Basu, Arnab

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)

TPS4642 Background: Although first-line (1L) EV + pembrolizumab has improved outcomes for patients with la/mUC, most experience disease progression, highlighting an unmet need for novel therapies. TROP2 is a glycoprotein highly expressed in several epithelial tumors, including UC. Dato-DXd is a TROP2-directed antibody–drug conjugate comprising an anti-TROP2 monoclonal antibody, a tetrapeptide-based cleavable linker, and a topoisomerase I inhibitor payload (DXd), designed to induce selective tumor-cell death and reduce systemic exposure to the payload. Dato-DXd has demonstrated durable efficacy and a manageable safety profile in patients with la/mUC both as monotherapy (TROPION-PanTumor01 [NCT03401385]) and in combination with immunotherapy (TROPION-PanTumor03 [NCT05489211]). We describe a Phase 2/3 trial comparing the efficacy and safety of Dato-DXd + platinum CT (carboplatin or cisplatin) vs gemcitabine + platinum CT in participants with la/mUC with disease progression on or after EV + pembrolizumab. Methods: TROPION-Urothelial03 (NCT07129993) is a global, multicenter, randomized, open-label, Phase 2/3 trial. In Part A (Phase 2), ~60 participants will be randomized 1:1 to receive Dato-DXd 4 or 6 mg/kg + platinum CT. In Part B (Phase 3), ~570 participants will be randomized 1:1 to receive Dato-DXd at the recommended Phase 3 dose (determined using Part A data) + platinum CT vs gemcitabine + platinum CT. Randomization in Part A is stratified by assignment of platinum CT (carboplatin vs cisplatin), and in Part B, by assignment of platinum CT, presence vs absence of liver metastasis at screening, and time to progression on 1L EV + pembrolizumab (<6 vs ≥6 months). Study treatment continues until disease progression, unacceptable toxicity, withdrawal of consent, death, loss to follow-up, or other reasons per protocol. Radiographic tumor assessments occur every 6 weeks for 54 weeks, then every 12 weeks. The primary endpoint of Part A is objective response rate (ORR) by investigator per Response Evaluation Criteria in Solid Tumours, version 1.1 (RECIST 1.1). The dual primary endpoints of Part B are progression-free survival (PFS) by blinded independent central review (BICR) per RECIST 1.1 and overall survival (OS). Secondary endpoints include duration of response by investigator per RECIST 1.1 (Part A), and PFS by investigator per RECIST 1.1 and ORR by BICR and investigator per RECIST 1.1 (Part B). In Part A, ORR will be summarized with 2-sided 95% exact confidence intervals using the Clopper–Pearson method by trial group. In Part B, PFS and OS will be compared between trial groups using a log-rank test stratified by randomization stratification factors, with hazard ratios from the stratified Cox regression model. Clinical trial information: NCT07129993 .

e13656 Background: Effective multidisciplinary cancer care is bottlenecked by the cognitive burden of synthesizing fragmented electronic health records across different specialties. Current AI tools often function as isolated chatbots rather than integrated workflow assistants. We developed and validated "OncoSphere AI," an Agentic AI multidisciplinary clinical workflow support tool designed to autonomously orchestrate the multidisciplinary review process by deconstructing complex cases into domain-specific insights. Methods: OncoSphere AI is a novel, multi-agent large language model–driven platform that simulates a multidisciplinary tumor board for cancer care. The architecture uses a parallel reasoning framework in which eight virtual specialty agents—Medical Oncology, Radiation Oncology, Surgical Oncology, Pathology, Radiology, Pharmacy, Clinical Research, and Nurse Navigation—independently analyze structured and unstructured patient data against specialty specific guidelines and contemporary peer-reviewed literature. System performance was assessed via built-in telemetry (task completion, latency, inter-agent agreement) and blinded review by eight expert oncologists using a structured survey to rate data extraction fidelity, guideline concordance, safety, and clinical utility compared with standard manual review. Results: The study cohort (n=20, median age 59.5 years; 55% female) had a median of 2.5 (range 0-7) prior lines of therapy and multiorgan metastatic disease (median: 2, range 1–4), including lung, breast, gastrointestinal (CRC, pancreatic, hepatobiliary), and genitourinary (prostate, renal) malignancies. OncoSphere AI demonstrated high workflow utility (mean rating: 4/5) and strong extraction fidelity, with more than 90% of biomarker extractions rated as a “perfect match” by expert reviewers. Operational efficiency gains were substantial with AI processing averaged 4.7 minutes per case versus 29.3 minutes for manual review, corresponding to an 84% reduction in time (p<0.001). System telemetry revealed adaptive compute behavior, with the platform autonomously allocating 1.73-fold more reasoning tokens to complex, refractory cases than to standard scenarios. Safety review identified only one critical contraindication issue in a heavily pretreated, multi-line setting. Conclusions: OncoSphere AI effectively automates the clinical “perception layer,” consistently transforming heterogeneous longitudinal data into decision-grade insights with high fidelity and marked time savings. By operating as a coordinated ensemble of domain-specialized AI agents rather than a single generic model, it offers a scalable approach to reduce physician cognitive burden and to standardize guideline-concordant care planning in diverse, heavily pretreated oncology populations.

e13611 Background: Decentralized cancer care models may reduce treatment burden for patients with advanced prostate cancer (CaP), particularly those facing mobility, transportation, of scheduling challenges. However, prospective patient-reported experience and quality-of-life data comparing home- versus clinic-based oncology care remain limited. Mayo Clinic Cancer CARE Beyond Walls (CCBW) is a patient-centered program designed to deliver selected standard-of-care cancer therapies in the home with virtual clinical oversight. This observational study evaluated patient-reported experience, feasibility, safety, and quality of life among men with advanced or metastatic CaP receiving care at home. Methods: Men with advanced/metastatic CaP receiving active standard-of-care therapy including hormone-based treatments, oral anticancer therapies, anti-bone resorptive agents, or intravenous chemotherapy who were participating in a CCBW randomized trial were enrolled into a companion observational study. After confirmation of treatment tolerance in clinic, patients received home-based care for 24 weeks (Arm A) or in-clinic care for 8 weeks followed by home-based care for 16 weeks (Arm B). Patient-reported outcomes were collected at baseline and Weeks 8, 16, and 24, using CAHPS-based measures and CCBW-specific surveys assessing overall cancer care experience, patient preference for site of care, comfort with home infusions, functional/symptom assessment and quality of life. Descriptive analyses were performed. Results: Eighteen participants were analyzed (Arm A n = 10; Arm B n = 8). Participants were medically stable, and predominantly insured by Medicare. Survey completion was 100% at baseline and 83% at all follow-up time points. At Weeks 16 and 24 when all respondents had received home-based care, 100% and 92% reported being “quite a bit” or “very much” comfortable receiving treatment at home. Communication with at-home and remote nursing teams was rated highly, and most respondents reported transition, scheduling, and equipment use easy or very easy. At Weeks 16 and 24, 45% and 40% of respondents reported improved quality of life from receiving home-based care, with no reports of worsening. No emergency department visits, hospitalizations, or urgent care visits occurred in either arm. Conclusions: In this prospective, observational study, home-based oncology care delivered through CCBW was feasible, safe, and associated with high patient satisfaction among men with advanced CaP receiving diverse systemic therapies. These findings support further evaluation and expansion of decentralized care integrating home-based treatment with virtual clinical oversight.

11177 Background: Early detection of recurrence after curative-intent treatment is a critical priority, traditional imaging often detects relapse when disease burden is high, potentially limiting treatment efficacy. ctDNA has emerged as an established prognostic biomarker for molecular residual disease (MRD) detection and early relapse. We performed a real-world meta-analysis on published datasets that utilized personalized, tumor-informed ctDNA testing to assess its performance across tumor types and clinical management approaches. Methods: Real-world studies published from December 2022 to December 2025 using the personalized, tumor-informed Signatera ctDNA assay (Natera, Inc.) were included. Eligibility required ctDNA assessment in the MRD (2–16 weeks post-surgery (histology dependent)/pre adjuvant therapy (ACT)) and/or surveillance (post-ACT/end of the MRD window if no ACT was given) settings, with outcomes stratified by ctDNA status and hazard ratios (HR) reported from univariable or time-dependent Cox regression analysis. Exclusions included: clinical trials, biobanks, metastatic disease treated with palliative intent, case reports, reviews, editorials, or commentaries. Clinical outcomes were harmonized into a composite event-free survival (EFS) endpoint (time to recurrence, progression, or death). Pooled HRs and 95% CIs were estimated using random and fixed effects models and heterogeneity between studies was assessed using the I 2 and Cochran’s Q test. Median lead time from ctDNA-positivity to radiographic or clinical recurrence was summarized when available. Results: We identified 18 eligible publications comprising 3,004 unique patients across 15 solid tumor types. During the MRD window, ctDNA positivity was associated with a significantly increased risk of an EFS event compared with ctDNA negativity (pooled HR: 8.15; 95% CI: 6.12–10.85, I 2 = 0.00%, P = 0.844, n = 13). During the surveillance window, inclusive of post-definitive treatment, ctDNA positivity conferred an even greater EFS risk (pooled HR: 18.30; 95% CI: 14.17–23.65, I 2 = 0%, P = 0.911 for heterogeneity, n = 16). Across studies, ctDNA detection during surveillance preceded radiographic or clinical recurrence by a median of 3.20 months (95% CI: 2.50-4.10, P < 0.0001, n = 10), although this varied substantially by study/tumor type (I 2 = 80.9%). Conclusions: In this pan-cancer real-world meta-analysis, ctDNA positivity assessed using a tumor-informed assay was highly prognostic following curative-intent therapy at all timepoints evaluated with low heterogeneity between studies. These findings support the broad clinical utility of tumor-informed ctDNA testing for post-treatment risk stratification and longitudinal disease monitoring across solid tumors.

794 Background: HER2 protein, encoded by ERBB2 gene, is the target of emerging therapies for patients (pts) with advanced urothelial carcinoma (aUC). High HER2 IHC expression (IHC 3+) is present in ~15% of pts with aUC, while >50% of pts have HER2 IHC 1+/2+/3+ expression. EVP is the preferred 1 st line regimen for pts with aUC, but the impact of HER2 status on EVP outcomes is unclear. Methods: We identified pts from the UNITE study with tumor samples tested for HER2 by IHC and another cohort with next-generation sequencing (NGS) data available. Outcomes were compared by HER2 IHC status (HER2 IHC 0/1+ vs 2+/3+) and separately based on ERBB2 status (alteration vs wild-type). X 2 test was used to compare observed response rate (ORR) and logistic regression was used to compare disease control rate (DCR). Duration of response (DOR), progression-free survival (PFS) and overall survival (OS) from EVP start were evaluated using log-rank test and Cox proportional hazards models. Multivariate analysis (MVA) included age, sex, race, ECOG PS, neutrophil:lymphocyte ratio, albumin, hemoglobin, primary tumor site, tumor histology, prior lines of treatment, and metastatic sites as covariates. Results: Among 521 pts treated with EVP from 16 sites, 112 had HER2 IHC data and 313 had NGS. Median age was 71, 74% were men, 84% Caucasian, 82% had ECOG PS 0-1, 65% pure urothelial histology, and 73% received frontline EVP. HER2 IHC distribution (n=112) was 27%/21%/37%/15% for 0/1+/2+/3+, respectively. ERBB2 alterations were present in 14% (44/313). Baseline characteristics were similar in HER2 0/1+ and 2+/3+ pts, except more primary bladder tumors in 2+/3+ pts (83% vs 63%, p=0.01). For the overall cohort (n=521), with a median follow up of 7.3 mos from EVP start, ORR was 54%, DCR 81%, median DOR (mDOR) 8.5 mos, median PFS (mPFS) 8.5 mos, and median OS (mOS) 19.8 mos. Pt outcomes by HER2 IHC status are shown (Table). In univariate analysis (UVA), pts with HER2 IHC 0/1+ had longer mOS compared to HER2 IHC 2+/3+, but this was not significant in MVA (HR: 0.5; CI: 0.2–1.3, p=0.17). No significant differences were observed in other outcomes. Evaluating clinical outcomes by ERRB2 status, no significant differences were seen for ERBB2 -altered vs wild-type pts. Conclusions: In this retrospective study, no significant differences were seen in EVP treatment outcomes based on HER2 IHC status or ERBB2 status. These findings are hypothesis-generating and require validation in larger and prospective cohorts. HER2 0/1+ (n=54)* HER2 2+/3+ (n=58) OR (95%CI, p) HR UVA (95%CI, p) HR MVA (95%CI, p) ORR 49% (23/47) 55% (27/49) 1.3 (0.6-2.9, p=0.5) - - DCR 79% (37/47) 88% (43/49) 1.9 (0.7-6.2, p=0.2) - - mDOR 11.7 mos 10.4 mos - 1.3 (0.7-2.4, p=0.4) 1.2 (0.4-3.3, p=0.8) mPFS 5.9 mos 7.1 mos - 0.8 (0.5-1.4, p=0.5) 1.1 (0.5-2.5, p=0.9) mOS 42.4 mos 18.7 mos - 2.3 (1.1-4.5, p=0.02) 2.0 (0.8-5.4, p=0.2) *HER2 0/1+ used as reference.

703 Background: The association of circulating tumor DNA (ctDNA) reduction with treatment response has been investigated across multiple malignancies. This study aims to assess the impact of ctDNA response on survival outcomes in patients receiving enfortumab-pembrolizumab (EV-P). Methods: We identified 32 patients with advanced urothelial carcinoma who received EV-P and had ≥2 ctDNA assessments with ≥1 detectable level. Deep molecular response was defined as >2-log reduction from highest ctDNA detection. Radiographic progression was determined from radiology reports documenting progression by RECIST criteria. Kaplan Meier analyses were performed to compute Hazard Ratios (HR) with 95% confidence intervals (CI) for radiographic progression free survival (rPFS), defined as time from treatment start to radiographic progression or death and overall survival (OS) defined as time from treatment start to death were computed. Log-rank tests were performed to compare PFS and OS in patients with deep molecular response to those without response. A p-value <0.05 indicated a statistically significant association with response. Results: Thirty-two patients were included (median age 71 years; range 39–90). Median follow-up was 10.6 months. At cutoff (10/1/2025), 26 patients were alive and 6 had died. Seventeen (53%) achieved a ≥2-log ctDNA reduction (responders), while 15 (47%) did not (non-responders). Progression-free survival was significantly longer in responders (log-rank P < 0.001); median PFS was 5.9 months (95% CI 2.4–NR) in non-responders and not reached in responders (HR 0.09, 95% CI 0.02–0.44; P = 0.0026). Median overall survival was 12.7 months (95% CI 10.0–NR) in non-responders versus not reached in responders (log-rank P = 0.011; HR 0.11, 95% CI 0.01–0.88; P = 0.037). At cutoff, 94% of responders and 67% of non-responders were alive. Conclusions: Achieving a >2-log decrease in ctDNA during EV + Pembrolizumab therapy was significantly associated with improved progression free and overall survival. These findings support ctDNA kinetics as a potential early biomarker of response warranting prospective validation. Clinical outcomes by >2-log ctDNA reduction status. Endpoint > 2-log drop (n = 17 ≤ 2-log drop (n = 15) Log-rank p Median (mo) ≤ 2-log > 2-log HR (95% CI) P C-index PFS 2 events (12%) 10 events (67%) 0.0002 5.9 (2.4–NA) NR 0.09 (0.02–0.44) 0.0026 0.77 OS 1 event (6%) 6 events (40%) 0.011 12.7 (10.0–NA) NR 0.11 (0.01–0.88) 0.037 0.73

567 Background: Selecting adjuvant therapy after nephrectomy for renal cell carcinoma (RCC) requires reviewing scattered electronic health records (EHR) and performing accurate eligibility and risk assessments. Large Language Models (LLMs) may assist physicians by extracting key clinical data and generating evidence-based recommendations between active surveillance and adjuvant pembrolizumab. Methods: We selected RCC patients who underwent nephrectomy at Mayo Clinic in Arizona, Florida, and Minnesota between December 2021 and September 2025. Data extraction from EHRs was performed by a HIPAA-compliant LLM (GPT-5 and Gemini 2.0-Flash-001) and compared to human labels. Data on patients’ demographics, medical history, pathology, laboratory results, and imaging were collected. GPT-5 generated treatment recommendations using a structured prompt that incorporated patients’ features and relevant clinical information that is available to clinicians. Concordance between the human labels and LLM was evaluated for extraction. For treatment recommendations, agreement between the LLM’s output and physician’s documented recommendation in clinical notes was assessed, with disagreements reviewed independently by two blinded genitourinary oncologists. Results: The dataset included 85 patients, median (range) age of 65 (34-84) years; 24.7% were females. Patients who declined adjuvant treatment for personal reasons were excluded. LLM achieved an extraction accuracy of 95% for presenting symptoms, 97% for BMI and ECOG, and > 99% for pathology variables. Concordance between LLM and physician treatment recommendations was 84.7% (72/85). Among 13 discrepant cases, one oncologist agreed with the LLM in 9 cases and the other in 8. The main reasons for disagreement between the physicians and the LLM were that the LLM favored active surveillance for patients with declining renal function, whereas physicians still considered adjuvant pembrolizumab appropriate in such cases. Additionally, the LLM demonstrated hesitancy to recommend pembrolizumab for patients with low performance status. Overall, the 13 discrepant cases were nuanced and challenging, with disagreement observed between the two oncologists themselves in three instances. Conclusions: LLMs demonstrated high accuracy in extracting clinically relevant variables and generating objective, evidence-based adjuvant treatment recommendations for RCC after nephrectomy. Importantly, LLM did not generate any inappropriate or potentially harmful recommendations. Our pipeline can serve as a supportive decision tool to enhance consistency in post-nephrectomy management and assist physicians in choosing between active surveillance and adjuvant pembrolizumab. However, final treatment decisions should remain the responsibility of the treating physician after comprehensive patient evaluation.

360 Background: Adding NHT to androgen deprivation therapy (ADT) improves survival in mCSPC, but their comparative impact on patient-reported QoL and time to deterioration (TTD) across FACT-P domains remains underexplored. Methods: MEDLINE and Embase were systematically searched from each database’s inception through August 21, 2025, to identify Phase III trials assessing NHT treatment in mCSPC. Time to FACT-P (total, physical, emotional, functional, and social/family well-being) deterioration across trials with the latest follow up were extracted. Hazards ratios (HRs) and 95% confidence intervals (CIs) were pooled and mixed treatment comparisons were made using network meta-analysis. P-scores were computed; a higher score indicated better QoL. Results: A total of 4 trials (12 references) with 4,037 patients met the inclusion criteria with median follow up ranging from 14.4 months to 30 months. Median time to QoL deterioration ranged from 11.4 to 12.9 months. In terms of FACT-P overall, darolutamide (DARO)+ADT (HR: 0.76; 95% CI: 0.62-0.93), abiraterone acetate (AAP)+ADT (0.85; 0.73-0.98) significantly improved TTD as compared to ADT alone. DARO+ADT was associated with significant improvement in TTD when compared to APA+ADT (0.75; 0.56-0.98). In terms of FACT-P social/family wellbeing, DARO+ADT significantly improved TTD when compared to AAP+ADT (0.75; 0.57-0.96), and ADT (0.79; 0.64-0.98). Likewise, ENZA+ADT significantly improved TTD when compared to AAP+ADT (0.77; 0.63-0.95), APA+ADT (0.88; 0.61-0.99) and ADT (0.82; 0.70-0.95). However, there was no statistically significant difference between DARO+ADT and ENZA+ADT. In terms of FACT-P physical wellbeing, AAP+ADT was associated with statistically significant improvement in TTD when compared to APA+ADT (0.66; 0.52-0.83), ENZA+ADT (0.74; 0.61-0.90), and ADT (0.75; 0.65-0.87). In terms of FACT-P functional wellbeing, DARO+ADT was associated with statistically significant improvement in TTD when compared to ADT (0.78; 0.63-0.96). The ranking analysis is showed in Table. Conclusions: Novel hormonal therapies significantly delay QoL deterioration in mCSPC, particularly across overall, emotional, and functional domains. Domain-specific analyses suggest darolutamide may preserve social/family and functional well-being, while abiraterone favors physical well-being. Rank table. Rank (P-score) FACT-Ptotal FACT-Psocial/family FACT-Pphysical FACT-Pfunctional FACT-Pemotional AAP+ADT 2 (0.73) 5 (0.18) 1 (0.98) 2 (0.62) 3 (0.52) APA+ADT 5 (0.17) 4 (0.22) 5 (0.07) 4 (0.33) 4 (0.52) DARO+ADT 1 (0.90) 1 (0.88) 2 (0.68) 1 (0.92) 2 (0.66) ENZA+ADT 3 (0.50) 2 (0.84) 4 (0.36) 3 (0.52) 1 (0.67) ADT 4 (0.18) 3 (0.38) 3 (0.42) 5 (0.11) 5 (0.13)