Citations

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 Hematoxylin and eosin (H&E) slides are routinely available in cancer care, but molecular profiling often requires additional tissue processing and turnaround time. We introduce HistoMethyl, a DNA methylation-aware pathology foundation model that aligns whole-slide histopathology with matched genome-scale methylation beta-value profiles during pre-training while requiring only an H&E slide at inference. We evaluated HistoMethyl across four task groups: gene mutation prediction, morphology-associated classification, overall survival prediction, and direct DNA methylation beta-value recovery. Evaluation spanned TCGA cross-validation, a disease-held-out lower-grade glioma cohort, and external cohort validation in CPTAC glioblastoma and SurGen rectal adenocarcinoma, totaling 14 cancer cohorts and 81 gene mutation tasks. The best-performing configuration improved mean mutation AUROC by 5.03 percentage points. By converting DNA methylation supervision into an H&E-only representation, His-toMethyl could support an early molecular triage layer that helps prioritize cases for confirmatory sequencing, methylation profiling, immunohistochemistry, and molecular tumor board review. These image-only predictions are intended to accelerate downstream molecular testing and tissue allocation while leaving final diagnosis and treatment selection anchored in validated molecular assays.

Background Patients with paraneoplastic syndromes (PNS) are excluded from clinical trials involving immune checkpoint inhibitors (ICIs) due to safety concerns. Moreover, real-world data on efficacy and safety is scarce. Methods In this retrospective study, data were collected on patients with PNS and solid tumors receiving ICI between 2015 and 2022 at nine institutions. Patients were classified into: Cohort 1 (pre-existing PNS before ICI initiation), cohort 2 (PNS during ICI treatment), and cohort 3 (PNS after ICI discontinuation). Patients with metastatic non-small cell lung cancer (NSCLC) (mNSCLC) from cohort 1 were matched to patients who were PNS-free at each institution up to a 1:3 ratio for age, sex, type of ICI, use of concurrent chemotherapy, and number of lines of systemic therapy prior to ICI initiation. Kaplan-Meier method was used to assess overall survival (OS) and time-to-next treatment (TTNT). Results Among 109 patients with PNS treated with ICIs, median age at ICI initiation was 67 years (IQR: 58–74). The most represented cancer type was NSCLC (n=39, 36%). In cohort 1 (n=55), PNS exacerbations occurred in 16 (29%) patients with median time to exacerbation after ICI of 1.1 months (IQR: 0.7–3.3). Exacerbation or de novo PNS prompted temporary/permanent interruption of ICIs in 14 (13%) patients. For cohort 2 (n=16), median time between ICI initiation and de novo PNS was 1.2 months (IQR: 0.4–3.5). Treatment-related adverse events (trAEs) occurred in 43 (39%) patients. Grade ≥3 trAEs occurred in 18 (17%) patients. PNS-directed immunosuppressive therapy was required in 55 (50%) patients. We matched 18 patients with mNSCLC and PNS (cohort 1) to 40 without PNS, treated with ICIs. There was no significant difference in OS or TTNT between patients with mNSCLC with and without PNS, although a trend was seen towards worse outcomes in patients with PNS. TrAEs occurred in 6/18 (33%) and 14/40 (35%), respectively. Grade ≥3 trAEs occurred in 4 (22%) patients with PNS and 7 (18%) patients without PNS. Conclusions Exacerbations of pre-existing PNS occurred in 29% of patients treated with ICIs and both exacerbations and de novo PNS occur early in the ICI course. TrAE from ICIs were similar between patients with and without PNS. Our data suggest that pre-existing PNS should not preclude consideration of ICI therapy although patients may not derive the same clinical benefit compared with patients without PNS.

LBA531 Background: Muscle-invasive urothelial carcinoma (MIUC) is an aggressive disease with high relapse rates. Neoadjuvant platinum-based chemotherapy (NAC) is the standard of care in patients (pts) who are cisplatin (Cis)-eligible. However, many pts are Cis-ineligible or have persistent muscle-invasive disease after NAC and surgery. We evaluated pembrolizumab (Pembro) as adjuvant therapy in pts with high-risk MIUC following surgical resection. Methods: The AMBASSADOR study is an open-label, randomized, phase III trial that enrolled pts with histologically confirmed MIUC of the bladder, upper tract, or urethra who (1) had ≥ pT2 and/or pN+ or margins+ at surgery (radical cystectomy, nephrectomy, nephroureterectomy, or ureterectomy) following NAC or (2) ≥ pT3 and/or pN+ or margins+ at surgery without NAC and were either Cis-ineligible or declined adjuvant Cis-based therapy. Pts were registered ≥4 to 16 weeks after surgery and randomized 1:1 to receive Pembro 200 mg every 3 weeks for 1 year or observation (Obs). Randomization was stratified by pathologic stage, centrally tested PD-L1 status, and prior NAC. The dual primary endpoints were disease-free survival (DFS) and overall survival (OS) with a target sample size of 739 pts and required 387 DFS and 320 OS events for final analysis. Secondary objectives included evaluation of DFS and OS in PD-L1-positive and -negative pts and assessing safety. Here we present the interim DFS and OS results. Results: Between 09/2017 and 08/2021 a total of 702 pts were randomized prior to early closure due to US FDA approval of nivolumab for MIUC pts: 354 pts to Pembro and 348 pts to Obs; 13.0% vs 21.6% withdrew from study without event in the Pembro vs Obs arms, respectively. 74 pts (21%) in the Obs arm received an immune checkpoint inhibitor. Median follow-up was 22.3 months (mos) for DFS and 36.9 mos for OS. The DFS endpoint (based on 319 events needed for interim analysis) crossed the efficacy boundary. Median DFS was 29.0 months (95% confidence interval (CI) 21.8‒not evaluable (NE)) with Pembro and 14.0 months (95% CI 9.7‒20.20) with Obs (hazard ratio (HR) 0.69 [95% CI 0.55–0.87]; p=0.0013). At the interim analysis (n = 257 events), median OS was 50.9 months (95% CI 43.9‒NE) with Pembro and 55.8 months (95% CI 53.3‒NE) with Obs (HR 0.98 [95% CI 0.76–1.26]; p=0.88). Grade 3+ adverse events occurred in 48.4% and 31.8% of pts in the Pembro and Obs arms, respectively. Conclusions: Adjuvant Pembro demonstrated a statistically significant and clinically meaningful improvement in DFS vs Obs for high-risk MIUC pts after radical surgery. The OS endpoint may have been impacted by pts on the Obs arm receiving a checkpoint inhibitor. Pembro was tolerable with no new safety signals. These results support adjuvant Pembro as a new therapeutic option for pts with MIUC with high risk for recurrence. Additional follow-up is ongoing for the final DFS/OS, PD-L1 subgroups, and ctDNA analyses. Clinical trial information: NCT03244384 .

Abstract Background Treatment options for penile squamous cell carcinoma are limited. We sought to investigate clinical outcomes and safety profiles of patients with penile squamous cell carcinoma receiving immune checkpoint inhibitors. Methods This retrospective study included patients with locally advanced or metastatic penile squamous cell carcinoma receiving immune checkpoint inhibitors between 2015 and 2022 across 24 centers in the United States, Europe, and Asia. Overall survival and progression-free survival were estimated using the Kaplan-Meier method. Objective response rates were determined per Response Evaluation Criteria in Solid Tumours 1.1 criteria. Treatment-related adverse events were graded per the Common Terminology Criteria for Adverse Events, version 5.0. Two-sided statistical tests were used for comparisons. Results Among 92 patients, 8 (8.7%) were Asian, 6 (6.5%) were Black, and 24 (29%) were Hispanic and/or Latinx. Median (interquartile range) age was 62 (53-70) years. In all, 83 (90%) had metastatic penile squamous cell carcinoma, and 74 (80%) had received at least second-line treatment. Most patients received pembrolizumab monotherapy (n = 26 [28%]), combination nivolumab-ipilimumab with or without multitargeted tyrosine kinase inhibitors (n = 23 [25%]), or nivolumab (n = 16 [17%]) or cemiplimab (n = 15 [16%]) monotherapies. Median overall and progression-free survival were 9.8 months (95% confidence interval = 7.7 to 12.8 months) and 3.2 months (95% confidence interval = 2.5 to 4.2 months), respectively. The objective response rate was 13% (n = 11/85) in the overall cohort and 35% (n = 7/20) in patients with lymph node–only metastases. Visceral metastases, Eastern Cooperative Oncology Group (ECOG) performance status of 1 or higher, and a higher neutrophil/lymphocyte ratio were associated with worse overall survival. Treatment-related adverse events occurred in 27 (29%) patients, and 9.8% (n = 9) of the events were grade 3 or higher. Conclusions Immune checkpoint inhibitors are active in a subset of patients with penile squamous cell carcinoma. Future translational studies are warranted to identify patients more likely to derive clinical benefit from immune checkpoint inhibitors.

PURPOSE Compared with people living without HIV (PWOH), people living with HIV (PWH) and cancer have traditionally been excluded from immune checkpoint inhibitor (ICI) trials. Furthermore, there is a paucity of real-world data on the use of ICIs in PWH and cancer. METHODS This retrospective study included PWH treated with anti–PD-1- or anti–PD-L1-based therapies for advanced cancers. Kaplan-Meier method was used to estimate overall survival (OS) and progression-free survival (PFS). Objective response rates (ORRs) were measured per RECIST 1.1 or other tumor-specific criteria, whenever feasible. Restricted mean survival time (RMST) was used to compare OS and PFS between matched PWH and PWOH with metastatic NSCLC (mNSCLC). RESULTS Among 390 PWH, median age was 58 years, 85% (n = 331) were males, 36% (n = 138) were Black; 70% (n = 274) received anti–PD-1/anti–PD-L1 monotherapy. Most common cancers were NSCLC (28%, n = 111), hepatocellular carcinoma ([HCC]; 11%, n = 44), and head and neck squamous cell carcinoma (HNSCC; 10%, n = 39). Seventy percent (152/216) had CD4+ T cell counts ≥200 cells/µL, and 94% (179/190) had HIV viral load <400 copies/mL. Twenty percent (79/390) had any grade immune-related adverse events (irAEs) and 7.7% (30/390) had grade ≥3 irAEs. ORRs were 69% (nonmelanoma skin cancer), 31% (NSCLC), 16% (HCC), and 11% (HNSCC). In the matched mNSCLC cohort (61 PWH v 110 PWOH), 20% (12/61) PWH and 22% (24/110) PWOH had irAEs. Adjusted 42-month RMST difference was –0.06 months (95% CI, –5.49 to 5.37; P = .98) for PFS and 2.23 months (95% CI, –4.02 to 8.48; P = .48) for OS. CONCLUSION Among PWH, ICIs demonstrated differential activity across cancer types with no excess toxicity. Safety and activity of ICIs were similar between matched cohorts of PWH and PWOH with mNSCLC.

531 Background: Neoadjuvant cisplatin-based chemotherapy (NAC) is established in the management of MIBC. Pathologic complete response (pCR, ypT0N0) and downstaging (<ypT2N0) are associated with long term survival in patients with MIBC. Several trials have evaluated neoadjuvant immunotherapy (NAI) and chemoimmunotherapy (NACI) for MIBC. We aimed to conduct a meta-analysis of available trials to compare the incidence of pathological response with NAC, NAI and NACI in MIBC patients undergoing radical cystectomy. Methods: Phase II/III clinical trials assessing neoadjuvant systemic therapies (NAC, NAI, or NACI) in patients with MIBC undergoing radical cystectomy were identified through a systematic search of MEDLINE and EMBASE. Outcomes of interest included pCR and pathological downstaging. A DerSimonian and Laird random-effect meta-analysis was conducted to estimate pooled incidence of events using the Freeman-Tukey transformation. Clopper-Pearson method was used to estimate the associated 95% confidence intervals (CI). Results: Of the 5357 records initially identified, 28 trials with a total of 2138 patients were considered eligible for inclusion. Neoadjuvant chemotherapy was assessed in 15 trials (1535 patients), NAI in 5 (288 patients), and NACI in 8 (315 patients). The pooled incidence of patients experiencing a pCR was 30.91% (95% CI: 26.54%-35.45%) with NAC, 30.92% (23.22%-39.14%) with NAI, and 42.25% (36.6%-48.0%) with NACI. The difference in pCR among these therapies was statistically significant (p <0.05). Similarly, the pooled incidence of achieving pathological downstaging was 40.91% (34.26%-47.72%) with NAC, 38.84% (26.73%-51.64%) with NAI, and 62.88% (57.22%-68.39%) with NACI. Consistently, there was statistically significant effect modification by the choice of neoadjuvant systemic therapy (p <0.01). The results were consistent across different clinically important subgroups (Table). Conclusions: NACI may achieve better pCR and pathological downstaging than NAC and NAI alone in MIBC patients which may potentially be associated with long-term survival. Definitive data from phase III trials are pending. [Table: see text]

5 Background: Management options for patients (pts) with advanced (locally advanced or metastatic) PeCa are limited. The GSRGT assembled an international cohort of pts with advanced PeCa treated with ICI to evaluate toxicity and clinical outcomes. Methods: We retrospectively collected data on pts with advanced PeCa receiving ≥1 cycle of ICI between 2015-2022 at 18 medical centers in the US, Europe, and Asia. Immune-related adverse events (irAE) were graded per the Common Terminology Criteria for Adverse Events v5.0. Overall survival (OS) and progression-free survival (PFS) were estimated by the Kaplan-Meier method. Objective response rate (ORR) was determined by the clinical investigator per RECIST 1.1 criteria, whenever feasible. Results: Among 72 pts with advanced PeCa treated with ICI, 24 (33%) were Hispanic and 7 (10%) were Black. 60 (83%) pts had metastases while the remainder had locally advanced disease. The median age was 64 (inter-quartile range (IQR): 54,70) years and 48 (67%) had ECOG performance status ≥1. Most pts (n=60, 83%) were treated in the ≥2nd line setting and received pembrolizumab (n=23), nivolumab (n=15), cemiplimab (n=15), nivolumab and ipilimumab (n=7), or other anti-PD1/L1-based therapies (n=12). Among 37 pts with available data on HPV status, 24 (65%) were HPV+. 3 (4%) pts were HIV+. irAE of any grade occurred in 18 (25%) pts, 7 (10%) were grade ≥3, 7 (10%) required steroids, 6 (9%) required hospitalization, and 8 (11%) led to treatment discontinuation. The median OS and 24-month OS and median PFS and 24-month PFS were 9.4 (95%CI: 6.8, 12.8) months and 19.3% (95%CI: 9.2, 32.1) and 2.8 (95%CI: 2.1, 3.9) months and 11.2 % (95%CI: 4.9, 20.2), respectively. Among 66 pts evaluable for response, ORR was 7/66 (11%) (2 with complete response, 5 with partial response), and 16 (24%) pts had stable disease for a disease control rate of 35%. The median duration of response was 7.9 (IQR: 3, not reached) months. Conclusions: In the largest retrospective cohort of ICI-treated advanced PeCa, ICI showed no new safety signals, however, overall anti-tumor activity was limited. Future translational studies are needed to identify pts that are more likely to derive clinical benefit from ICI.