline and quality of clinical trial projects. Through in-depth analysis of 4 core processes of participants recruitment and retention, this study explores the application of the Failure Mode and Effects Analysis (FMEA) method to identify four high-risk failure modes and three medium-high risk failure modes within the core processes of participants recruitment and retention. By implementing a multifaceted and systematically proactive approach, including leveraging established relationships with community providers, appointing dedicated full-time personnel, providing targeted incentives for both research staff and participants, and maintaining persistent communication, this exploratory study integrates preemptive risk identification and mitigation strategies aimed at participant recruitment and retention. This methodology seeks to establish a sustainable model for continuous improvement in clinical trial efficiency and participant engagement, focusing on the early detection and resolution of potential barriers.

Machine learning models for causal inference aim to adjust for confounding factors that are associated with both an exposure and an outcome, creating a spurious biased association. But, these methods are rarely empirically evaluated to assess their success in mitigating such bias. Recent advances in knowledge representation, including both foundation models and knowledge graphs, could enrich these models, but rigorous evaluations are needed in order to assess their potential. Here, we ask whether enriching existing causal inference models with knowledge representations from foundation models can improve confounding control. Rather than using semi-simulated data to address this question, we focus on examples of real confounding: we emulate target randomized active comparator trials that are subject to confounding by indication. Our results can guide researchers aiming to develop or apply methods for discovering causal effects from observational data.

Large language models are increasingly used for clinical guidance while their parent companies introduce advertising. We tested whether pharmaceutical ads embedded in the prompts of 12 models from OpenAI, Anthropic, and Google shift drug recommendations across 258,660 API calls and four experiments probing distinct epistemic conditions. When two drugs were both guideline-appropriate, advertising shifted selection of the advertised drug by +12.7 percentage points (P < 0.001), with some model-scenario pairs shifting from 0% to 100%. Google models were the most susceptible (+29.8 pp), followed by OpenAI (+10.9 pp), while Anthropic models showed minimal change (+2.0 pp). When the advertised product lacked evidence or was clinically suboptimal, models resisted. This reveals a structured vulnerability: advertising does not override medical knowledge but fills the space where clinical evidence is underdetermined. An open-response sub-analysis (2,340 calls across three representative models) confirmed that advertising restructures free-text clinical reasoning: models echoed ad claims at 2.7 times the baseline rate while maintaining high stated confidence and rarely disclosing the ad. Susceptibility was provider-dependent (Google: +29.8 pp; OpenAI: +10.9 pp; Anthropic: +2.0 pp). Because this bias operates within clinically correct answers, it is invisible to accuracy-based evaluation, identifying a class of AI safety vulnerability that standard testing cannot detect.

BackgroundIn 2022, the Centers for Medicare & Medicaid Services (CMS) issued its final national coverage policy for aducanumab, a novel FDA-approved treatment for Alzheimers disease, deciding to Cover with Evidence Development (CED). CMS will thus only pay for the treatment of AD patients enrolled in an approved randomized controlled trial (RCT). We sought to understand whether, given current evidence, CED was best from a societal perspective. MethodsWe conducted a modeling-based expected value of sample information analysis to estimate the expected net decision-theoretic value of a further RCT to evaluate the clinical efficacy of high-dose (10 mg/kg) aducanumab and to determine what sized trial, if any, is optimal conditional on an initial decision to cover or not. We also evaluated the expected net benefit of the manufacturers proposed RCT ( ENVISION). We considered two post-trial decision criteria: cost-effectiveness given updated evidence ( efficiency) and does the new trial demonstrate a statistical significant (p<0.05) clinical benefit. Results were used to calculate the expected population net monetary benefit (NMB) of four decision alternatives (including CED) depending on an initial coverage and trial decision. We ranked alternatives and calculated the expected opportunity loss of a suboptimal decision. We used a societal perspective and focused on willingness-to-pay (WTP) values for a quality-adjusted life year (QALY) between $50K-$200K. We conducted scenario analyses using different assumptions about population size, efficacy, and drug cost. FindingsCMSs decision to not cover aducanumab avoids an expected societal loss (NMB) of $15B-$110B. Even an optimally designed RCT would confer no or negative decision-theoretic value for WTP[&le;]$100K or with statistical significance as a post-trial decision criterion, respectively, and thus denying coverage without a trial (rather than CED) is clearly preferable. For WTP=$150K (WTP=$200K) and assuming an efficiency criterion, CED with ENVISION or a similar trial is reasonable (decidedly optimal). The case for future research would become less ambiguous if the manufacturer again voluntarily dropped the price [&ge;]50%. InterpretationThe societal net value of a future trial (and thus CED) depends on how CMS would use the trial results to update its coverage decision and the WTP per QALY. Assuming CMS policymakers can avoid the pitfalls of a legal framework that limits their ability to consider costs in coverage decisions, the CED decision is at least reasonable, if not optimal, if a QALY is valued [&ge;]$150K.

BackgroundThe benefits from responsible sharing of individual-participant data (IPD) from clinical studies are well recognized, but stakeholders often disagree on how to align those benefits with privacy risks, costs, and incentives for clinical trialists and sponsors. Recently, the International Committee of Medical Journal Editors (ICMJE) required a data sharing statement (DSS) from submissions reporting clinical trials effective July 1, 2018. We set out to evaluate the implementation of the policy in three leading medical journals (JAMA, Lancet, and New England Journal of Medicine (NEJM)). MethodsA MEDLINE/PubMed search of clinical trials published in the three journals between July 1, 2018 and April 4, 2020 identified 487 eligible trials (JAMA n = 112, Lancet n = 147, NEJM n = 228). Two reviewers evaluated each of the 487 articles independently. Captured outcomes were declared data availability, data type, access, conditions and reasons for data (un)availability, and funding sources. Findings334 (68.6%, 95% confidence interval (CI), 64.1%-72.5%) articles declared data sharing, with non-industry NIH-funded trials exhibiting the highest rates of declared data sharing (88.9%, 95% CI, 80.0%-97.8) and industry-funded trials the lowest (61.3%, 95% CI, 54.3%-68.3). However, only two IPD datasets were actually deidentified and publicly available as of April 10, 2020. The remaining were supposedly accessible via request to authors (42.8%, 143/334), repository (26.6%, 89/334), and company (23.4%, 78/334). Among the 89 articles declaring to store IPD in repositories, only 17 articles (19.1%) deposited data, mostly due to embargo and regulatory approval. Embargo was set in 47.3% (158/334) of data-sharing articles, and in half of them the period exceeded 1 year or was unspecified. InterpretationMost trials published in JAMA, Lancet, and NEJM after the implementation of the ICMJE policy declared their intent to make clinical data available. However, a wide gap between declared and actual data sharing exists. To improve transparency and data reuse, journals should promote the use of unique pointers to dataset location and standardized choices for embargo periods and access requirements. All data, code, and materials used in this analysis are available on OSF at https://osf.io/s5vbg/.