Citations

58 Background: Skeletal-related events (SREs) are a common cause of morbidity and mortality in metastatic castration-resistant prostate cancer (mCRPC). These complications are managed using bone-modifying agents (BMAs), radiation therapy, and surgery, and lead to diminished quality of life irrespective of their castration status. We sought to evaluate the mortality trends for patients with PC and SREs in patients of >65 years from 1999 to 2023 using the CDC WONDER (Wide-ranging Online Data for Epidemiologic Research) database. Methods: The database was searched from January 1, 1999, to December 31, 2023. Mortality trends were assessed by identifying prostate cancer international classification of diseases-10th revision code (ICD-10 code C61) as the underlying cause of death in individuals aged >65 years, and (SREs) as contributing causes using the following ICD-10 codes: S02, S12, S22, S32, S42, S52, S62, S82, S92, T02, T08, T10, T12, T14.2, Y93.2, E55, M80, M81, M84.4. Joinpoint 5.0 software was used to analyze mortality trends. Results: From 1999 to 2023, a total of 3,231 deaths were reported. Trend analysis revealed a decline in overall mortality from 1999 to 2014, with an annual percentage change (APC) of -1.60 (CI: -2.6, -0.5; p = 0.005), followed by a significant increase in mortality from 2015 to 2023, with an APC of 5.30 (p = 0.0001). When stratified by age, individuals aged 65-69 showed an insignificant decline in mortality over the entire study period. For those aged 70-74, mortality significantly decreased from 1999 to 2014 (APC: -3.02, CI: -5.0, -0.9; p = 0.008), but experienced an exponential increase from 2014 to 2023 (APC: 14.07, CI: 8.7, 19.6; p = 0.00005). In the 75-79 age group, mortality significantly decreased from 1999 to 2011 (APC: -5.6, CI: -9.5, -1.6; p = 0.0091), followed by a robust increase from 2011 to 2023 (APC: 7.06, CI: 2.7, 11.5; p = 0.002). For individuals aged >80 years, the trend showed an insignificant decline in mortality from 1999 to 2023 (APC: -0.260, CI: -0.8, 0.3; p = 0.380). Conclusions: Our study revealed significant trends in prostate cancer mortality attributable to SREs, contributing to a sharp increase from 2015 to 2023. Age-stratified analysis revealed that substantial mortality is reported in patients aged 70-79, with the most pronounced rise occurring after 2014. These findings emphasize the growing burden of SREs in older prostate cancer patients, underscoring the need for targeted interventions along with a focus on patient-reported outcomes to improve survival and quality of life.

57 Background: Cardiovascular disease (CVD) is the leading non-cancer cause of death among men with prostate cancer (PC), particularly during androgen-deprivation therapy. While CVD risk-prediction models are well established in the general population, barriers to integration within oncology workflows remain poorly defined. Guided by the PRISM and RE-AIM frameworks, this study evaluated multilevel determinants influencing adoption, feasibility, and sustainability of CVD risk-assessment tools in PC care. Methods: A national, cross-sectional survey of 45 physicians (medical oncology 49%, urology 38%, radiation oncology 13%) from 18 U.S. states was conducted. Respondents represented early- (27%), mid- (36%), and late-career (38%) clinicians across academic (87%) and community (13%) settings. Quantitative items mapped to PRISM/RE-AIM domains assessed adaptability, cost/resource feasibility, effectiveness, equity, and sustainability; open-ended responses explored contextual barriers and facilitators. Descriptive statistics summarized quantitative data. Results: Physicians reported moderate-to-high readiness to adopt a CVD risk-assessment tool, emphasizing workflow alignment and EHR integration over algorithmic complexity. Table 1 summarizes domain-specific findings. Ninety-three percent of respondents favored lab-independent models that minimize workflow disruption. Conclusions: Pre-implementation analysis highlighted that structural barriers, rather than algorithmic or statistical limitations, remain the primary obstacles to adopting CVD risk-assessment tools in prostate-cancer care. Physicians emphasized the need for EHR-integrated, non-laboratory-based, and time-efficient tools that align with routine oncology workflows and minimize cross-specialty friction. Addressing these practical and organizational determinants may be pivotal to achieving equitable, scalable, and sustainable cardio-oncology implementation across diverse practice settings. PRISM and RE-AIM constructs evaluation results. Construct Likely/Very Likely (%) Interpretation Perceived Effectiveness 87% Broad confidence in clinical utility Equity in Patient Reach 75% Strong perceived equity in reach Equity in Effectiveness 73% Expected to perform consistently across populations Adaptability to Workflow 71% High compatibility with existing clinic workflows Cost/Resource Feasibility 65% Feasible for resource-constrained settings Likelihood of Adoption 60% Moderate readiness for implementation Equity in Adoption 58% Equitable uptake possible despite resource differences Reach to Target Population 57% Moderate likelihood of reaching most patients Sustainability Over Time 51% Moderate long-term feasibility Consistency of Implementation 42% Need for standardized protocols

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.

5 Background: Penile cancer (PC) is often diagnosed late, which leads to advanced disease. Studies have reported that less than 18% of individuals with metastatic PC receive any kind of palliative therapy. Place of Death (PoD) is a crucial determinant of patient and caregiver preferences and the cost of caregiving at the end of life (EOL). We aimed to evaluate the trends in PoD for patients with PC in the U.S. from 2003 to 2023 based on the CDC WONDER (Wide-ranging Online Data for Epidemiologic Research) database. Methods: We analyzed data from January 01, 2003, to December 31, 2023. The data for deaths due to PC were pooled using the International Classification of Diseases-10th Revision code as C60 for malignant neoplasm of the penis. The inclusion criteria included patients aged > 25 years. PoD is defined as deaths at home and hospice (H&H) versus medical facilities. Average annual percentage change (AAPC) was calculated using the Joinpoint Regression Program, version 5.0.2. Results: The analysis demonstrated a total of n = 6,074 deaths from PC. A total of 52.1% of deaths, n = 3,167, were reported at H&H. The trend analysis demonstrates an exponential rise in H&H utilization from 34.4% in 2003 to 57.5% in 2023, AAPC [5.4284, CI (4.7-6.10, p < 0.000001]. A significant increase in the utilization of H&H is notable in both African Americans (AA), AAPC [2.8514, CI (1.0-4.7), p = 0.004], and the White population, AAPC [5.3497, CI (4.6-6.0); p < 0.000001]. However, a racial disparity is also noted, with 45.4% in AA, while the White population is at 52.5% deaths in H&H-based care. PC mortality in medical facilities has a 29.7% decline from 2003 to 2023. On age stratification, there was a significant increase in utilizing H&H as PoD for the 45-64 years cohort AAPC [4.29, CI (3.2, 5.2), p < 0.000001], similarly, with the older age group 65+ cohorts AAPC [6.26, CI (5.4, 7.0), p < 0.000001]. In contrast, the younger cohort deduced no significant trend AAPC [2.92, CI (-1.5, 7.6), p = 0.14]. Conclusions: This is the first study to our knowledge providing valuable insights into the evolving PoD preferences among PC patients in the U.S.The age-related trends underscore the importance of tailoring end-of-life care strategies to specific age cohorts. The findings emphasize the importance of policies promoting and supporting H&H care for PC patients. Targeted efforts are warranted to address notable disparities in PoD preferences among racial and age groups.

46 Background: Primary results from the randomized, multicenter, open-label, phase 3 TRITON3 (NCT02975934) study demonstrated that rucaparib significantly improved radiographic progression-free survival (rPFS) vs physician’s choice of DTX or an ARPI. ECOG PS is used as a predictor of prostate cancer survival. In this post hoc analysis, we report results from the BRCA1/2 subgroup of pts stratified by ECOG PS. Methods: Pts with disease progression after 1 prior second-generation ARPI in any setting were randomized 2:1 to rucaparib 600 mg BID or physician’s choice of DTX or ARPI (abiraterone acetate or enzalutamide). Primary endpoint was rPFS by independent radiology review (IRR). Objective response rate (ORR) for pts with measurable disease and overall survival (OS) were key secondary endpoints. Crossover from physician’s choice to rucaparib was allowed after radiographic progression was confirmed by IRR. Data cutoff was August 25, 2022. Results: Of 302 pts in the BRCA1/2 population, 147 had an ECOG status of 0 (97 rucaparib; 50 physician’s choice) and 155 had ECOG 1 (104 rucaparib; 51 physician’s choice). Baseline characteristics and demographics were generally similar between arms for pts in the ECOG 0 and ECOG 1 groups. Median rPFS favored rucaparib vs physician’s choice for pts with ECOG 0 (11.4 vs 8.2 months; HR, 0.48 [95% CI, 0.30-0.76]) and ECOG 1 (10.9 vs 6.1 months; HR, 0.48 [95% CI, 0.31-0.74]) (Table). Additionally, rPFS favored rucaparib for ECOG 0 and ECOG 1 including those whose prespecified physician’s choice was ARPI or docetaxel (Table). The confirmed ORR (95% CI) of rucaparib vs physician’s choice for ECOG 0 was 54% (37%–69%) vs 16% (3%–40%); the ORR for ECOG 1 was 37% (22%–53%) vs 18% (5%–40%), respectively. Median OS was similar between arms for rucaparib or physician’s choice; median OS was 23.7 vs 21.7 months (HR, 0.91 [95% CI, 0.61–1.37]) for ECOG 0 and 21.8 vs 20.4 months (HR, 0.85 [95% CI, 0.58–1.24]) for ECOG 1. Conclusions: Rucaparib improved rPFS and ORR in pts with BRCA-mutated prostate cancer with an ECOG PS of 0 or 1. OS was similar for these groups in both arms. These data support rucaparib as a treatment option for pts with BRCA-mutated mCRPC, whether ECOG 0 or 1. Clinical trial information: NCT02975934 . Median rPFS in patients in the ECOG 0 and ECOG 1 groups. Rucaparib, n Physician’s choice, n rPFS Rucaparib, median months rPFS Physician’s choice, median months HR (95% CI) rPFS, ECOG 0 97 50 11.4 8.2 0.48 (0.30–0.76) rPFS, ECOG 1 104 51 10.9 6.1 0.48 (0.31–0.74) rPFS, physician’s choice ARPI, ECOG 0 51 24 11.2 5.6 0.42 (0.22–0.78) rPFS, physician’s choice docetaxel, ECOG 0 46 26 11.5 8.8 0.49 (0.24–0.98) rPFS, physician’s choice ARPI, ECOG 1 52 17 11.2 4.5 0.33 (0.16–0.68) rPFS, physician’s choice docetaxel, ECOG 1 52 34 10.7 7.1 0.59 (0.34–1.01)

392 Background: Cancer-associated fibroblasts (CAFs) are a major component of the tumor microenvironment, exerting both tumor-promoting and tumor-restrictive effects, and in some cases, dual roles. CAFs are heterogeneous and have been implicated in tumor progression, metastasis, and therapeutic resistance. Recent translational evidence suggests the existence of at least two CAF subtypes: CAV1+ CAFs and FAP+FN1+ CAFs. The clinical utility of subclassifying CAFs in prostate cancer remains undefined. In this study, we aimed to examine the correlation of CAF subtypes in localized prostate cancer with Decipher genomic classifier (GC) scores, Grade Group (GG), and the presence of adverse pathological features (APF). Methods: We retrieved de-identified transcriptomic and pathological data from 55,329 radical prostatectomy (RP) samples tested with the Decipher prostate genomic classifier (Veracyte, San Diego, CA) between 2016 and 2024, available through the GRID registry (NCT02609269). A subset of 855 patients (META855 cohort) treated with RP was utilized for outcomes analyses. CAF subtypes were defined based on the expression of key biomarkers (FN1, FAP, and CAV1), normalized to total CAF estimates. Correlation analyses and multivariable Cox regression models were performed to evaluate the association of CAF subtypes with GC scores, GG, APF. Metastasis-free survival (MFS) estimates were generated using Kaplan–Meier methodology. Results: FAP and FN1 normalized expressions were highly correlated (r=0.61) and weakly correlated with CAV1 suggesting that they represent two distinct subtypes of CAF. Samples with high expression (top 25%) of FN1 (FN1+) and FAP (FAP+) were enriched with very high Decipher (> 0.85, VHD) and GG4-5. As such, 37% and 32% of FAP+ were VHD and GG4-5 respectively compared to ~18% for low FAP(low 25%). While for FN+ 33.9% were VHD, and 31.5% GG4-5. Patients with FAP+FN1+, 42.5% and 37.1% were VHD, and had GG4-5. Samples with high CAV1 (CAV1+) were associated with lower rate of VHD (15.5%) or GG4-5 (20.8%) compared to low CAV1 (35.4% for VHD and 24.5% for GG4-5). When evaluating VHD in FAP+FN1+CAV1- subgroup, 58% were VHD compared to 6.4% in FAP-FN1-CAV1+. The 5-year MFS was 78% in FAP+FN1+CAV- compared to 96% in FAP-FN1-CAV1+ (p<0.01). When evaluating the biological implications of FAP+FN1+, we found that FAP and FN1 expression is highly correlated to angiogenesis and EMT pathways, radiation sensitivity signature (PORTOS). Conclusions: FAP+FN1+CAV- is a unique subset of CAFs that is associated with aggressive prostate cancer behavior. Our findings may lay the foundation for future efforts to modulate CAFs in localized prostate cancer with existing therapeutics targeting FAP or FN1.

382 Background: Androgen receptor signaling drives prostate cancer (PC) progression, and genes involved in androgen production, uptake, and conversion (APUC) are key therapeutic targets. We previously defined a 6-APUC gene set ( HSD3B1, HSD3B2, CYP3A43, CYP11A1, CYP11B1, and CYP17A1 ), that identified metastatic PC with favorable prognosis and predict docetaxel benefit. However, the expression landscape and clinical relevance of APUC genes have not been interrogated in localized PC. Methods: We analyzed the expression of 21 APUC genes in 55,329 radical prostatectomy tumor samples that underwent Decipher prostate genomic classifier (GC) testing (Veracyte Inc) between 2016-2024. Baseline clinical/pathologic factors were retrieved from Decipher GRID database (NCT02609269). Associations between gene expression quartiles with PathoGenomic Risk Factors (PGRF) including Very High Decipher score >0.85 (VHD), Grade Group (GG) 4-5, LNI and SVI were examined. Associations with distant metastatic (DM) outcomes and metastasis free survival (MFS) were evaluated using logistic regression in Mayo545 and Cox regression in Meta855 RP cohorts, respectively. Results: High expression of SRD5A2 was associated with favorable PGRF (lower odds ratio (OR), p<0.0001), lower rate of development of DM (OR 0.36, p<0.05), and longer MFS (Hazard ratio (HR) 0.51, p<0.05). HSD17B6 showed similar results for PGRF (lower OR, p<0.0001), DM (OR 0.58, p<0.05) and longer MFS (HR 0.39, p<0.05) despite being associated with GG4-5 and SVI+ in samples with VHD. In contrast, high expression of SRD5A3 and HSD17B10 had high OR for PGRFs (p<0.0001). High expression of SLCO2B1 , an androgen uptake gene, had the strongest association with GG4-5 and SVI/LNI (p<0.001 for both). We noted that SRD5A2 lacked correlation with SRD5A1 and SRD5A3 , which instead correlated with each other. Compared to all APUC genes, SRD5A2 had limited correlations with almost all APUC genes, including ones that exhibited robust cross correlation with each other . Similar to our prior finding, most APUC genes demonstrated negative correlation with AR expression and AR activity signatures. Key outcomes were validated based on transcriptomic data from 492 samples from TCGA database. Conclusions: Comprehensive analysis of 21 APUC genes in over 55,000 PC identified distinct, context-dependent associations with metastatic outcomes and clinical and molecular risk factors. Notably, high SRD5A2 expression was associated with lower odds of high-risk features and metastatic outcomes, whereas SRD5A1 and SRD5A3 linked to adverse outcomes, suggesting critical yet divergent roles among 5-alpha-reductase family members. These findings underscoring androgen-metabolic heterogeneity and highlighting SRD5A2 as a potential biomarker of indolent biology in localized prostate cancer.

369 Background: MRI-guided transurethral ultrasound ablation (TULSA) uses MRI to target, monitor, and control thermal ablation of prostate tissue with high intensity directional ultrasound, demonstrating favorable histologic control and preservation of genitourinary function in single-arm studies [1,2]. CAPTAIN (NCT05027477) is a post-market, multi-center randomized controlled trial (RCT) of TULSA vs RARP for intermediate-risk prostate cancer, and is the first such study to meet its enrolment target. Here we compare baseline patient characteristics, treatment parameters, and periprocedural outcomes. Methods: CAPTAIN enrolled men with treatment-naïve, organ-confined, Grade Group 2 or 3 prostate cancer at 23 centers including academic and private clinics in the USA (20), Canada (2), and Finland (1). Patients were randomized 2:1 to TULSA or RARP, without crossover. Endpoints include pad-free urinary continence (EPIC) and erections sufficient for penetration (IIEF Q2) at 1 year, salvage treatment at 3 years, and survival to 10 years. Periprocedural metrics include length of stay, blood loss, catheter time, and 30-day patient diary (EQ-5D-5L, NRS pain score). Results: 211 patients were treated from Jan 2022 to Aug 2025 (70% TULSA, 30% RARP), surpassing the enrolment target of 201. Arms had similar baseline characteristics: median (IQR) age for TULSA vs. RARP was 63 (58–68) vs 65 (60–69) years (p=0.12), PSA was 6.5 (4.9–9.6) vs 7.2 (5.6–9.7) ng/mL (p=0.63), Grade Group 2/3 proportions were 76%/24% vs 77%/23% (p=0.88). RARP involved bilateral, unilateral, or no nerve sparing in 89%, 5.7%, and 5.7% of cases. TULSA ablation plans included whole-gland (68%) and ≥hemi-ablation (32%). Median (IQR) blood loss was lower during TULSA vs RARP: 0 (0–0) vs 100 (100–200) mL (p<0.001). Hospital stay was shorter for TULSA: 0.29 (0.27–0.32) vs 1.24 (1.12–1.36) days (p<0.001). Catheter duration was longer post-TULSA: 13 (11–15) vs 8 (8–10) days (p<0.001). Patients reported lower NRS pain scores after TULSA to post-operative day 6, and less decline in overall health on the EQ-5D-5L 0-100 visual analog scale over 30 days (p<0.05). Fewer patients reported extreme inability on EQ-5D-5L after TULSA vs RARP over the first month for mobility (0% vs 15%), self-care (1% vs 18%), and usual activities (18% vs 38%) (all p<0.05). Conclusions: CAPTAIN is the first multicenter RCT comparing ablation to radical prostate cancer treatment to meet its enrollment target. Early perioperative results from CAPTAIN demonstrate that TULSA had no blood loss or overnight hospitalization, reduced post-procedural pain, and faster return to baseline activities and overall health. Data collection to 1 and 3 years post-treatment continues to readout of the primary safety and oncological outcomes. [1] Klotz et al 2021, JUrol 205(3):769-779. [2] Eggener et al 2024, UrolOnc 42:S83. Clinical trial information: NCT05027477 .

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)

357 Background: Prostate cancer (PCa) is the most incident cancer in adult males in the United States, and the second highest in the world. CTCAE is the standardized method to grade the severity of treatment-related adverse events (AEs), but are tedious to collect and subject to inter-observer variability. This study aimed to evaluate the performance of LLMs in automatically extracting and grading CTCAE toxicities from clinical notes and patient-reported outcomes of PCa patients from a clinical trial. Methods: A total of 55 patients from NCT02874014 trial, undergoing proton therapy (PT) for localized PCa, were included. The ground truth of CTCAE graded toxicities was obtained from the trial’s results. Clinical notes within 3 days, and the most recent patient questionnaires within 90 days of toxicity assessment were retrieved from the electronic health record. A comprehensive prompt was developed to extract and CTCAE grade toxicities from the retrieved records. LLMs used for this analysis were OpenAI GPT-4o and Gemini 2.0 Flash. Phase 1 evaluated both LLMs' performance against the ground truth. To ensure the metrics reflected the models’ performance based only on the information available in the text provided, cases where both LLMs contradicted the ground truth were manually reviewed. Phase 2 recalculated LLMs' performance metrics using this adapted ground truth. A conservative stepwise ensemble model was developed to leverage the complementary strengths of both LLMs. Results: In the conservative stepwise ensemble model, Gemini 2.0 Flash served as the initial screening platform to identify AEs in step 1. All cases with identified toxicity events were then reviewed and confirmed using OpenAI GPT-4o in step 2. This effectively reduced false positive rates while preserving high sensitivity. The ensemble model achieved strong performance metrics: binary accuracy (0.954), sensitivity (0.961), specificity (0.953), F1 score (0.836), and grade accuracy (0.932). Conclusions: This study demonstrates the feasibility of the use of LLMs for extraction and CTCAE toxicity grading in PCa patients treated with RT. The ensemble model achieved strong performance metrics. Further validation is needed for other cancer diagnosis and treatment modalities. Phase Model Binary Accuracy Precision Sensitivity Specificity F1 Score Grade Accuracy 1 Gemini 0.946 0.672 1.000 0.939 0.804 0.919 1 OpenAI 0.940 0.785 0.637 0.978 0.703 0.895 2 Gemini 0.956 0.735 1.000 0.950 0.848 0.927 2 OpenAI 0.954 0.917 0.679 0.991 0.780 0.903