Jean-Philippe Jais

Summary: Using LYmphoma Study Association clinical trials combining 438 first‐line large B‐cell lymphomas (LBCL) patients >80 years old, we defined SENIOR‐IPI, an easily applicable prognostic index for this elderly population treated with an anti‐CD20‐miniCHOP (cyclophosphamide, doxorubicin, vincristine and prednisone)‐based regimen. SENIOR‐International Prognostic Index (SENIOR‐IPI) takes into account accessible variables (number of aaIPI predictors, lactate dehydrogenase > 3x Upper Limit of Normal (LDH >3× ULN), albumin 80 years old fit for age‐adapted immunochemotherapy than IPI, aaIPI and NCCN‐IPI.

Background: Objective Structured Clinical Examinations (OSCEs) are used as an evaluation method in medical education, but require significant pedagogical expertise and investment, especially in emerging fields like digital health. Large language models (LLMs), such as ChatGPT (OpenAI), have shown potential in automating educational content generation. However, OSCE generation using LLMs remains underexplored. Objective: This study aims to evaluate 3 GPT-4o configurations for generating OSCE stations in digital health: (1) standard GPT with a simple prompt and OSCE guidelines; (2) personalized GPT with a simple prompt, OSCE guidelines, and a reference book in digital health; and (3) simulated-agents GPT with a structured prompt simulating specialized OSCE agents and the digital health reference book. Methods: Overall, 24 OSCE stations were generated across 8 digital health topics with each GPT-4o configuration. Format compliance was evaluated by one expert, while educational content was assessed independently by 2 digital health experts, blind to GPT-4o configurations, using a comprehensive assessment grid. Statistical analyses were performed using Kruskal-Wallis tests. Results: Simulated-agents GPT performed best in format compliance and most content quality criteria, including accuracy (mean 4.47/5, SD 0.28; P=.01) and clarity (mean 4.46/5, SD 0.52; P=.004). It also had 88% (14/16) for usability without major revisions and first-place preference ranking, outperforming the other configurations. Personalized GPT showed the lowest format compliance, while standard GPT scored lowest for clarity and educational value. Conclusions: Structured prompting strategies, particularly agents’ simulation, enhance the reliability and usability of LLM-generated OSCE content. These results support the use of artificial intelligence in medical education, while confirming the need for expert validation.

BACKGROUND Objective Structured Clinical Examinations (OSCEs) are used as an evaluation method in medical education, but require significant pedagogical expertise and investment, especially in emerging fields like digital health. Large language models (LLMs), such as ChatGPT (OpenAI), have shown potential in automating educational content generation. However, OSCE generation using LLMs remains underexplored. OBJECTIVE This study aims to evaluate 3 GPT-4o configurations for generating OSCE stations in digital health: (1) standard GPT with a simple prompt and OSCE guidelines; (2) personalized GPT with a simple prompt, OSCE guidelines, and a reference book in digital health; and (3) simulated-agents GPT with a structured prompt simulating specialized OSCE agents and the digital health reference book. METHODS Overall, 24 OSCE stations were generated across 8 digital health topics with each GPT-4o configuration. Format compliance was evaluated by one expert, while educational content was assessed independently by 2 digital health experts, blind to GPT-4o configurations, using a comprehensive assessment grid. Statistical analyses were performed using Kruskal-Wallis tests. RESULTS Simulated-agents GPT performed best in format compliance and most content quality criteria, including accuracy (mean 4.47/5, SD 0.28; P=.01) and clarity (mean 4.46/5, SD 0.52; P=.004). It also had 88% (14/16) for usability without major revisions and first-place preference ranking, outperforming the other configurations. Personalized GPT showed the lowest format compliance, while standard GPT scored lowest for clarity and educational value. CONCLUSIONS Structured prompting strategies, particularly agents’ simulation, enhance the reliability and usability of LLM-generated OSCE content. These results support the use of artificial intelligence in medical education, while confirming the need for expert validation.

Diffuse large B-cell lymphoma (DLBCL) is the most frequent lymphoid malignancy affecting adults. The NF-κB transcription factor family is activated by 2 main pathways, the canonical and the alternative NF-κB activation pathway, with different functions. The alternative NF-κB pathway leads to activation of the transcriptionally active RelB NF-κB subunit. Alternative NF-κB activation status and its role in DLBCL pathogenesis remain undefined. Here, we reveal a frequent activation of RelB in a large cohort of DLBCL patients and cell lines, independently of their activated B-cell–like or germinal center B-cell–like subtype. RelB activity defines a new subset of patients with DLBCL and a peculiar gene expression profile and mutational pattern. Importantly, RelB activation does not correlate with the MCD genetic subtype, enriched for activated B-cell–like tumors carrying MYD88L265P and CD79B mutations that cooperatively activate canonical NF-κB, thus indicating that current genetic tools to evaluate NF-κB activity in DLBCL do not provide information on the alternative NF-κB activation. Furthermore, the newly defined RelB-positive subgroup of patients with DLBCL exhibits a dismal outcome after immunochemotherapy. Functional studies revealed that RelB confers DLBCL cell resistance to DNA damage–induced apoptosis in response to doxorubicin, a genotoxic agent used in the front-line treatment of DLBCL. We also show that RelB positivity is associated with high expression of cellular inhibitor of apoptosis protein 2 (cIAP2). Altogether, RelB activation can be used to refine the prognostic stratification of DLBCL and may contribute to subvert the therapeutic DNA damage response in a segment of patients with DLBCL.

Rituximab plus polychemotherapy is the standard of care in diffuse large B-cell lymphoma (DLBCL). GAINED, a randomized phase 3 trial, compared obinutuzumab to rituximab. Transplant-eligible patients (18-60 years) with an untreated age-adjusted International Prognostic Index (aaIPI) score ≥1 DLBCL were randomized (1:1) between obinutuzumab or rituximab and stratified by aaIPI (1; 2-3) and chemotherapy regimen (doxorubicin, cyclophosphamide, prednisone plus vindesine, bleomycin [ACVBP] or vincristine [CHOP]). Consolidation treatment was determined according to response to interim positron emission tomography (PET). Responders after cycle 2 and 4 (PET2−/PET4−) received immunochemotherapy. Responders after only cycle 4 (PET2+/4−) received transplantation. The primary objective was an 8% improvement (hazard ratio [HR] = 0.73; 80% power; α risk, 2.5%; 1-sided) in 2-year event-free survival (EFS) in the obinutuzumab arm. From September 2012, 670 patients were enrolled (obinutuzumab, n = 336; rituximab, n = 334). A total of 383 (57.2%) were aaIPI 2-3, 339 (50.6%) received CHOP. Median follow-up was 38.7 months. The 2-year EFS was similar in both groups (59.8% vs 56.6%; P = .123; HR = 0.88). The 2-year PFS in the whole cohort was 83.1% (95% confidence interval, 80% to 85.8%). PET2−/4− and PET2+/4− had similar 2-year progression-free survival (PFS) and overall survival (OS): 89.9% vs 83.9% and 94.8% vs 92.8%. The 2-year PFS and OS for PET4+ patients were 62% and 83.1%. Grade 3-5 infections were more frequent in the obinutuzumab arm (21% vs 12%). Obinutuzumab is not superior to rituximab in aaIPI ≥1 DLBCL transplant-eligible patients. This trial was registered at www.clinicaltrials.gov as #NCT01659099.