Citations

PURPOSE Although most patients with follicular lymphoma (FL) can expect an indolent course, progressive lymphoma remains the primary cause of death during the first decade after diagnosis. Progression of disease within 24 months (POD24) of starting first-line (1L) immunochemotherapy defines a high-risk population with poor survival, but better risk stratification at diagnosis is needed. METHODS The FLIPI24 model was developed and internally validated to predict 24-month event rates using individual data from 4,485 patients treated with 1L immunochemotherapy from 10 observational cohorts of FL. Overall and cause-specific survival was further evaluated in FLIPI24 risk groups. External validation in the 1L immunochemotherapy setting was performed using the prospective observational Lymphoma Epidemiology of Outcomes (LEO) cohort (N = 565) and three randomized phase III trials (N = 3,192); extension to all patients with FL (any 1L therapy) was performed in the LEO cohort (N = 1,445) and its Molecular Epidemiology Resource subcohort (N = 1,074). RESULTS The FLIPI24 model uses age and four blood-based variables (hemoglobin, lactate dehydrogenase, beta-2 microglobulin, and WBC count). FLIPI24 showed consistent performance across validation and extension data sets, which was superior to existing prognostic tools. Across the four external immunochemotherapy validation data sets, patients with high-risk FLIPI24 (23%-32% of patients) had significantly higher 24-month event rates (22%-35%) and inferior 5-year overall survival (77%-83%) compared with patients with low-risk FLIPI24 (29%-31% of patients, 24-month event rates: 10%-12%; 5-year OS: 96%-97%). Results were consistent when evaluating lymphoma-related death and when extended to all patients with FL. CONCLUSION The FLIPI24 model robustly stratifies, at diagnosis, patients with FL at increased risk of lymphoma-related death versus patients with very low lymphoma-related mortality during the first decade after diagnosis. FLIPI24 can be used to enrich future clinical trial designs in newly diagnosed FL.

Abstract Anaplastic large cell lymphomas (ALCLs) are CD30+ T-cell lymphomas that share pathologic features but differ in presentation, outcome, and genetics. Current classification incorporates clinical presentation and anaplastic lymphoma kinase (ALK) status but inadequately addresses molecular heterogeneity and therapeutic vulnerabilities. We studied 689 patients with ALCL in the LLMPP (Lymphoma/Leukemia Molecular Profiling Project) and performed expert consensus review, genetic subtyping (ALK, DUSP22, TP63, and triple negative), and immunohistochemistry for phosphorylated STAT3Tyr705. RNA sequencing with unsupervised gene expression profiling in 393 patients identified 2 main molecular types of ALCL that could be predicted with 91% accuracy based on the presence (type I) or absence (type II) of phosphorylated STAT3Y705 expression. Type I ALCLs included ALK+ ALCL and a subset of triple-negative ALCLs (TN-I); type II ALCLs included tumors with DUSP22 and/or TP63 rearrangements and the remaining triple-negative ALCLs (TN-II). Type I ALCLs were enriched for JAK-STAT3, whereas type II ALCLs were enriched for non–tyrosine kinase pathways, particularly epigenetic regulators such as EZH2. Immunohistochemistry showed overexpression of EZH2 and its trimethylated substrate H3K27. Prognosis in systemic ALCL was favorable for DUSP22-rearranged ALCL (5-year overall survival, 95%) and ALK+ ALCL (88%), intermediate for triple-negative ALCL (TN-I, 52% and TN-II, 37%), and poor for TP63-rearranged ALCL (0%). We introduce an integrated molecular classification that preserves currently diagnosed ALCL entities but identifies 4 molecularly distinct ALK− ALCL subtypes (DUSP22-rearranged, TP63-rearranged, TN-I, and TN-II). This classification can be easily implemented on paraffin tissue in routine practice or clinical trials, and stratifies ALCL into diagnostically, prognostically, biologically, and potentially therapeutically relevant subtypes.

Objective: Generative artificial intelligence is rapidly evolving and is now being explored in health care to support patient and clinician education. This study evaluated the accuracy, completeness, and readability of four large language models (LLMs): ChatGPT 3.5, Gemini, ChatGPT 4.0, and OpenEvidence in answering questions about menopause and hormone therapy. Methods: A total of 35 questions (20 patient-level, 15 clinician-level) were entered into each LLM. OpenEvidence was only used for clinician-level questions. Four blinded expert reviewers rated responses as accurate and complete, accurate but incomplete, or inaccurate. Readability of patient-level responses was assessed using the Flesch Reading Ease Score (FRES) and word count. Analysis used ANOVA for readability, odds ratios for accuracy comparisons. Results: For patient-level questions, ChatGPT 3.5 achieved the highest accuracy (70%), followed by ChatGPT 4.0 (60%) and Gemini (30%); Gemini had significantly lower odds of accuracy compared with ChatGPT 3.5 (OR=0.18, 95% CI=0.05-0.71; P =0.014). FRES scores differed significantly ( P <0.001): Gemini scored 38.9±7.3 (“difficult”), ChatGPT 3.5 scored 31.0±11.2, and ChatGPT 4.0 scored 26.5±8.6 (both “very difficult”). For clinician-level questions, ChatGPT 4.0 achieved the highest accuracy (67%), followed by ChatGPT 3.5 and OpenEvidence (60% each) and Gemini (47%); no significant differences were observed among models (all P >0.05). Conclusion: LLMs demonstrated limited accuracy and frequent incorrect or incomplete responses to menopause-related queries, highlighting the need to improve model performance to ensure accurate and reliable information for both patients and clinicians.