Citations

Objective Meaningful assessments of how large language models (LLMs) incorporate clinical guidelines require large-scale testing over many queries. Here, we evaluate the prevalence of clinical guideline omissions and hallucinations in a large sample of diagnostic LLM outputs. Methods We used simulated case vignettes and zero-shot prompting to generate diagnostic outputs and rationales from GPT-4.1 and DeepSeek-V3. English case vignettes were created for hypercholesterolaemia and type-2 diabetes mellitus. Each vignette contained identical medical information, while sociodemographic characteristics varied in terms of sex, ethnicity and location. We calculated the prevalence of existing and hallucinated clinical guidelines in LLM outputs across disease, LLM and sociodemographic characteristics. Results We analysed a total of 12 197 LLM outputs, which quantifies three hazard areas: omissions (up to 97% for DeepSeek-V3 and 46% for GPT-4.1), hallucinations (up to 9%) and inconsistencies (guideline citation rate ranging from 0% to 78.39% across sociodemographic vignettes). Omission and hallucination rates were generally similar across vignettes with different sex or ethnicity data, yet were particularly sensitive to patient location. Discussion This study highlights significant variability in clinical guideline prediction across two different diseases, three different sociodemographic variables and two LLMs, even when the LLMs were instructed by identical prompts, establishing clinical guideline prediction in LLM outputs as a stochastic event. Conclusion The stochastic nature of LLMs creates a unique challenge for evidence generation and clinical deployment. Being able to measure and capture this stochasticity within high-quality research designs will be a prerequisite to advancing the responsible deployment of LLMs in healthcare.

Importance Accurate prediction of patient mortality after kidney transplant is an unmet need. Objective To develop and validate an integrative prediction model that predicts short- and long-term patient mortality for kidney recipients. Design, Setting, and Participants This international cohort study included patients who underwent transplants between 2004 and 2023 from 14 academic medical centers from Europe and the US. The derivation cohort consisted of 1566 consecutive adult kidney recipients in a deeply phenotyped cohort prospectively recruited in 1 French center between 2004 and 2014. External validation cohorts consisted of 10 951 kidney recipients from 7 centers in France, 3 centers in Europe, and 3 centers in North America who underwent kidney transplants between 2006 and 2023. Data were analyzed from January 2023 to June 2025. Main Outcome and Measures All-cause mortality was assessed, and 121 candidate prognostic factors, comprising demographics and clinical, biological, imaging, and immunological parameters, were collected. Results Among 12 517 kidney transplant recipients, including 1566 in the derivation cohort (mean [SD] age, 50.05 [14.31] years; 942 male [60.15%]) and 10 951 in validation cohorts (mean [SD] age, 53.32 [13.97] years; 6766 male [61.78%]), 2486 patients (19.9%) died after a median (IQR) follow-up of 5.08 (2.97-7.00) years. Fourteen prognostic factors (including clinical, biological, and imaging risk factors) were independently associated with patient death (eg, patient age: hazard ratio per 1-year increase in age, 1.07 [95% CI, 1.06-1.08]; P < .001) and were combined into a risk prediction model (mBox). The model exhibited accurate calibration and discrimination in the derivation cohort, with C statistics of 0.82 (95% CI, 0.77-0.87) and 0.80 (95% CI, 0.78-0.82) at 1 and 10 years, respectively, after the transplant. Abbreviated models were developed and validated to ensure model generalizability. Performance of abbreviated models was confirmed in external validation cohorts from France (C statistic, 0.76 [95% CI, 0.73-0.78]), Europe (C statistic, 0.74 [95% CI, 0.72-0.76]), the US (C statistic, 0.74 [95% CI, 0.70-0.78]), the Greater Paris University Hospital database (C statistic, 0.79 [95% CI, 0.77-0.81]), and the University of California at San Francisco database (C statistic, 0.70 [95% CI, 0.65-0.74]). The model was also validated in a series of subpopulations (eg, women vs men: C statistic, 0.81 [95% CI, 0.77-0.84] vs 0.79 [95% CI, 0.77-0.82]) and clinical scenarios (eg, before COVID-19 era: C statistic, 0.79 [95% CI, 0.77-0.81]) with good and stable performance. Conclusions and Relevance In this study, an accurate prediction model for mortality among kidney recipients, computable at the time of transplant, was developed and externally validated. Results suggest that this model may help stratify patient risk of death, allowing for improved medical decisions.

Abstract Pancreatic cancer disproportionately affects Black individuals in the United States, but they have limited representation in genetic studies of pancreatic ductal adenocarcinoma (PDAC). To address this gap, we performed admixture mapping and genome-wide association analysis (GWAS) in genetically inferred African ancestry individuals (1,030 cases and 889 controls). Admixture mapping identified three regions with a significantly higher proportion of African ancestry in cases compared to controls (5q33.3, 10p1, 22q12.3). GWAS identified a genome-wide significant association at 5p15.33 ( CLPTM1L , rs383009:T>C, T Allele Frequency=0.51, OR:1.45, P value=1.24×10 -8 ), a locus previously associated with PDAC. Known loci at 5p15.33, 7q32.3, 8q24.21 and 7q25.1 also replicated (P value <0.01). Multi-ancestral fine-mapping identified two potential causal SNPs (rs3830069 and rs2735940) at 5p15.33. Collectively these findings identified novel PDAC risk loci and expanded our understanding of this deadly cancer in underrepresented populations, emphasizing the multifactorial nature of PDAC risk including inherited genetic and non-genetic factors. Statement of Significance To understand how genetic variation contributes to PDAC risk in Black people in North American, we studied individuals of genetically-inferred African ancestry. We identified novel risk loci and differences in the contribution of known loci. This demonstrates that ancestry-informed genetic analyses improve our understanding of PDAC risk and enhances discovery.

Abstract The hallmark of heparin-induced thrombocytopenia (HIT) is the presence of immunoglobulin G (IgG) antibodies against platelet factor 4/heparin (PF4/H) complexes, typically detected by PF4/H enzyme-linked immunosorbent assay (ELISA); thus, negative ELISA results are commonly used to exclude this diagnosis. Here, we report a prevalent yet previously unrecognized subset of antibodies that are undetectable by PF4/H ELISA (ELISA−) but activate platelets in the PF4-dependent P-selectin expression assay (PEA+). In 11 patients with clinically confirmed HIT who tested positive in both PF4/H ELISA and platelet activation assays, ELISA−PEA+ antibodies accounted for 65% ± 19% of total platelet-activating IgG activity and coexisted with ELISA+PEA+ antibodies. Consistent with this finding, single-cell cloning from 7 patients with HIT identified 23 PEA+ antibody-producing B-cell clones, of which 17 were ELISA−, outnumbering the ELISA+ clones. Functionally, ELISA−PEA+ antibodies closely resembled ELISA+PEA+ antibodies: platelet binding and activation required exogenous PF4 and were inhibited by FcγRIIA blockade, high-dose heparin, or Fab fragments made from ELISA+PEA+ antibodies. Importantly, these antibodies induced thrombocytopenia in a humanized mouse model of HIT. Despite lacking PF4/H reactivity in ELISAs, they recognize PF4 on platelets and showed no appreciable binding to neutrophil-activating peptide-2, interleukin-8, or PF4 alone. Structurally, these antibodies were heterogeneous, with a subset sharing heavy-chain features with ELISA+PEA+ antibodies. Collectively, our findings demonstrate that ELISA−PEA+ antibodies are a common, previously unrecognized feature of HIT, with functional relevance, supporting the possibility that they play an important, perhaps even central, role in HIT pathogenesis. Defining their prevalence, kinetics, and clinical impact deserves high priority for further investigation.

Objective This study aims to identify factors associated with patient global assessment (PtGA) and physician global assessment (PhGA) and discordance between them in systemic sclerosis (SSc). Methods Data from adults with early SSc (<5 years) from the Collaborative National Quality and Efficacy Registry were included. PtGA and PhGA (0–10 scale), clinical evaluations, and patient‐reported outcomes (PROs) were collected every six months. Multivariable mixed‐effects linear regression identified factors associated with PtGA and PhGA using (1) clinical variables and (2) clinical variables plus PROs. Relative weight analysis (RWA) determined the relative importance of each variable. Discordance (≥2 points between PtGA and PhGA) was assessed using multinomial mixed‐effects logistic regression. Results Among 956 patients (83% women, 33% limited disease), mean PtGA and PhGA at enrollment were 4.2 (SD 2.6) and 3.4 (SD 2.0), respectively ( P < 0.001). RWA of clinical variables identified modified Rodnan skin score (mRSS) and New York Heart Association (NYHA) functional class as most influential for both global assessments. After including PROs, PtGA was most influenced by measures of pain, skin symptoms, and physical function. Discordance occurred in 53% of patients (35% PtGA worse, 18% PhGA worse). Worse PtGA was associated with higher overall pain and discomfort. Worse PhGA was associated with higher mRSS, worse NYHA class, higher pain interference, and lower diffusing capacity of the lung. Conclusion Discordance between PtGA and PhGA occurs commonly, highlighting the need for comprehensive symptom management and measurement of disease burden in this complex disease. In SSc, differences in PtGA and PhGA reflect dissimilar weighting of data elements.

We previously reported initial results of the pivotal phase II trial of atezolizumab for patients with alveolar soft part sarcoma (ASPS; ClinicalTrials.gov identifier: NCT03141684 ). Here, we report on three additional years of observation. Fifty-three patients with ASPS received atezolizumab. Median duration of response increased to 37.0 months. Objective response rate (ORR) and median progression-free survival (mPFS) remained essentially as previously reported (35.8% [95% CI, 23.1 to 50.2] and 20.8 months [IQR, 7.6-not reached], respectively). ASPSCR1::TFE3 fusion type was determined for 47/53 patients; ORR and mPFS were higher among the 41 patients expressing type 1 (43.9% [95% CI, 28.5 to 60.2] and 28.3 months [IQR, 9.2-not reached], respectively) than the six patients expressing type 2 (0% [95% CI, 0 to 45.9] and 7.5 months [IQR, 3.9-not reached], respectively, PFS HR, 3.2 [95% CI, 1.01 to 10.2]). Eleven patients chose a per-protocol drug holiday (range, 3.5-26.4 months) after ≥2 years of treatment; two experienced disease progression during the holiday. Nine eligible patients elected to receive bevacizumab plus atezolizumab after progressing on monotherapy; ORR was 0% and mPFS was 18.5 months (IQR, 7.9-21.1) in this small cohort. Long-term results support using atezolizumab to treat ASPS, even for several years; a drug holiday with careful monitoring may be an option for some patients.

Abstract Oncology Advanced Practice Providers (APPs), including Advanced Practice Registered Nurses, Physician Assistants, and Clinical Pharmacists (CPs), play a vital role in delivering high-quality, patient-centered cancer care across the United States. Despite their widespread presence in oncology practices, APPs remain underutilized in cancer clinical research, representing a missed opportunity to expand trial access and improve patient outcomes. Clinical trials are essential to advancing oncology care, yet participation remains critically low among adult patients. Given Oncology APPs’ and CPs’ central role in cancer care delivery, their meaningful engagement in oncology clinical trial research is imperative as the standard of oncology care. This position article, with contributions from 5 professional societies, presents solutions and resources to barriers limiting APPs’ and CPs’ involvement in cancer clinical research, including gaps in education, role expectations, limited protected time, restrictive policies, insufficient financial support, and underrecognition of contributions. Integrating APPs and CPs more fully into the clinical research enterprise is essential to improving trial access, patient outcomes, closing equity gaps, and accelerating innovation in oncology care.

Background: The most critical exam components in patients seeking rhinoplasty have not been established. Objective: To identify concordance among experienced rhinoplasty surgeons regarding the most important exam features for various deformities. Methods: An online questionnaire was distributed to board-certified facial plastic and plastic surgeons experienced in rhinoplasty, with seven nasal deformity categories: dorsal hump, crooked nose, boxy tip, pinched tip, droopy tip, saddle nose, and nonallergic nasal obstruction. For each, respondents were asked to rank the three most important anatomic regions to examine from the following: nasal bones, midvault cartilage, skin, radix, facial profile, lip, tip, septum, internal nasal valve, nostrils, inferior and middle turbinates. For each ranked region, up to three important features were chosen. Normalized Attention Scores (NAS) were calculated to determine agreement amongst surgeons. Results: A total of 13 facial plastic and 8 plastic surgeons were invited to participate and completed the questionnaire. All cases had high agreement demonstrated by NAS, with the dorsal hump case exhibiting the greatest variability in responses. Conclusions: In this survey of 21 rhinoplasty surgeons, there was general agreement regarding the most relevant anatomical structures and features, which may provide a conceptual foundation for identifying key exam features for seven common complaints in patients seeking rhinoplasty.

Treatment guidelines for Clostridioides difficile infection (CDI) have been published by infectious disease and gastroenterology professional societies; however, adherence in clinical practice remains poorly characterized, particularly for recurrent disease. We conducted a retrospective chart review of 100 patients with CDI (350 episodes: 115 initial, 235 recurrent) referred to a tertiary complicated CDI clinic between 2018 and 2023. Guideline adherence was assessed by comparing treatment with IDSA/SHEA and ACG recommendations, and referring diagnoses were compared with final specialist diagnoses. Guideline adherence was significantly higher in initial compared to recurrent episodes (70.4% vs. 41.3%, p < 0.0001). Among guideline non-adherent recurrent episodes, 51.3% used standard antibiotic regimens inappropriate for the recurrence tier. Specialist review reclassified 12.0% of episodes, with colonization increasing from 2.6% to 8.9%. Misdiagnosed colonization cases had a 6.2-fold higher treatment failure rate than confirmed CDI (39.3% vs. 6.3%, p < 0.0001). Guideline non-adherence showed a non-significant trend toward treatment failure (10.0% vs. 6.7%, p = 0.31). Guideline adherence for recurrent CDI is inadequate in pre-referral settings, and diagnostic misclassification is common. Early specialist involvement may improve both diagnostic accuracy and treatment appropriateness for patients with recurrent CDI.