Sciascia, Savino

Introduction: Large language models (LLMs) match or exceed physician accuracy on benchmark diagnostic tasks. Whether higher accuracy reflects human-like diagnostic behavior, which bears on how safely clinicians can supervise these systems, is rarely assessed. Objectives: To determine whether diagnostic accuracy and alignment with human reader responses move together across widely accessible LLMs under default, point-of-care conditions. Methods: This cross-sectional evaluation submitted the 29 most recent New England Journal of Medicine (NEJM) case challenges to 5 publicly accessible LLMs (GPT-5, Claude, Google Gemini, DeepSeek, and Grok) through each model’s web interface (November 5-8, 2025) at default settings, one query per case, no system prompt. Each case offered a clinical vignette and 6 candidate diagnoses; published human reader response distributions (4386-17 714 per case; median, 7732) were the comparator. Diagnostic accuracy against the published gold-standard diagnosis, compared with the human reader majority by the exact McNemar test; agreement by Cohen κ; alignment by a human-error similarity index (the share of model errors on cases the reader majority also missed) and by the Jensen-Shannon divergence (JSD) between each model’s response and the reader distribution; and ordinary least squares regression of per-case JSD on normalized reader-disagreement entropy. Results: Reader majority accuracy was 48.3% (95% CI, 31.4-65.6). All 5 reached or exceeded this value, from 51.7% for DeepSeek (95% CI, 34.4-68.6) to 82.8% for Grok (95% CI, 65.5-92.4); Grok (P = .02), GPT-5 (P = .004), and Gemini (P = .04) significantly exceeded the reader majority, whereas Claude and DeepSeek did not. Grok, the most accurate model, showed the weakest alignment (similarity index, 0.40; mean JSD, 0.50 bits), with 3 of 5 errors on cases readers answered correctly. GPT-5 showed the strongest alignment (index, 1.000), failing only on cases that also challenged readers. For 4 of 5 models, divergence rose with reader-disagreement entropy; DeepSeek was the exception. Conclusions and Relevance: In this evaluation, diagnostic accuracy and alignment with human reader responses diverged across LLMs. Because a model can outperform clinicians in aggregate yet fail where oversight is least likely to detect it, evaluation of clinical AI should weigh error structure alongside accuracy.

Background/Purpose APS ACTION Registry was created to study the natural course of antiphospholipid syndrome (APS) over 10 years in persistently antiphospholipid antibody (aPL) positive patients with or without systemic autoimmune rheumatic diseases (SARDs). Our primary objective was to compare the characteristics of aPL-positive patients with or without thrombocytopenia (TP) and/or autoimmune hemolytic anemia (AIHA). Methods The registry inclusion criteria are positive aPL based on the Revised Sapporo APS Classification Criteria, tested at least twice within 1 year prior to enrollment. For the primary comparison of demographic, clinical, and serologic characteristics in this retrospective study, we divided patients into two groups: TP/AIHA ever and never. Thrombocytopenia was defined as a platelet count of <100,000 x 10 9 /L tested twice at least 12 weeks apart, and AIHA was defined as anemia with hemolysis and a positive direct antiglobulin test (DAT). For the secondary analysis, we compared patients with TP versus AIHA, and the immunosuppressive use stratified by systemic lupus erythematosus (SLE) classification. Results As of April 2022, of 1,039 patients (primary aPL/APS: 618 [59%]; SLE classification: 334 [31%]) included in the registry, 228 (22%) had baseline (historical or current) TP and/or AIHA (TP only: 176 [17%]; AIHA only: 35 [3%], and both: 17 [2%]). Thrombocytopenia and/or AIHA was significantly associated with Asian race, SLE classification, cardiac valve disease, catastrophic/microvascular APS, triple aPL (lupus anticoagulant, anticardiolipin antibody, and anti-β 2 -glycoprotein-I antibody) positivity, and SLE-related serologic and inflammatory markers. When 101/618 (16%) primary aPL/APS patients and 101/334 (34%) SLE patients with TP and/or AIHA were compared, azathioprine and mycophenolate mofetil were more commonly reported in lupus patients, however corticosteroid, intravenous immunoglobulin, and rituximab use were similar between groups. Conclusion In our large multi-center international cohort of persistently aPL-positive patients, approximately one-fifth had active or historical TP and/or AIHA at registry entry; half of these patients had additional SLE. Cardiac valve disease, catastrophic/microvascular APS, and triple aPL-positivity were aPL-related clinical and laboratory manifestations associated with TP and/or AIHA, suggesting a more severe APS clinical phenotype in aPL-patients with TP and/or AIHA.

ObjectiveThe objective was to determine the mortality rate as well as the causes and predictors of death in antiphospholipid antibody (aPL)–positive patients with and without antiphospholipid syndrome (APS) classification.MethodsThe inclusion criterion for the multicenter international Antiphospholipid Syndrome Alliance for Clinical Trials and International Networking (APS ACTION) registry is positive aPLs according to the Revised Sapporo Classification Criteria tested within one year before enrollment. Patients are observed every 12 ± 3 months with clinical data and blood collection. For this prospective analysis, we first analyzed the causes of death for patients reported as “deceased.” Secondly, we analyzed risk factors for death using the adjusted Cox proportional hazards model and calculated survival probability using the Kaplan‐Meier model based on different age groups.ResultsOf 967 patients, 43 (5%) were deceased after a median follow‐up of 5.3 years. Based on the univariate analysis, deceased patients, compared to living patients, were more likely to be older and have a history of arterial thrombosis, catastrophic APS, concomitant systemic autoimmune diseases (SAIDs), and baseline cardiovascular disease (CVD) risk factors. Based on the Cox proportional hazards model adjusted for age and for each of the strongest predictors of death, arterial thrombosis (hazard ratio [HR] 2.94, 95% confidence interval [CI] 1.50–5.76), concomitant SAIDs (HR 2.97, 95% 1.56–5.63), and baseline any CVD risk factor (HR 2.43, 95% CI 1.05–5.71) were significantly associated with mortality.ConclusionIn our cohort of persistently aPL‐positive patients, the mortality rate was 5% after a median follow‐up of five years and was highest for patients ≥60 years old at registry entry. History of arterial thrombosis, concomitant SAIDs, and baseline any CVD risk factor independently predicted future death.

Objective To develop new antiphospholipid syndrome (APS) classification criteria with high specificity for use in observational studies and trials, jointly supported by the American College of Rheumatology (ACR) and EULAR. Methods This international multidisciplinary initiative included 4 phases: 1) Phase I, criteria generation by surveys and literature review; 2) Phase II, criteria reduction by modified Delphi and nominal group technique exercises; 3) Phase III, criteria definition, further reduction with the guidance of real‐world patient scenarios, and weighting via consensus‐based multicriteria decision analysis, and threshold identification; and 4) Phase IV, validation using independent adjudicators’ consensus as the gold standard. Results The 2023 ACR/EULAR APS classification criteria include an entry criterion of at least one positive antiphospholipid antibody (aPL) test within 3 years of identification of an aPL‐associated clinical criterion, followed by additive weighted criteria (score range 1–7 points each) clustered into 6 clinical domains (macrovascular venous thromboembolism, macrovascular arterial thrombosis, microvascular, obstetric, cardiac valve, and hematologic) and 2 laboratory domains (lupus anticoagulant functional coagulation assays, and solid‐phase enzyme‐linked immunosorbent assays for IgG/IgM anticardiolipin and/or IgG/IgM anti–β 2 ‐glycoprotein I antibodies). Patients accumulating at least 3 points each from the clinical and laboratory domains are classified as having APS. In the validation cohort, the new APS criteria versus the 2006 revised Sapporo classification criteria had a specificity of 99% versu s 86%, and a sensitivity of 84% versus 99%. Conclusion These new ACR/EULAR APS classification criteria were developed using rigorous methodology with multidisciplinary international input. Hierarchically clustered, weighted, and risk‐stratified criteria reflect the current thinking about APS, providing high specificity and a strong foundation for future APS research.