Yunus, Furhan

21 Background: Large language models (LLMs) are becoming an increasing source of medical information for both patients and providers. Clinical decision making is occasionally complex and may use factors outside of traditional guidelines. We conducted a comparative performance evaluation of frontier models on standardized genitourinary (GU) oncology knowledge assessments derived from the American Society of Clinical Oncology Self-Evaluation Program 2025 (ASCO SEP 2025) in addition to challenging questions developed by GU oncology experts. Methods: Three frontier LLMs were selected: ChatGPT-5.2 (OpenAI, San Francisco, CA, USA), Claude Opus 4.6 (Anthropic, San Francisco, CA, USA), and Gemini 3 Pro (Google DeepMind, Mountain View, CA, USA). Model selection was based on publicly reported performance on “Humanity’s Last Exam,” a benchmark of advanced scientific reasoning (Scale AI leaderboard Jan-Feb 2026). The OpenEvidence (OpenEvidence Inc., MA, USA) search engine was also selected for evaluation. All models were asked 145 ASCO SEP 2025 GU oncology questions and 12 real world challenging scenarios using an uniform prompt. Subject matter experts peer reviewed LLM generated responses to challenging questions and credit was given only for correct rationale and option selection. Results: All models demonstrated high and comparable accuracy on Standardized knowledge assessment, with Gemini 3 Pro leading numerically at 97.24% (95% CI, 93.1% – 99.0%), OpenEvidence followed with an accuracy of 94.48% (95% CI, 89.5% – 97.3%), while Chat GPT -5.2 and Claude OPUS 4.6 scored 92.41% (95%, CI 86.9% – 95.7%) and 79.31% (95% CI, 71.9% – 85.1%) respectively. On challenging real world scenarios, ChatGPT-5.2 demonstrated the highest numerical accuracy at 100% (95% CI, 75.8% – 100%), Gemini 3 Pro and Open Evidence scored 75.0% (95% CI, 46.8%–91.1%) each, followed by Claude Opus 4.6 at 66.7% (39.1–86.3). Conclusions: All current frontier AI models appear to provide accurate medical information and rationale on a large majority of guideline based medical knowledge assessments for GU oncology. Frontier AI models also appear to have emerging competence in highly challenging real-world scenarios. AI based tools may become a valuable and reliable tool for patients and providers. Comparative LLM performance. Model ASCO SEP 2025 % (95% CI) Real World Scenarios % (95% CI) Gemini 3 Pro 97.24% (93.1–99.0) 75.0% (46.8–91.1) OpenEvidence 94.48% (89.5–97.3) 75.0% (46.8–91.1) ChatGPT-5.2 92.41% (86.9–95.7) 100% (75.8–100) Claude Opus 4.6 79.31% (71.9–85.1) 66.67% (39.1–86.3)

Abstract Background: There remains an opportunity for new therapies to improve long-term outcomes for patients with low-tumor burden follicular lymphoma (FL). Mosunetuzumab is an off-the-shelf CD20xCD3 bispecific antibody, approved as an intravenous therapy for patients with relapsed/refractory FL after ≥2 lines of systemic therapy. Mosunetuzumab can be administered in the outpatient setting as a fixed-duration regimen. In an open-label, multicenter, Phase II study conducted mainly across US community sites (MorningSun; NCT05207670), mosunetuzumab administered subcutaneously (SC) induced high response rates with a manageable safety profile in patients with previously untreated low-tumor burden FL (Burke et al. ASH 2024). Here we present updated efficacy and safety data for mosunetuzumab SC in patients with previously untreated, low-tumor burden FL. Methods: Patients had untreated Grade 1 or 2 low-tumor burden FL per Groupe d'Etude des Lymphomes Folliculaires criteria, Ann Arbor stage III/IV disease, and Eastern Cooperative Oncology Group performance status (ECOG PS) of 0–2. Mosunetuzumab SC was administered with step-up dosing in Cycle 1 (Day [D]1, 5mg; D8, 45mg; D15, 45mg), then 45mg on D1 of each 21-day cycle. Hospitalization was not mandatory. Patients with a complete metabolic response after 8 cycles could complete therapy and patients with a partial metabolic response or stable disease continued treatment for up to 17 cycles. The primary endpoint was progression-free survival (PFS) at 24 months. Key secondary endpoints included objective response rate (ORR), time to response (TTR), overall survival (OS), duration of response (DOR), duration of complete response (DOCR), and safety. Exploratory analyses assessed pharmacodynamic changes in B-cell subsets. Here, we report interim data. Results: As of February 10, 2025 (clinical cut-off date), 101 patients were enrolled from US community (n=56) and academic (n=45) centers; 52 patients had completed treatment, 21 had discontinued (mainly due to adverse events [AEs; n=6] and progressive disease [n=5]), and 28 were ongoing on treatment. Median age was 65 years (range: 34–82). Most patients had an ECOG PS of 0 (81.2%) or 1 (18.8%) and 43 patients (42.6%) had extranodal involvement. Median duration of follow-up was 13.8 months (95% confidence interval [CI]: 11.5–16.9) and median number of cycles received was 10 (range: 3–17). The ORR was 98.0% (95% CI: 93.0–99.8) and the complete response rate was 89.1% (95% CI: 81.3–94.4). Median TTR was 2.8 months (range: 1.4–8.3). Landmark DOR and DOCR rates at 12 months were 91.8% (95% CI: 81.5–96.5) and 95.7% (95% CI: 82.6–99.0), respectively. The 12-month PFS and OS rates were 92.0% (95% CI: 83.8‒96.1) and 99.0% (95% CI: 93.2‒99.9), respectively. The most common AE was injection-site reaction (ISR, 81.2%; Grade 1, 74.3%; Grade 2, 5.9%; Grade 3, 1.0%). Grade 3/4 AEs were reported in 51.5% of patients; the most common (≥10%) events were cytopenias (36.6%) and infections (11.9%). Cytokine release syndrome (CRS) was reported in 48.5% of patients (Grade 1, 40.6%; Grade 2, 7.9%). Overall,infections occurred in 63.4% of patients (Grade 1, 7.9%; Grade 2, 42.6%; Grade 3, 11.9%; Grade 5, 1.0%) with the most common (≥10%) being upper respiratory tract infection (21.8%), COVID-19 (17.8%), sinusitis (13.9%), and pneumonia (11.9%). At the time of data cut-off, ISR, CRS, and infection AEs were resolved at rates of 98.2%, 100%, and 89.0%, respectively. Two patients had a Grade 5 AE (sepsis, n=1; unknown cause, n=1).No immune effector cell-associated neurotoxicity syndrome events were reported. B-cell depletion was observed following the first dose of mosunetuzumab, with subsequent B-cell recovery observed within six months of treatment completion. Conclusions: Fixed-duration mosunetuzumab SC demonstrates high efficacy rates with deep and sustained responses in patients with previously untreated, low-tumor burden FL. The manageable safety profile supports ease of outpatient administration without mandatory hospitalization, including in community oncology practices. Additionally, exploratory analyses demonstrating B-cell recovery after treatment completion also substantiate the use of mosunetuzumab SC as a fixed-duration regimen. These data support further evaluation of mosunetuzumab SC in patients with previously untreated, low-tumor burden FL.