Citations

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.

e16598 Background: Brain metastases occur in a substantial proportion of patients with renal cell carcinoma (RCC) and are associated with poor prognosis. Stereotactic radiosurgery (SRS) is frequently used for intracranial disease control; however, contemporary real-world survival outcomes and the influence of systemic therapy exposure and metastatic burden remain incompletely characterized. Methods: We conducted a retrospective observational study using the TriNetX global federated health research network. Adult patients (≥18 years) with RCC and documented brain metastases who underwent stereotactic radiosurgery between January 2013 and December 2023 were identified. The index date was defined as the date of first SRS. Baseline demographics, comorbidities, and prior systemic therapy exposure including immune checkpoint inhibitors and VEGF-targeted therapies within 6 months before SRS were assessed. Metastatic burden was approximated using the number of distinct extracranial metastatic organ sites (single-organ vs multi-organ involvement). Overall survival (OS) was evaluated using Kaplan–Meier methods. Results: A total of 232 patients with RCC and brain metastases treated with SRS were included. Mean age was 61.2 ± 10.6 years, and 72% were male. Common comorbidities included hypertension (53%), diabetes mellitus (30%), chronic kidney disease (28%), and cerebrovascular disease (26%). Prior to SRS, 78% of patients received systemic therapy, including immune checkpoint inhibitors (49%) and VEGF-targeted therapies (46%). Using organ-site count as a proxy for disease burden, 59% demonstrated single-organ extracranial metastatic involvement. During follow-up, 95 patients (40.9%) experienced all-cause mortality. Median overall survival was 967 days (31.8 months). Estimated long-term survival probability at maximum follow-up was 21.2%. Kaplan–Meier analysis demonstrated early survival attrition followed by prolonged survival in a subset of patients. Patients receiving immunotherapy and those with limited extracranial metastatic burden demonstrated more favorable survival patterns. Conclusions: In this large contemporary real-world cohort of RCC patients with brain metastases treated with stereotactic radiosurgery, median overall survival exceeded 2.5 years, providing important modern survival benchmarks. Favorable outcomes were observed among patients receiving systemic immunotherapy and those with limited metastatic burden, supporting the continued role of SRS within multidisciplinary management strategies for selected patients with metastatic RCC.

e16586 Background: Guideline-discordant care disproportionately affects racial minorities, uninsured, and underinsured patients, yet prior studies lack quantification of absolute survival time lost. We evaluated equity-adjusted survival penalties from non-adherence to guideline-concordant therapy in muscle-invasive bladder cancer (MIBC). Methods: From the National Cancer Database, we identified patients with non metastatic muscle invasive bladder cancer (cT2 to T4a, N0 or N1, M0). Overall survival was defined from diagnosis to death or last contact. Guideline-concordant therapy was defined as radical cystectomy (± perioperative systemic therapy) or trimodality therapy (definitive bladder radiotherapy with chemotherapy); all other patterns were classified as discordant. Of 55,873 eligible patients, 28,025 received concordant therapy (27,401 cystectomy; 524 trimodality therapy) and 27,848 received discordant careInverse probability of treatment weighting (IPTW) was used to balance demographic, socioeconomic, facility, tumor, and temporal covariates. We analyzed survival using weighted Kaplan-Meier and Cox models censored at 36 and 60 months. Equity-adjusted survival penalties were calculated as the difference between observed survival and counterfactual survival predicted from concordant-only models. Restricted mean survival time losses were extrapolated to years lost per 1,000 patients. Results: Guideline concordance was associated with significantly improved survival (36-month HR 0.64; 60-month HR 0.68; both p < 0.01). Among concordant modalities, radical cystectomy conferred an early survival advantage over trimodality therapy at 36 months (HR 0.23; p < 0.05) which dissipated by 60 months (HR 0.53; p > 0.05). Survival penalties were most severe for vulnerable populations. At 36 months, uninsured patients incurred a 0.90-month survival penalty (75.0 life-years lost per 1,000 patients). By 60 months, penalties widened: uninsured patients lost 1.94 months (161.7 life-years/1,000), Filipino patients lost 1.53 months (127.0/1,000), Medicaid beneficiaries lost 0.97 months (80.8/1,000), and patients at community cancer programs lost 0.21 months (54.1/1,000). Conclusions: In this national cohort of 55,873 patients, guideline concordance reduced the propensity-weighted risk of death by 32% to 36% at 3 and 5 years. However, non-adherence results in substantial absolute survival losses for marginalized groups. By quantifying these "survival penalties," we demonstrate that targeting the factors underlying discordance offers a direct opportunity to reduce inequities and achieve measurable population-level survival gains in MIBC.

e16564 Background: Urothelial carcinoma (UC) predominantly affects older adults with a high burden of comorbidities, particularly chronic kidney disease (CKD), which affects 30–40% of patients at diagnosis and is more prevalent in advanced disease. Enfortumab vedotin plus pembrolizumab (EV+P) has become a frontline standard for advanced UC; however, patients with moderate-to-severe CKD were underrepresented in pivotal trials. Enfortumab vedotin targets Nectin-4 expressed in renal tubular epithelium, and immune checkpoint inhibitors are associated with immune-mediated nephrotoxicity, raising concern for renal vulnerability in patients with reduced renal reserve. Real-world data evaluating renal and clinical outcomes of EV+P in patients with baseline CKD are limited. Methods: We conducted a retrospective cohort study using the TriNetX Global Collaborative Network, identifying adult patients with UC treated with EV+P between January 2020 and December 2025. Baseline CKD status (stages 3–5; eGFR < 60 mL/min/1.73 m²) was determined by diagnostic codes and laboratory data obtained prior to EV+P initiation.Patients were stratified into CKD and non-CKD cohorts. Propensity score matching (1:1) was performed for demographics, comorbidities, laboratory values, and prior therapies. Outcomes assessed within 180 days of treatment initiation included overall survival (OS), major adverse kidney events (MAKE; composite of acute kidney injury, dialysis initiation or dependence, or death), and selected treatment-related toxicities. Kaplan–Meier and Cox proportional hazards models were used. Results: After matching, 386 patients were included in each cohort (median age 75 years), with well-balanced baseline characteristics (all post-matching standardized mean differences < 0.1). Prior to matching, patients with CKD had a higher burden of cardiovascular comorbidities. Baseline renal function differed by design, with lower mean eGFR and higher serum creatinine in the CKD cohort. Overall survival was similar between CKD and non-CKD cohorts (HR 0.96; 95% CI, 0.69–1.35; p = 0.83), with 180-day survival rates of 79.9% and 79.3%, respectively. Patients with CKD had a higher incidence of MAKE, reflected by lower 180-day event-free survival (49.4% vs 59.1%; HR 1.41; 95% CI, 1.13–1.76; p = 0.002). Treatment-related toxicities were also more frequent in the CKD cohort (HR 1.35; 95% CI, 1.05–1.74; p = 0.020), with a higher mean number of toxicity events (4.3 vs 3.4). Conclusions: In this large real-world analysis, EV+P use was associated with comparable short-term overall survival in UC patients with and without CKD but with higher rates of renal events and treatment-related toxicities among patients with baseline CKD. These findings support the use of EV+P in select UC patients with CKD, with close monitoring of renal function and treatment-related toxicity.

e16535 Background: RMC is a rare and highly aggressive malignancy that predominantly affects young individuals with sickle hemoglobinopathies. Due to its rarity, no validated prognostic models exist to guide clinical decision-making. We aimed to develop and internally validate a prognostic model for patients with RMC using routinely available clinical and laboratory variables. Methods: We analyzed a single-institution cohort of patients with RMC. Candidate variables were identified based on biological plausibility and represented in a directed acyclic graph. Missing data were handled via multiple imputation by chained equations ( m = 50). Variables were ranked using elastic net Cox regression ( α = 0.5, λ = λ min ) with stability selection derived from 100 bootstrap resamples per imputed dataset. Predictors achieving > 60% selection frequency were retained for a nested parsimony analysis. Bootstrap internal validation ( B = 2000) estimated optimism-corrected discrimination and calibration performance, with the calibration slope applied as global shrinkage to penalize final coefficients for overfitting. Clinical utility was evaluated via decision curve analysis (DCA) at 18- and 24-month horizons. Results: Among 180 patients (median age 29; IQR 23–37), 142 deaths occurred with a median follow-up of 12.4 months. The final model included four predictors (Table). The optimism-corrected c-index was 0.667 (optimism = 0.009), and the calibration slope was 0.964, indicating minimal overfitting. Calibration plots showed excellent predicted-observed agreement at 12, 18, 24, and 36 months. DCA showed net benefit across threshold probabilities of 10%–75% at 18 months and 10%–80% at 24 months compared to default strategies. In sensitivity analysis, inclusion of TP53 mutation status improved discrimination ( c -index 0.686) and model fit ( ΔAIC −10.6). An interactive web-based risk calculator was developed to facilitate clinical implementation. Conclusions: This study presents the first internally validated prognostic model for RMC, demonstrating acceptable discrimination and excellent calibration using routinely available variables. By stratifying patients into distinct risk groups, this model may inform treatment intensity—guiding escalation to multi-agent regimens or sequential chemotherapy cycling in high-risk patients, while identifying lower-risk patients who may benefit from de-escalation strategies such as metastasis-directed therapy for oligometastatic disease or consideration of cytoreductive nephrectomy. External validation is warranted. Variable Time Ratio 95% CI p value ECOG performance status 0.64 0.53–0.77 <0.001 Distal (non-retroperitoneal) lymph node metastases 0.69 0.52–0.92 0.011 Neutrophil-to-lymphocyte ratio 0.98 0.94–1.02 0.25 Corrected calcium (mg/dL) 0.77 0.56–1.06 0.10 Sensitivity Analysis Only TP53 Mutation 0.63 0.36-1-10 0.10

e16510 Background: Early-onset RCC may reflect distinct biology and hereditary predisposition, yet data on stage-specific clinical features and outcomes remain limited. We comprehensively evaluated age-stratified, stage-specific clinical features and survival outcomes using the National Cancer Database. Methods: Adults with histologically confirmed RCC diagnosed between 2004 and 2022 were analyzed. Patients were categorized into three groups by age at diagnosis: early-onset ( < 40 years), middle-aged (40–50 years), and late-onset ( > = 50 years). The sociodemographic and clinical characteristics, comorbidities, histology, and treatment patterns were compared using descriptive statistics. Overall survival (OS) was evaluated using Cox proportional hazards models, adjusting for the aforementioned variables. Stage-specific analyses were conducted, including a metastatic disease cohort. Results: Among 742,114 patients, 4.3% were classified as early-onset, 10.9% as middle-aged, and 84.8% as late-onset. Overall, 12.6% presented with metastatic disease. Compared with late-onset, early-onset patients were more likely to be from racial minority populations (non-White: 16.9 % vs. 20.2 % in late-onset), uninsured, and had fewer comorbidities (Charlson–Deyo score 0: 85.2% vs. 65.2%). Early-onset RCC was characterized by a lower proportion of papillary RCC (8.9% vs. 12.5%) and higher proportions of clear cell (54.9% vs. 50.8%) and chromophobe RCC (9.3% vs. 4.5%) than late-onset RCC and more likely to undergo partial nephrectomy (50.4% vs. 30.2%). They also presented with earlier-stage disease, with lower rates of stage III (7.4% vs. 10.2% vs. 13.4%) and stage IV disease (5.4% vs. 8.9% vs. 13.5%) compared with middle-aged and late-onset groups. Among patients with metastatic disease, early-onset RCC demonstrated higher minority representation (30.8% vs. 14.1%) and a higher prevalence of non-clear cell histology (70.6% vs. 58.1%). In multivariable analysis, compared with early-onset RCC, middle-aged (HR 1.44, 95% CI 1.39–1.50; P < 0.001) and late-onset patients (HR 2.83, 95% CI 2.74–2.93; P < 0.001) experienced significantly worse OS. However, among metastatic disease, this survival advantage was attenuated. Early-onset patients had similar OS to middle-aged patients (HR 0.99, 95% CI 0.93–1.05, p = 0.69) and late-onset patients only had a modestly worse OS than the early-onset group (HR 1.09, 95% CI 1.03–1.15 p = 0.003). Conclusions: Early-onset RCC was associated with more favorable baseline characteristics, earlier stage at diagnosis, and higher utilization of nephron-sparing nephrectomy, corresponding to improved OS. However, this survival advantage diminishes in metastatic disease despite better baseline health, highlighting stage-dependent differences in outcomes and suggesting distinct histologic patterns in advanced early-onset RCC.

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.

e16488 Background: Small bowel adenocarcinoma (SBA) and ampullary adenocarcinoma (ampullary AC) are rare gastrointestinal (GI) malignancies. Treatment strategies are largely extrapolated from colorectal cancer, where oxaliplatin-based regimens are the most common first-line therapy in advanced disease. Although bevacizumab is considered safe in metastatic SBA, its efficacy remains unproven, and evidence is limited to small retrospective studies. We assessed the association of adding bevacizumab to FOLFOX with survival and toxicity in a large real-world cohort. Methods: We conducted a retrospective cohort study using the TriNetX Global Collaborative Network (January 2000-December 2025), including data from 105 healthcare organizations. Adults (≥18 years) with SBA or ampullary AC treated with FOLFOX with or without bevacizumab were identified. Cohorts were propensity score-matched on age, sex, race, comorbidities, and metastatic sites. Outcomes included overall mortality, GI perforation, bleeding, venous thromboembolism (VTE), hospital admission, and emergency department (ED) visits. Kaplan-Meier and Cox proportional hazards models were used for time-to-event analyses. Results: After matching, 818 patients were included in each cohort. In the FOLFOX + bevacizumab cohort, mean age was 66.1 ± 12.0 years and 53.8% were male; 63.7% were White, 13.8% Black, 5.6% Asian, 12.0% Unknown/Other, and 4.4% Hispanic/Latino. Over a median follow-up of 13 months, FOLFOX + bevacizumab was associated with higher overall mortality compared with FOLFOX alone (55.3% vs 45.2%; HR 1.24 [95% CI 1.1-1.4]; p = 0.002) and shorter median OS (15.1 vs 19.5 months). Rates of GI perforation (15.5% vs 16.7%; HR 0.90 [95% CI 0.71-1.15]; p = 0.39), GI bleeding (13.1% vs 13.6%; HR 0.96 [95% CI 0.73-1.25]; p = 0.74), and VTE (19.1% vs 21.4%; HR 0.88 [95% CI 0.71-1.09]; p = 0.23) were similar between groups. Hospital admissions were higher in the FOLFOX + bevacizumab cohort (2.8% vs 1.3%; HR 2.1 [95% CI 1.0-4.3]; p = 0.04). Findings were consistent at 5-year follow-up. Conclusions: In this real-world cohort, the addition of bevacizumab to FOLFOX was not associated with improved survival and was associated with higher healthcare utilization. Prospective, disease-specific studies are needed to define its role in this population. Outcomes at median 13-month follow-up after propensity matching (FOLFOX + bevacizumab vs FOLFOX). Outcome % vs % HR p Overall mortality 55.3 vs 45.2 1.24 0.002 Hospital admission 2.8 vs 1.3 2.1 0.04 ED visit 45.2 vs 42.4 1.02 0.75 GI perforation 15.5 vs 16.7 0.90 0.39 GI bleeding 13.1 vs 13.6 0.96 0.74 VTE 19.1 vs 21.4 0.88 0.23

e16486 Background: Allele-specific KRAS G12C inhibitors have revolutionized lung and colorectal cancer treatment, yet the biology of G12C in other gastrointestinal (GI) malignancies remains undefined. We aimed to elucidate the clinicopathologic and genomic landscape of KRAS G12C-mutated tumors compared to other G12 variants in a large non-colorectal GI cohort. Methods: We queried cBioPortal for pancreatic, esophagogastric, hepatobiliary, ampullary, and appendiceal adenocarcinomas. Following data quality control, the final cohort (n = 6,352) was stratified into KRAS wild-type (n = 3,293), G12C (n = 63), and other G12 variants (n = 2,996). Clinicopathologic features, tumor mutational burden (TMB), microsatellite instability status, and co-alteration in 96 key genes across 11 signaling pathways were compared. Statistical significance was assessed using Fisher’s exact or Chi-square tests for categorical variables and Kruskal-Wallis tests for continuous variables. The Benjamini-Hochberg method was applied to correct for multiple hypothesis testing. Results: KRAS G12C prevalence was 1.0% overall, with significant heterogeneity across cancer subtypes (p < 0.001). Appendiceal adenocarcinoma demonstrated the highest G12C prevalence at 5.71% (10/175), followed by ampullary (1.44%, 2/139), pancreatic (1.17%, 34/2906), hepatobiliary (0.88%, 12/1360), and esophagogastric cancers (0.28%, 5/1772). G12C patients were significantly older than wild-type patients (64.8±11.2 vs 62.5±12.9 years, mean±SD, p < 0.001) but younger than other G12 variant carriers (65.8±10.4 years). G12C tumors demonstrated intermediate genomic instability, with a TMB of 4.82±6.17 compared to 4.98±10.55 in wild-type tumors and 3.67±10.12 in other G12 variants (p < 0.001). Notably, G12C tumors were universally microsatellite stable (0% MSI-H vs 4.0% wild-type vs 0.9% other G12, p < 0.001). Genomic analysis revealed that AKT1 mutations were significantly enriched in the G12C cohort (6.3% vs 0.4%; q = 0.025) as compared to other G12 variants. A trend of increased mutation frequency in HRAS was observed specifically in pancreatic cancers in G12C tumors as compared to other variants (8.8% vs 0.7%; q = 0.083). In pancreatic adenocarcinoma, G12C tumors were enriched for chromatin modifiers, notably ARID1A (17.6% vs 7.7%, p = 0.045) and SETD2 / SETD26 (8.8% vs 1.1%, p = 0.006), though these did not reach FDR significance. Notably, the immunotherapy-resistance drivers STK11 and KEAP1 were absent (0%) in the whole G12C cohort. Conclusions: KRAS G12C represents a biologically distinct subgroup of exon 2 KRAS mutations compared to other KRAS G12 pathogenic variants. The absence of resistance markers and enrichment of actionable AKT1 and chromatin-modifier mutations suggest distinct therapeutic vulnerabilities in non-colorectal KRAS G12C-mutated GI cancers.

e16469 Background: Pancreatic cancer remains among the most lethal malignancies worldwide, with limited improvement in survival despite therapeutic advances. Contemporary real-world data characterizing recent diagnosis acceleration, demographic disparities, and future burden remain limited. We evaluated national trends in pancreatic cancer burden using a large federated electronic health record (EHR) network. Methods: We conducted a retrospective multicenter real-world cohort analysis using the TriNetX Research Network. Adult patients (≥18 years) with pancreatic cancer were identified using ICD-10-CM code C25. Annual newly recorded diagnosis proportions, prevalence, and incidence rates were calculated for calendar years 2020–2024 and stratified by age, sex, race, and ethnicity. New diagnoses were defined as first recorded pancreatic cancer encounters within each calendar year. Exploratory future burden projections were generated through 2028 using autoregressive integrated moving average (ARIMA) time-series modeling. Results: Between 2020 and 2024, newly recorded pancreatic cancer diagnoses increased from 0.032% (21,118 cases) to 0.046% (23,660 cases), representing a descriptive relative increase exceeding 40%. Prevalence increased from 0.101% (65,971 cases) to 0.145% (74,139 cases), reflecting rapid expansion of disease burden. Diagnosis burden increased markedly with advancing age, with the highest burden observed among individuals aged ≥65 years. Males consistently demonstrated higher diagnosis proportions compared with females. Racial stratification demonstrated higher recorded diagnosis proportions among White and Asian populations, while substantial prevalence burden persisted across all racial groups, potentially reflecting differences in healthcare utilization and access. Non-Hispanic individuals exhibited higher diagnosis and prevalence proportions compared with Hispanic individuals. Exploratory ARIMA forecasting projected continued growth in pancreatic cancer burden through 2028, with widening uncertainty intervals over time, indicating escalating long-term healthcare demand. Conclusions: Pancreatic cancer burden in the United States is rapidly increasing, with disproportionate impact among older adults and males and persistent demographic disparities. Exploratory projection modeling suggests continued expansion of disease burden in the near term. These findings highlight the urgent need for earlier detection initiatives, risk-adapted screening strategies, and healthcare system preparedness to address the growing national impact of pancreatic cancer.