Citations

ABSTRACT Background Men with castration‐resistant prostate cancer (CRPC) who have a pre‐existing history of cardiovascular disease (CVD) or other comorbidities are often excluded from clinical trials involving oral androgen receptor pathway inhibitors (ARPi). In this study, we compared all‐cause and prostate cancer‐specific mortality among CRPC patients, with and without a pre‐existing history of CVD, receiving ARPi compared to chemotherapy. Methods Men with CRPC were identified using the Surveillance, Epidemiology, and End Results‐Medicare Linked Database from 2004 to 2015. Patients were grouped into two analytical cohorts by drug use. Inverse probability treatment weights (IPTW)‐adjusted Cox models and Fine‐Gray subdistribution hazards models were used to evaluate associations between ARPi and chemotherapy, and between ENZ and AA for all‐cause mortality and cancer‐specific mortality separately. Results The study cohort included 1438 men with CRPC. Nearly 54.4% of patients had a pre‐existing history of CVD. Patients with a pre‐existing history of CVD had a higher incidence of all‐cause and prostate cancer‐specific mortality compared to patients without a history of CVD (all‐cause mortality: 69.7% vs. 59.3%, prostate cancer‐specific mortality: 56.0% vs. 49.5%, respectively). In the pre‐existing history of CVD cohort, the IPTW‐adjusted Cox model showed a significantly lower all‐cause mortality in patients who received APRi, enzalutamide, and abiraterone, versus chemotherapy (IPTW‐adjusted hazard ratio [AHR], 0.56; 95% Confidence Interval [CI], 0.48–0.64; p ‐value < 0.001). Further, the IPTW‐adjusted competing risk model showed significantly lower prostate cancer‐specific mortality in patients who received ARPi compared with those who received chemotherapy (IPTW‐AHR, 0.48; 95% CI, 0.41 to 0.57; p ‐value < 0.001). In the without pre‐existing history of CVD cohort, the adjusted Cox model showed significantly lower all‐cause mortality in patients who received APRi than those who received chemotherapy (IPTW‐AHR, 0.49; 95% CI, 0.41–0.60; p ‐value < 0.001). Whereas the IPTW‐adjusted competing risk model showed significantly lower prostate cancer‐specific mortality in patients who received ARPi compared with those who received chemotherapy (IPTW‐AHR, 0.52; 95% CI, 0.42–0.64; p ‐value < 0.001). Conclusion In this population‐based cohort of older men with castration‐resistant prostate cancer, treatment with oral androgen receptor pathway inhibitors was associated with lower estimated risks of all‐cause and prostate cancer–specific mortality compared with chemotherapy in patients with and without pre‐existing CVD. These findings add real‐world comparative effectiveness evidence for patient populations not well represented in randomized clinical trials; however, given the observational design and limitations of administrative data, residual confounding and unmeasured clinical differences may have influenced the results.

As machine learning (ML)-based decision support tools proliferate in clinical practice, understanding how clinicians integrate personalized ML predictions alongside randomized controlled trial (RCT) evidence is critical. We designed a web-based clinical decision support system (CDSS) presenting survival and adverse event data from a simulated RCT and ML model across 12 synthetic multiple myeloma scenarios. In a within-subjects study with 32 physicians, we evaluated how clinicians synthesize competing evidence sources to make treatment decisions. When ML and RCT outputs were concordant, physicians reported greater confidence than with RCT data alone. When results were discordant, most physicians shifted toward the ML-supported treatment, often before reviewing any information about model training or validation, suggesting a tendency toward automation bias rather than algorithm avoidance. Despite reporting higher perceived reliability after viewing model quality disclosures, physicians were largely unable to describe the validation procedures they had reviewed. Taken together, these findings reveal that clinicians may over-rely on ML recommendations even when equipped with tools designed to support critical appraisal. We discuss implications for CDSS design, clinician training, and the institutional safeguards needed before ML-based systems are deployed in high-stakes oncology settings.

e18607 Background: Progression to myelofibrosis (MF) is associated with worse survival and influences treatment decisions, yet predictors of fibrotic progression in polycythemia vera (PV) and essential thrombocythemia (ET) remain poorly characterized. Methods: Patients with pathologically confirmed PV or ET were identified from a retrospective chart review. Demographics (age, sex, race, ethnicity), clinical features (thrombosis and bleeding history, performance status, symptom burden, organomegaly), bone marrow pathology (cellularity, trilineage hematopoiesis, megakaryocyte proliferation and atypia, fibrosis grade), molecular markers ( JAK2 , CALR , MPL , ASXL1 , EZH2 , TET2 , SRSF2 , IDH1/2 , SF3B1 , TP53 ), cytogenetics, laboratory values, and treatment history (phlebotomy, hydroxyurea, interferon, ruxolitinib) at first bone marrow biopsy were recorded. MF progression was defined as hematologist-confirmed transformation to MF. Kaplan-Meier analyses estimated median MF-free survival (MFS), defined as time to MF progression or death. Univariable followed by multivariable Cox proportional regression were performed to identify candidate MFS predictors. The proportional hazards assumption was assessed using Schoenfeld residuals. Hazard ratios (HR) with 95% confidence intervals (CI) were computed, with p<0.05 indicating statistical significance. Results: A total of 505 PV and 198 ET patients were identified. The majority were White (PV: 92%; ET: 86%). PV patients were predominantly male (54%) whereas ET patients were predominantly female (66%). MF progression occurred in 160 (32%) PV and 32 (16%) ET patients. Death was observed in 139 (28%) PV and 10 (5%) ET patients. Median MFS was 158.5 months (95% CI: 143.0-183.0) for PV and 308.0 months (95% CI: 175.0-NR) for ET. In PV, older age at diagnosis (HR 1.04; 95% CI: 1.02-1.07; p<0.001) and shorter time since PV or ET diagnosis to biopsy (HR 0.995; 95% CI: 0.990-0.999; p=0.048) independently predicted worse MFS. In ET, bone marrow hypercellularity (HR 5.67, 95% CI: 1.40-23.0, p=0.015) independently predicted worse MFS. Conclusions: PV patients demonstrated shorter MFS compared to ET. Older age at diagnosis in PV, whereas bone marrow hypercellularity in ET was significantly associated with fibrotic progression or death. Active surveillance of at-risk patients may facilitate early detection and timely treatment initiation. Disease Variable HR (95%CI) P-value PV Age at diagnosis 1.04 (1.02-1.07) <0.001 Time since PV or ET diagnosis to biopsy 0.995 (0.990-0.999) 0.048 ET Bone marrow hypercellularity 5.67 (1.40-23.0) 0.015

e17005 Background: We previously developed a hybrid framework combining LLMs and rule-based algorithms (RBA) for automated risk stratification in localized prostate cancer (locPCa), achieving 89% accuracy. However, clinical T-stage (cTS) inference from narrative MRI reports remained the primary source of errors. Herein, we evaluated whether ensemble approaches using multiple state-of-the-art LLMs could improve cTS inference and overall NCCN risk stratification accuracy. Methods: This study included patients with locPCa presenting with at least one positive prostate biopsy and MRI report available. Building on our previously validated hybrid framework that extracts structured phenotypic variables (PSA, prostate volume, cTS, Gleason patterns, grade group, positive and examined cores) from unstructured reports using LLMs and applies NCCN-concordant RBA for risk stratification, we focused on improving cTS inference. Four HIPAA-compliant LLMs (GPT-4.1, GPT-5.1, Gemini-2.5-Pro, Gemini-2.5-Flash) were evaluated using structured zero-shot prompts iteratively refined through oncologist feedback. Majority voting was employed to generate consensus cTS predictions. Performance was evaluated using weighted accuracy. Results: A total of 358 patients were included. The median age at diagnosis was 65 years (IQR: 60-69); the majority were White (92%) and non-Hispanic (94%). cTS distribution was T2a (51%), T2c (20%), T3a (16%), T3b (7%), T2b (4%), and T4 (2%). The most prevalent risk category was unfavorable intermediate (38%) followed by favorable intermediate (20%), high risk (19%), very high risk (16%), low (5%), and very low (1%). Individual LLM accuracy for cTS inference ranged from 85-90%, with Gemini-2.5-Pro achieving highest accuracy (90%), followed by GPT-4.1 (88%), GPT-5.1 (87%), and Gemini-2.5-Flash (85%). The four-model ensemble reached consensus in 92% of cases, achieving 93% cTS inference accuracy. When integrated into NCCN risk stratification schema, single-model approaches achieved 89-92% accuracy, while ensemble consensus cases achieved 95% accuracy. Conclusions: The ensemble framework using multiple LLMs significantly improves cTS inference and achieves 95% NCCN risk stratification accuracy in consensus cases (92% of patients), representing a 6-percentage-point improvement over our previous single-model framework (89%). This strategy demonstrates that ensemble methods can overcome individual model limitations for complex clinical tasks and holds promise for reliable AI-assisted risk stratification.

e17004 Background: Living clinical practice guidelines offer the opportunity to keep recommendations aligned with the latest evidence, but they also present a fundamental challenge: as evidence accumulates, panelists cannot sustainably query and synthesize a growing body of literature. In developing ASCO's living metastatic castration-resistant prostate cancer (mCRPC) guideline, we encountered this challenge firsthand. To address it, we developed and evaluated an interactive agentic framework that enables guideline panelists to query evidence in natural language and retrieve structured facts, tables, and figures—with the goal of extending this tool as a clinician-facing companion to enhance guideline accessibility. Methods: The mCRPC systematic review dataset supporting ASCO's clinical practice guideline was used, comprising 212 structured columns capturing trial characteristics, patient populations, and outcomes extracted from 188 references across 88 clinical trials. A large language model (LLM)-based agentic framework was developed employing a coding agent that translates natural language queries into executable scripts for structured evidence retrieval. Claude (Haiku 4.5, Sonnet 4.5, and Opus 4.5) served as the core LLM agents. We constructed a benchmark of 50 fact-based queries with gold-standard answers targeting trial identification, treatment modalities, and endpoint characteristics (e.g., "Which trials evaluated systemic monotherapy?"). Performance was evaluated using exact-match accuracy for final answers, and precision, recall, and F1 score for trial-level retrieval. Results: Across 50 queries, exact-match accuracy ranged from 82% to 86%. Haiku 4.5 and Sonnet 4.5 each achieved 82% (42/50), while Opus 4.5 achieved 86% (43/50). Precision ranged from 0.93–0.97, recall from 0.94–0.96, and F1 scores from 0.91–0.94. Conclusions: This agentic framework enables scalable, auditable, natural language interrogation of guideline evidence, addressing a critical bottleneck in living guideline maintenance. For panelists, it supports rapid, reproducible evidence surveillance during recommendation development. For clinicians, it offers a foundation for an interactive companion tool that transforms static guidelines into queryable resources—enabling personalized evidence retrieval at the point of care. LLM EM Precision Recall F1 claude-haiku-4-5 0.82(42/50) 0.95 0.94 0.91 claude-sonnet-4-5 0.82(42/50) 0.97 0.96 0.93 claude-opus-4-5 0.86(43/50) 0.93 0.96 0.94

e17003 Background: Accurate ascertainment of prostate cancer (PCa) disease states—biochemical recurrence (BCR), castration-sensitive prostate cancer (CSPC), and castration-resistant prostate cancer (CRPC)—is critical for automated clinical trial enrollment, quality improvement initiatives, and real-world evidence generation. Essential information for disease state ascertainment resides in unstructured EHR data, requiring labor-intensive manual abstraction that is inconsistent and not scalable. We developed an automated hybrid AI framework using large language models to identify PCa disease states from real-world EHR data. Methods: This retrospective study included patients with histopathologically confirmed PCa (2017-2022) and ≥3 years follow-up. Sequential EHR data—including PSA and testosterone levels, clinical notes, pathology and radiology reports—were retrieved across temporal points. BCR was identified using rule-based algorithms (post-radical prostatectomy [RP]: PSA ≥0.2 ng/mL ×2; post-radiation therapy [RT]: nadir + 2 ng/mL ×2). CRPC was determined using androgen deprivation therapy (ADT) timing, PSA kinetics, testosterone < 50 ng/dL, and radiographic progression per PCWG/EAU-ASCO criteria. GPT-4o with structured parametrized prompts extracted metastatic status and radiographic progression. Independent clinicians performed manual annotation for validation. The framework was developed on 10% of data with evaluation on 90%. External validation used a held-out cohort of clinical trial patients. Results: Among 200 PCa patients (median age 66 years [IQR 60-71]; 91% White, 93% non-Hispanic), the framework demonstrated high accuracy: PCa diagnosis (92%), RP dates (94%), RT dates (93%), and ADT initiation (92%). For disease state identification, BCR detection achieved 90% accuracy, castration status 93%, and metastatic state (M0/M1) 87%. External validation on trial-enrolled patients confirmed robust performance: 26/31 (84%) CRPC trial patients and 26/27 (96%) CSPC trial patients were correctly classified, validating the framework’s reliability for clinical trial matching. Conclusions: This hybrid LLM-based framework accurately identifies PCa disease states using real-world EHR data with minimal manual intervention. The system offers a scalable, reproducible approach for automated disease state identification, enabling streamlined clinical trial matching and enhanced real-world data curation in oncology.

e16502 Background: Accurate postoperative risk stratification in localized clear-cell renal cell carcinoma (ccRCC) informs prognosis, surveillance intensity, and eligibility for adjuvant systemic therapy. Although validated prognostic models (Leibovich 2018, SSIGN, and UISS) exist, they are inconsistently applied in routine practice due to workflow and documentation barriers. Using AI assisted tools within the electronic medical record (EMR) may enable rapid, standardized risk assessment and improve clinical use. Methods: We retrospectively analyzed 92 patients with localized ccRCC undergoing radical nephrectomy and postoperative oncology evaluation (December 2021–September 2025) across three Mayo Clinic sites. Pathology reports and oncology notes were extracted via structured SQL queries. Clinical and pathologic variables were parsed using a zero-shot GPT-4o pipeline. Gold-standard prognostic risk classification and adjuvant therapy eligibility were established by an expert oncology clinician using validated prognostic calculators and KEYNOTE-564 eligibility criteria. AI-generated outputs and novice manual abstraction by two first-year fellows were compared against expert assessment using accuracy, precision, recall, and F1 score. Leibovich and SSIGN were evaluated as three-class models (low/intermediate/high), and UISS as binary risk stratification (high vs non-high). Results: Among 92 patients with localized ccRCC (median age 64 years [range 41–78]), AI-generated outputs demonstrated near-perfect agreement with expert assessment for KEYNOTE-564 adjuvant immunotherapy eligibility (accuracy 97%, precision 100%, recall 97%, F1 score 98%). For three-class prognostic stratification, AI achieved perfect agreement for Leibovich disease-free survival risk (accuracy, precision, recall, and F1 all 100%), outperforming novice abstraction. For SSIGN, AI demonstrated strong multi-class performance (accuracy 91%, precision 91%, recall 93%, F1 score 92%), compared with novice abstraction (accuracy 92%, F1 score 90%). For binary UISS risk stratification, AI achieved 81% accuracy, precision 74%, recall 100%, and F1 score 85% showing a tendency towards false positives. Conclusions: This study demonstrates AI can use existing prognostic and systemic therapy eligibility to accurately risk stratify and identify patients eligible for adjuvant systemic immunotherapy post nephrectomy. For our next steps, we plan to develop a multimodal AI platform integrating pathologic, radiologic, and clinical data to improve prognostication for outcomes and predictive capabilities for patients who would benefit from adjuvant systemic therapy for more personalized medicine.

4589 Background: Recent years have seen a shift in aUC management following approval of enfortumab vedotin–pembrolizumab (EV-P) and gemcitabine/cisplatin plus nivolumab (GC-NIVO) as first-line therapies. Despite improved clinical outcomes, their relative economic value remains uncertain. Herein, we conducted a cost-effectiveness analysis of current 1L regimens in aUC from a U.S. payer perspective using 2026 pricing. Methods: A partitioned survival model with three health states (progression-free [PF], progressed disease [PD], and death) was developed over a 10-year horizon from a U.S. payer perspective. Analyses focused on cisplatin-eligible patients. Survival data were reconstructed from published Kaplan–Meier curves and extrapolated using parametric survival models. Treatment dosing, schedules, and duration were modeled according to trial protocols. Patients receiving gemcitabine/cisplatin without progression were modeled to receive avelumab maintenance (GC→AVEL). Subsequent-line therapy costs were applied at incident progression, and grade ≥3 adverse event costs and disutilities were applied as one-time events. Drug acquisition costs (2026 USD) were based on average sales prices from the Centers for Medicare & Medicaid Services. Costs and quality-adjusted life years (QALYs) were discounted at 3% annually. Incremental cost-effectiveness ratios (ICERs) and net monetary benefit (NMB) were evaluated at willingness-to-pay (WTP) thresholds of $100,000, $150,000, and $200,000 per QALY. Parameter uncertainty was characterized using probabilistic sensitivity analysis. Results: In the base-case analysis, total discounted costs and QALYs were $90,269 and 1.62 for GC→AVEL, $131,836 and 2.09 for GC-NIVO, and $1,451,885 and 2.77 for EV-P. Compared with GC→AVEL, GC-NIVO gained 0.46 QALYs at an incremental cost of $41,567, resulting in an ICER of $89,501 per QALY and positive NMB across all evaluated WTP thresholds. Compared with GC→AVEL, EV-P gained 1.15 QALYs at an incremental cost of $1.36 million (ICER $1.19 million per QALY). Compared with GC-NIVO, EV-P gained 0.68 QALYs at an incremental cost of $1.32 million (ICER $1.93 million per QALY), exceeding all evaluated WTP thresholds. In a scenario analysis limiting EV-P to 10 treatment cycles, total costs decreased to $287,360, with ICERs of $172,000/QALY versus GC→AVEL and $225,000/QALY versus GC-NIVO, remaining above the $150,000/QALY threshold. Conclusions: Using updated 2026 pricing, GC-NIVO is likely to be the most cost-effective first-line regimen for cisplatin-eligible patients with aUC. EV-P offers the greatest clinical benefit but is not cost-effective at current prices. A major limitation of this study is cross-trial variability in study populations and design, limiting the validity of head-to-head economic comparisons.

1617 Background: Large language models (LLMs) are increasingly evaluated for oncology decision support, yet guideline concordance and hallucinations remain variable across cancer types and tasks. Retrieval-augmented generation (RAG) may improve safety by constraining the evidence base used during generation. Prior oncology LLM studies report discordance with guideline-based care, but few describe a reproducible health-system “vetting mechanism” that couples guideline-only retrieval with clinician-scored safety rubrics across multiple oncology domains. Methods: We assembled 316 de-identified oncology cases in a tumor-board template: medical oncology (breast, GI, GU, leukemia; n=216), gynecologic oncology (n=50), and neuro-oncology/CNS metastatic cancer (n=50). Each case was run on three systems: baseline GPT-4 (O1), NCCN-guideline anchored RAG (O2; guideline-constrained retrieval), and a literature-anchored system (OpenEvidence; O3). Oncologists independently scored outputs using a modified Generative Performance Score (mGPS; −1 to +1) combining guideline concordance with hallucination penalties; readability/rationality was scored on a 1–5 Likert scale. Inter-rater reliability was assessed (κ =.63). This workflow was designed as a pre-deployment model credentialing step for health-system governance (source policy → case bank → clinician-scored acceptance criteria). Results: Across all 316 cases (case-averaged), O2 achieved the highest mean total mGPS (0.75) compared with O1 (0.43) and O3 (0.59). In the medical oncology cohort (n=216), O2 outperformed O1 and O3 (mean mGPS 0.72 vs 0.35 and 0.54), with higher guideline concordance (0.85 vs 0.68 and 0.75) and lower hallucination penalties (−0.09 vs −0.25 and −0.19); readability was also highest for O2 (4.17 vs 3.22 and 3.86). In gynecologic oncology (n=50), O2 again performed best (mGPS 0.827) relative to O3 (0.697) and O1 (0.647). In the CNS cohort (n=50), mean composite mGPS favored O2 (0.82) over O3 (0.68) and O1 (0.55), while readability remained stable across models (3.55–3.73). Severe hallucination/unsafe recommendation rate: O1 [26%], O2 [5%], O3 [17%] (pre-specified definition; e.g., major fabricated citation or guideline-discordant harmful recommendation). Conclusions: In a 316-case, multi-domain clinician-scored evaluation, limiting retrieval to oncology guideline sources improved safety-aligned performance relative to unconstrained generation or literature-only anchoring. A guideline-constrained RAG configuration paired with clinician-scored rubrics and tail-risk reporting can serve as a practical health-system vetting mechanism for governance, re-validation, and change control of LLM decision support, including in resource-limited settings where reliable guideline-consistent recommendations are essential.

PURPOSE 177Lu-PSMA-617 (LuPSMA) is an approved prostate-specific membrane antigen (PSMA)–targeted radiopharmaceutical therapy for men with metastatic castration-resistant prostate cancer (mCRPC). Older patients, particularly octogenarians, represent a substantial proportion of men with mCRPC but have been traditionally underrepresented in key registrational trials of LuPSMA. METHODS This retrospective multi-institutional study evaluated patients age 80 years and older with mCRPC after chemotherapy treated with ≥1 cycle of LuPSMA from August 2022 to December 2024. Clinical and demographic data were abstracted from electronic medical records. Outcomes included prostate specific antigen (PSA) response (PSA50 and PSA90), progression-free survival (PFS), overall survival (OS), and toxicities. Kaplan-Meier methods estimated PFS/OS; descriptive statistics summarized baseline and safety data. RESULTS Ninety-five patients (median age, 83 years) were included, of whom 36 (38%) and 21 (22%) had a history of cardiac disease or chronic kidney disease, respectively. Median follow-up was 24.1 months (IQR, 6.7-28.1). Patients received a median of five cycles of LuPSMA. Fifty-two (57%) and 21 (23%) patients achieved a PSA50 and PSA90, respectively. Median PFS and OS were 7.3 months (95% CI, 6.4 to 8.7) and 13.5 months (95% CI, 9.7 to 18.8), respectively. Grade ≥3 hematologic toxicities included anemia (n = 19; 20%) and thrombocytopenia (n = 4, 4%); grade ≥3 acute kidney injury occurred in one patient. Fourteen (15%) patients had dose delays, seven (7%) required reductions, and 10 (11%) discontinued therapy due to toxicity. Thirty-seven patients (39%) were hospitalized during therapy, with intensive care unit-level care required in two patients (2%), and there was one treatment-related death. CONCLUSION LuPSMA had comparable outcomes in octogenarians with mCRPC to patients on registrational trials, although 2/5 of patients required hospitalization during therapy. These findings support the feasibility and efficacy of LuPSMA in well-selected older men with mCRPC and suggest a role for closer monitoring of older patients.