Search

Introduction: Chromosomal instability (CIN) is a continuum of chromosomal gains and losses present in tumor cells. While increasing CIN is linked to poor patient outcomes in select cancers, the prognostic significance of CIN, and its association with molecular phenotypes and therapeutic response in advanced prostate cancer (PCa) is poorly understood. We characterize CIN in aggressive Pca by integrating DNA/RNA sequencing with detailed clinical information. Methods: From 4/2005-7/2021, 290 men with advanced PCa underwent IRB-approved tissue collection for tumoral RNA-sequencing and tumor/normal whole exome sequencing. DNA/RNA sequencing data for 328 libraries across 304 biopsies was processed using Turnkey Precision Oncology. CIN was measured using the weighted genome integrity index (wGII). Driver mutations (ETS fusions, SPOP, FOXA1 class 1, and CDK12 mutations), TP53 mutations and AR amplification status were determined. GSEA and EPIC were used for pathway/immune deconvolution. Associations with overall survival (OS) from the time of biopsy were determined via multivariable analysis (MVA). Results: 36 samples were from patients with localized disease (LD), 101 from metastatic hormone sensitive PCa (mHSPC) and 191 from metastatic castrate resistant PCa (mCRPC). wGII was associated with disease state: LD (median:0.28, IQR:[0.18-0.48]); mHSPCa (0.36, [0.21-0.58]); mCRPC (0.52, [0.35-0.62]) (p Conclusions: We identify CIN as one of the strongest genetic predictors of poor patient outcomes in metastatic PCa. Future research is needed to identify therapeutic vulnerabilities in this newly-described genetic subset of aggressive prostate cancers. Citation Format: Ryan J. Rebernick, Liat Hammer, Matthew McFarlane, Thomas Westbrook, Munna Hazime, Tanya Hammoud, Pin-en Chiu, Yi-Mi Wu, Dan Robinson, Daniel Eidelberg Spratt, Ajjai S. Alva, William C. Jackson, Zachery R. Reichert, Arul Chinnaiyan, Joshi J. Alumkal, Robert T. Dess, Marcin Cieslik. Association of chromosomal instability with poor prognosis in metastatic hormone sensitive and castrate-resistant prostate cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2023; Part 1 (Regular and Invited Abstracts); 2023 Apr 14-19; Orlando, FL. Philadelphia (PA): AACR; Cancer Res 2023;83(7_Suppl):Abstract nr 2244.

136 Background: rPFS is often used as an intermediate clinical endpoint (ICE) for overall survival (OS) in randomized trials in mCRPC. However, the current literature shows conflicting results on the surrogacy of rPFS for OS. Moreover, it remains unknown if TTrP, which does not consider death as an event, is an ICE for OS. We performed a combined analysis of COU-AA-302 and ACIS to determine if TTrP and rPFS can be used as ICE. Methods: In COU trial, docetaxel-naïve mCRPC patients were randomized to abiraterone (abi) versus placebo. In ACIS, a similar patient population was randomized to abi alone or abi with apalutamide (abi+apa). We applied weighted Cox regression models to evaluate the effect of treatment on TTrP and OS and used landmark analyses to determine the if the treatment effect on OS is mediated by that on radiographic progression. We estimated a semiparametric Spearman correlation between the ICE and OS at the patient level. We determined the trial level correlation of treatment effect on the ICE and OS in the 2 trials where each of them was subdivided into 9 pseudo-trial centers and then calculating the adjusted R2 between center level estimates of treatment effect for ICE and OS. The procedure of creating pseudo-trial centers was repeated 500 times and the presented R2 is the average across 500 repetitions after excluding those with negative association. Results: Overall, 2016 patients were eligible for this study – 1053 from COU and 963 from ACIS. Abi was associated with superior TTrP (HR 0.55 [95%CI 0.45-0.66]) and OS (HR 0.80 [0.70-0.92]). Similar results were seen with abi+apa (0.51 [0.41-0.64], 0.77 [0.65-0.91]). Radiographic progression was associated with significantly higher hazard of death in the state arrival extended Markov proportional hazard model (3.64 [1.54-8.62]) while longer TTrP was associated with reduced hazard of death (0.94 [0.93-0.95]). At the patient level, the correlation between TTrP & OS and rPFS & OS was 0.58 [0.54-0.63] and 0.68 [0.65-0.71], in the overall cohort. In the abi and abi+apa group, the correlation between TTrP and OS was 0.60 [0.53-0.66] and 0.73 [0.66-0.79] and that for rPFS and OS was 0.72 [0.67-0.75] and 0.79 [0.74-0.83], respectively. At the trial level, the treatment effect on rPFS & OS and TTrP & OS were correlated with average R2 of 0.84, 0.84, 0.85, and 0.86, respectively. The mean surrogate threshold effect over 500 permutations for HRrPFS and HRTTrP was 0.78 and 0.70 in ACIS and 0.54 and 0.45 in the COU-AA-302 trials, respectively. Conclusions: TTrP and rPFS were found to have significant association with OS in chemo-naïve mCRPC patients. We noted a modest to strong correlation between the treatment effect on both the ICE and OS at the trial level. Larger meta-analytic studies are needed to validate these findings.

391 Background: Event-free survival, a PSA-driven endpoint, was shown to not be surrogate endpoint for overall survival (OS) in the ICECAP two-stage meta-analytic approach. However, time to biochemical recurrence (TTBCR) in NRG/RTOG 9202 met Prentice criteria for surrogacy. We performed an individual patient data (IPD) meta-analysis of 11 randomized controlled trials evaluating RT dose escalation, ADT use, and adjuvant ADT prolongation to evaluate the surrogacy of time to BCR (TTBCR), censoring for non-prostate cancer deaths, using both approaches to evaluate surrogacy. Methods: This individual patient level meta-analysis was performed using data from the MARCAP consortium, and 11 radiotherapy trials were included. TTBCR was defined as time to developing a BCR or experiencing prostate cancer-specific mortality (PCSM), with censoring at time of other-cause death or loss to follow-up. Landmark analyses were used to test the Prentice criteria for surrogacy. For patient level correlation between TTBCR and OS, we applied a bivariate Copula model to estimate the Kendall’s τ. For trial level correlation of the treatment effect on TTBCR and true endpoints, a weighted linear regression model was applied between the effects of treatment (natural log of hazard ratio [log-HR]) on OS versus TTBCR using a weightage that was inverse variance of BCR log-HR estimate. Results: Based on Prentice criteria, BCR at the landmark time point of 48 months was associated with increased risk of mortality in trials that compared treatment intensification with adjuvant ADT prolongation (HR 2.18 [95% CI 1.95-2.42]), the addition of ADT (HR 1.38 [1.25-1.54]), and RT dose escalation (HR 2.12 [1.83-2.46]) on uni- and multi-variable analyses. At the patient level, there was a low to moderate level correlation between BCR and OS with Kendall’s τ of 0.34 and a R2 of 0.55 for correlation of treatment effect on TTBCR and OS. At the trial level, there was a poor correlation between treatment effect on TTBCR and OS (R2=0.16). Conclusions: This IPD meta-analysis demonstrates that while BCR is prognostic, it is not a surrogate endpoint for OS in localized prostate cancer for patients treated with a diverse array of radiotherapeutic strategies. This highlights the importance of other cause mortality in prostate cancer. Our results highlight the differences in interpretability of Prentice criteria and the two-stage meta-analytic approach and suitability of endpoints for clinical trial design.

263 Background: Prostate cancer (PCa) is characterized by considerable genetic heterogeneity, and complex genomic features may influence prognosis and treatment response. We created a database of aggressive PCa that integrates comprehensive genomic sequencing with detailed clinical outcomes to better understand the optimal use of genomic sequencing. Methods: From 4/2005-7/2021, PCa cancer patients older than 18 years of age underwent tissue collection for tumoral RNA-sequencing and tumor/normal whole exome sequencing at our institution (HUM00046018, HUM00048105, HUM00067928, SU2C). Genomic and transcriptomic sequencing data was processed using Turnkey Precision Oncology. Genetic alterations, including ETS fusions, SPOP, FOXA1 class 1, and CDK12 mutations, as well as TP53 and RB1 mutations were analyzed. Clinical data was collected from 05/2021-01/2022, and clinical associations (metastasis free survival (MFS), time to castrate resistant prostate cancer (CRPC), and overall survival (OS)) were determined. Results: Data was available for 325 men. Median follow up from diagnosis was 106 months (IQR, 90-121), median age at diagnosis was 61 (IQR, 54-67), and most (91%) presented with PCa adenocarcinoma (n=292/325). At diagnosis, 51% (n=165) had localized, 5% (n=18) had clinical node positive, and 40% (n=128) had de-novo M1 disease. At time of tissue sampling, 87% (n=283) had metastatic disease, and 59% (n=192) were castrate resistant. Established PCa driver mutations included 140 ETS fusions (49%), 26 SPOP mutations (9%), 22 FOXA1 class 1 mutations (8%), and 15 (5%) CDK12 mutations. For men with localized disease at diagnosis (n=197/325), ETS fusion was associated with improved MFS (HR: 0.55; 95% CI: 0.37-0.81), time to CRPC (HR: 0.53; 0.35-0.80), and OS (HR: 0.56; 0.35-0.89). SPOP mutations were also associated with improved prognosis in this population (n=197/325): MFS (HR: 0.45; 0.24-0.84), time to CRPC (HR: 0.36; 0.18-0.73), and OS (HR: 0.46; 0.21-0.99). TP53 mutations were identified in 38% (n=122) of all patients and were associated with worse OS from the time of biopsy after adjusting for PCa castration state and disease spread at biopsy (HR: 2.2; 1.7-2.9, p

241 Background: We previously reported on the use of gene expression profiling to characterize four primary subtypes in an analysis of over 100,000 prostate cancer primary tumors. Here we examine these subtypes and response to radiation (RT) after prostatectomy or chemotherapy in addition to androgen deprivation therapy (ADT) in hormone-sensitive metastatic prostate cancer (mHSPC). Methods: We examined a 215-gene prostate subtyping classifier (PSC) which groups luminal and basal gene expression profiles into one of four subtypes; luminal differentiated (LD), luminal proliferating (LP), basal immune (BI) and basal neuroendocrine-like (BN). The log-rank method was used to compare (i) metastasis-free survival between patients who did and did not receive salvage RT in the META855 cohort of localized disease patients treated with radical prostatectomy (n=855) and the (ii) overall survival of patients on the Phase III mHSPC EA3805 CHAARTED trial of ADT or ADT + docetaxel chemotherapy (n=160). Results: After radical prostatectomy, patients with BI tumors derived benefit from RT in terms of metastasis-free survival ( P=9.23e-4) while those with other subtype tumors did not (Each Log-rank P≥0.5). Patients with metastatic disease and LP tumors derived benefit from docetaxel (Log-rank P=0.002) while those with other subtypes tumors did not (Log-rank P=0.2). Conclusions: Basal-luminal subtyping segments tumors by relevant biological processes with potential implications for identifying patients who benefit from salvage radiation post prostatectomy and addition of docetaxel to ADT in mHSPC.

TPS398 Background: Androgen deprivation therapy (ADT) with gonadotropin-releasing hormone (GnRH) receptor agonists or antagonists are the foundation of advanced prostate cancer (aPC) treatment. Relugolix is an oral non-peptide GnRH receptor antagonist approved in the US for the treatment of aPC. In the phase 3 HERO trial, relugolix maintained suppression of testosterone to castration levels in 96.7% of men from Day 29 to 48 weeks, with superiority to leuprolide acetate. Relugolix was well tolerated and associated with a 54% lower risk of major adverse cardiovascular events relative to leuprolide acetate (Shore N, NEJM 2020;382;23). The goal of this current study (OPTYX) is to generate real-world evidence about the safety and effectiveness of relugolix in patients with prostate cancer during routine clinical care. Methods: This prospective observational cohort study (registry) is enrolling patients who have planned initiation with relugolix within 1 month prior to the time of study enrollment. The study is designed to prospectively understand the actual experience of patients with prostate cancer treated with relugolix through data collection on patients selected for treatment, treatment patterns, adherence, patient reported quality of life (QoL) and safety data. Additionally, a better understanding of the trajectory of patients with prostate cancer following cessation of relugolix therapy will be gained by observing the clinical course of the disease, health outcomes, and health-related quality-of-life (HRQoL) in these patients. Inclusion criteria include patients > 18 years old with prostate cancer and planned total treatment with relugolix of at least 4 months. Estimated enrollment is 1000 patients, and patients will be followed for up to 5 years. In addition to assessing routine clinical care and disease progression, patient reported outcomes will be assessed with the Functional Assessment of Cancer Therapy-Prostate (FACT-P) for HRQoL and Simplified Medication Adherence Questionnaire (SMAQ) for adherence at 3 and 6 months then every 6 months. Clinical trial information: NCT05467176 .

366 Background: The kinetics of testosterone recovery (TR) vary following cessation of androgen deprivation therapy (ADT) of various durations when given in combination with radiotherapy for prostate cancer. Time to TR will impact on quality of life. We aim to identify factors affecting the time to TR following ADT use. Methods: We included trials of prostate radiotherapy and ADT in the Meta-Analysis of Randomized trials in Cancer of the Prostate (MARCAP) consortium for which prospectively collected serial testosterone data is available. Time to non-hypogonadal TR (NHTR) (>8.0nmol/L) and time to full TR (FTR) (>10.5nmol/L) were estimated from the date of first available testosterone at trial enrolment to the date of TR using the Kaplan-Meier method. The effect of interactions between duration of ADT and patients’ age on TR was evaluated. Results: There were 1439 men with non-castrate testosterone at baseline (>1.7nmol/L) who met the inclusion criteria for analysis, of which 220, 765 and 454 men had 3-, 6-, and 18-months of ADT. There were 959 (67%) men who had FTR at last follow-up. For men who had 3-, 6-, and 18-months of ADT, the median time (range) to NHTR were 5.5 (1.6-76.3), 12.2 (0.8-53.6), and 30.1 (2.8-90.4) months respectively, while the median time (range) to FTR were 6.2 (1.8-75.7), 15.2 (0.8-86.0), and 36.0 (18.1-85.5) months respectively. In the subset of 1160 men who had normal testosterone at baseline (>10.5nmol/L), 851 (69%) men had FTR, with a median time (range) to FTR of 5.5 (1.8-75.7), 12.7 (1.8-86.0), and 30.8 (18.1-84.1) months for men who had 3-, 6- and 18-months ADT, respectively. For any given duration of ADT, men aged above 65 years were less likely to have FTR compared to those aged under 65 years – HR=0.67 (95%CI=0.46-0.99), HR=0.80 (95%CI=0.67-0.96), HR=0.64 (95%CI=0.51-0.81) for men who had 3-, 6-, and 18-months ADT respectively. There was no evidence of interaction between the effect of ADT duration on time to FTR and age (interaction P=0.3 for entire cohort). Conclusions: This is the largest pooled analysis of prospectively collected serial testosterone data from randomized trials, indicating substantial delay in FTR in men receiving longer durations of ADT. Approximately one-third of the men did not achieve FTR, which may have life-long impact on quality of life.

303 Background: Six months of a GnRH agonist with SRT is a standard of care for patients with unfavorable features and a detectable PSA post-RP. FORMULA-509 was designed to evaluate whether adding six months of AAP and Apa to this regimen could improve outcomes. Methods: FORMULA-509 (NCT03141671) is an investigator-initiated, multi-center, open-label, randomized trial. Patients had PSA≥0.1 post-RP and one or more unfavorable features (Gleason 8-10, PSA>0.5, pT3/T4, pN1 or radiographic N1, PSA doubling time 0.5 vs.≤0.5) and pN0 vs pN1. Analyses within these subgroups were pre-planned. Results: 345 participants (332 evaluable) from 9 sites were randomized from 11/24/2017 to 3/25/2020 (172 bicalutamide, 173 AAP/Apa). Median follow-up was 34 (6-53) months; 29% were pN1 and 31% had PSA >0.5 ng/mL. The HR for PFS was 0.71 (90% CI 0.49-1.03), stratified one-sided log-rank p=0.06 (3-year PFS was 68.5% bicalutamide vs 74.9% AAP/Apa). The HR for MFS was 0.57 (90% CI 0.33-1.01), stratified one-sided log rank p=0.05 (3-year MFS was 87.2% bicalutamide vs 90.6% AAP/Apa). In a pre-planned analysis by stratification factors, AAP/Apa was significantly superior for patients with PSA >0.5 for PFS [HR 0.50, (90% CI 0.30-0.86), p=0.03 (2-sided); 3-year PFS 46.8% bicalutamide vs. 67.2% AAP/Apa] and for MFS [HR 0.32 (90% CI 0.15-0.72), p=0.01 (2-sided); 3-year MFS 66.1% bicalutamide vs. 84.3% AAP/Apa.] No statistically significant benefit was detected in pre-planned analyses of stratification subgroups defined by PSA≤0.5, pN0, or pN1. Adverse events were consistent with the known safety profiles of the agents being studied, with more rash and hypertension in the AAP/Apa arm. Conclusions: Although this primary analysis did not meet the pre-specified threshold for statistical significance, it does strongly suggest that the addition of AAP/Apa to SRT+6 months of ADT may improve PFS and MFS, particularly in the subgroup of patients with PSA>0.5 where a pre-planned subgroup analysis by stratification factors observed a statistically significant benefit for both PFS and MFS. Clinical trial information: NCT03141671 .

299 Background: Recently, an MMAI prognostic biomarker, ArteraAI Prostate, was trained and validated in localized prostate cancer to more accurately risk stratify patients for multiple endpoints compared to NCCN risk groups (Esteva et al., 2022). Prognostication within an NCCN risk group remains clinically important given the multiple treatment decisions required within each risk group (e.g., radiotherapy dose or hormone therapy use). Herein, we validated the MMAI biomarker in high-risk prostate cancer where an increasing number of therapeutic decisions is required. Methods: This study leveraged histopathology image and clinical data from patients with at least one high-risk feature (HRF; cT3-cT4, Gleason 8-10, PSA > 20 ng/mL, primary Gleason pattern 5) from six NRG/RTOG phase III randomized trials (n=1,088). Patients from two trials not part of the initial MMAI biomarker training/validation (RTOG 0521 [n=344] and 9902 [n=318]) and the MMAI validation cohort (RTOG 9202, 9408, 9413, and 9910 [n=426]) were included. Fine-Gray, cumulative incidence, and time dependent area under the curve (tdAUC) analyses were performed for time to distant metastasis (DM) and prostate cancer-specific mortality (PCSM) for standard clinicopathologic variables (age, PSA, Gleason score, T-stage, number of HRFs) and the MMAI model, as a continuous score (per standard deviation increase) and categorically by quartile. Death from other causes were treated as competing risks. Results: The analyzed cohort had a median follow-up of 10.4 years. Median PSA was 21 ng/mL, 60% had Gleason 8-10 disease, 37% had cT3-T4 disease, and 20% were African American. On univariable analysis, the MMAI model was significantly associated with DM (subdistribution hazard ratio [sHR] 2.05, 95% CI 1.74-2.43, p

345 Background: Decipher is a tissue-based genomic classifier (GC) developed and validated in the post-radical prostatectomy (RP) setting to help guide adjuvant treatment decisions for prostate cancer (PCa). We conducted the first prospective, randomized trial assessing the impact of GC testing on adjuvant therapy use. Here, we determine the impact of GC testing on patient reported (PRO) quality of life outcomes (QoL) in men at high-risk of post-RP recurrence. Methods: The G-MINOR trial is a prospective, unblinded, randomized trial which enrolled 356 patients from 12 centers in the Michigan Urological Surgery Improvement Collaborative (MUSIC). Patients were enrolled between Aug 2016-July 2018. Eligible patients had undergone RP within 9 months of enrollment, had pT3-4 disease and/or positive surgical margins, and a post-RP PSA

108 Background: Artificial intelligence (AI) tools can display racial bias as a result of existing systemic health inequities and biased datasets. We have previously developed multi-modal AI (MMAI) prognostic models based on digital pathology images from five phase III randomized radiotherapy prostate cancer trials that outperform NCCN risk groups for prediction of distant metastasis (DM), biochemical failure (BF), prostate cancer-specific mortality (PCSM) and all-cause mortality (OS). In this study, we assessed the algorithmic fairness of the locked MMAI models between African American (AA) and non-AA populations in the five randomized trials. Methods: Patients enrolled in NRG/RTOG 9202, 9408, 9413, 9910, and 0126 with digitized biopsy histopathology slides were included in this study. The locked MMAI models were applied, and subgroup analyses were conducted by comparing distributions of clinical variables and MMAI scores (medians for continuous variables and proportions for categorical variables reported), and evaluating MMAI models’ prognostic ability among AA and non-AA men. The performance of the models were compared using DM as the primary endpoint and secondary endpoints of BF, PCSM, OS (death without an event as a competing risk) with Fine-Gray or Cox Proportional Hazards models. Either Kaplan Meier or cumulative incidence estimates were computed and compared using log-rank or Gray’s test. Results: This study included 5,624 men: 932 (17%) AA, 4503 (80%) white, and 189 (3%) other races. AA had younger median age (69 vs 71 year [yr]), higher median baseline PSA (12 vs 10 ng/mL), more T1-T2a (62% vs 57%), more Gleason < 7 (42% vs 36%) and 8-10 (15% vs 12%), and more NCCN low and high risk (12% vs 10% and 41% vs 33%). AA and non-AA had estimated 5-yr BF rates 27% and 27%, 5-yr DM rates 5% and 5%, 10-yr PCSM 5% and 7%, and 10-yr OS 58% and 60%, respectively. The median (interquartile range) score of the model optimizing for 5-yr DM (5-yr DM MMAI) was 0.044 (0.037–0.059) in AA and 0.043 (0.036–0.057) in non-AA. Similarly, all other MMAI models had differences in the medians between AA and non-AA ranging from 0.001 to 0.02. For all endpoints, the 5-yr DM MMAI model showed strong prognostic signal (hazard ratio [HR] per one standard deviation increase: 1.6 for DM, 1.4 for BF, 1.6 for PCSM and 1.3 for OS, all p-values < 0.001) and had comparable trends within AA vs. non-AA in the entire cohort (e.g., HR for DM 1.4 vs 1.6). Similar results were observed for the MMAI model optimizing for 10-yr PCSM. Conclusions: To our knowledge, this represents the first comparative analyses of a digital pathology AI prognostic model in AA vs. non-AA prostate cancer patients. The prognostic performance of the AI models was found to be comparable between subgroups. Our data supports the use of these models across racial groups, though further validation in AA cohorts is ongoing.

5039 Background: The Intermediate Clinical Endpoints (ICEs) in Cancer of the Prostate (ICECaP) working group identified metastasis-free survival as a valid surrogate endpoint for overall survival (OS) for patients with localized prostate cancer. No comparably validated surrogate endpoints for OS exist in advanced prostate cancer. Methods: In this meta-analysis, PubMed was searched for trials in advanced prostate cancer, defined as node positive (N1M0), metastatic castration-sensitive (mCSPC), non-metastatic (M0CRPC), or metastatic castration-resistant prostate cancer (mCRPC). Eligible randomized trials were required to report OS and ≥1 intermediate clinical endpoint (ICE). ICEs included biochemical-failure (BF), clinical failure (CF), BF-free survival (BFS), progression-free survival (PFS), radiographic PFS (radiographic +/- other study defined endpoints). Candidacy for surrogacy was assessed using the second condition of the meta-analytic approach, correlation of the treatment effect of the ICE and OS, using R2 weighted by the inverse variance of the log ICE hazard ratio and defined as an R2 > 0.70. Results: A total of 143 randomized trials (n = 75,601 patients) were included. No candidate endpoints met criteria for surrogacy; R2 BF (n = 28,922) 0.42 (95%CI 0.18-0.64), BFS (n = 25,741) 0.57 (95%CI 0.37-0.73), CF (n = 22,616) 0.31 (95%CI 0.0075-0.56), PFS (n = 52,639) 0.50 (95%CI 0.35-0.63), and radiographic PFS (n = 52,548) 0.50 (95%CI 0.35-0.63). Within preplanned subgroups by castration sensitive or resistant disease, or by treatment type, neither BFS nor PFS met criteria for surrogacy. When assessing radiographically-defined progression (exclusive or with clinical progression), PFS for the overall group and by castration status did not meet criteria for surrogacy. Sensitivity analyses demonstrated that candidacy for surrogacy of all endpoints tested did not change over time. Conclusions: Our aggregate screening method for surrogate endpoints in advanced prostate cancer demonstrated commonly used clinical endpoints are not valid surrogate endpoints for OS, and further composite endpoint construction is necessary.

e17006 Background: In 2021, national data between the Decipher 22-gene prognostic gene expression classifier (GC) for men with prostate cancer and the Surveillance, Epidemiology, and End Results (SEER) cancer registries were linked. The purpose of the work is to report on the linkage by characterizing national GC usage and its association with treatment decisions for men with prostate cancer. Methods: Patients in the SEER registries with primary prostate cancer diagnosis from 2010 to 2018 were included and linked to data from GC testing conducted between 2014 to 2020 (Veracyte, San Diego, CA). GC scores (range 0-1) and GC risk groups (low, intermediate, and high) were used for continuous and categorical analyses. Multivariable logistic regression was used to quantify the association between GC and active surveillance and watchful waiting (AS/WW) use and adverse pathology at radical prostatectomy (RP). Adverse pathology was defined as pathological grade group ≥3, pathological stage ≥pT3b, or lymph node invasion. Results: A total of 575,363 patients were eligible for analysis, of which 10,528 patients underwent GC testing (5,015 GC biopsy test, and 5,513 GC RP test). The median age was 67 for both tested and untested, but more white patients underwent testing (82% vs 76%, p < 0.001). For GC biopsy tested patients, AS/WW was highest for those with GC low risk results (41%) as compared to those with intermediate (32%) or high (17%) GC risk (p < 0.001). RP rates were lower in the tested compared to untested (25% vs. 36%, p < 0.001), and among the tested patients, RP use increased by GC risk group (19% of low, 25% of intermediate, and 34% of high GC risk, p < 0.001). A similar trend by GC risk group in management for radiation therapy was observed (13% of low, 19% of intermediate, and 29% of high GC risk, p < 0.001). In a multivariable logistic regression adjusted for age, race, NCCN risk group, and year of diagnosis, GC tested patients were more likely to undergo AS/WW compared to untested (OR 2.9 [95% CI, 2.8-3.1], p < 0.001). Within the subset of patients classified as NCCN low/favorable intermediate risk at biopsy and who were subsequently treated with RP (n = 594), GC high risk (> 0.6) was associated with more than 3 times the odds of harboring adverse pathology (OR 3.2 [95% CI 1.6-6.4], p < 0.001). Conclusions: Using the first ever linked SEER-Decipher data, we demonstrate that population-based treatment patterns are independently associated with GC test results. Patients with lower GC scores are independently more likely to undergo AS/WW; those with higher scores are more likely to have adverse pathology at time of RP.

269 Background: The 22-gene Decipher genomic classifier (GC) is a prognostic biomarker that has been validated in phase III trials in high-risk localized, post-prostatectomy, and metastatic and non-metastatic castration-resistant prostate cancer. Herein, we report the first validation of the biopsy GC in intermediate-risk prostate cancer from the phase III randomized trial NRG/RTOG 0126. Methods: After National Cancer Institute approval, biopsy slides were collected from the NRG biobank from RTOG 0126, a phase III randomized trial of men with intermediate-risk prostate cancer randomized to 70.2 Gy versus 79.2 Gy of radiotherapy without the use of concomitant hormone therapy. RNA was extracted from the highest grade tumor foci and processed through a quality control (QC) pipeline prior to generation of the previously locked 22-gene GC model. After GC data was generated it was linked with clinical outcomes to assess prognostic performance. The primary endpoint for this ancillary project was disease progression, defined as biochemical failure, local failure, distant metastasis or prostate cancer-specific mortality, as well as use of salvage therapy. Secondary endpoints included the previous individual endpoints, metastasis-free survival, and overall survival. Independent GC prognostic performance was assessed using cause-specific Cox or competing risk adjusted Fine-Gray multivariable models that included randomization arm and prognostic stratification factors. Death without events were treated as competing risks. Results: A total of 215 patient samples passed QC of the 449 that had suitable cDNA for expression analysis. The median follow-up was 12.8 years (range 2.4-17.7), and 61% had Gleason 3+4, 24% had Gleason 4+3, and the median PSA was 7.2 ng/mL (IQR 5.0-10.2). On multivariable analysis the 22-gene GC (per 0.1 unit) was independently prognostic for disease progression (subdistribution hazard ratio [sHR] 1.13, 95%CI (1.01-1.26), p = 0.03), biochemical failure (sHR 1.23, 95%CI 1.10-1.37, p < 0.001), distant metastasis (sHR 1.28, 95%CI 1.06-1.54, p = 0.01), and PCSM (sHR 1.45, 95%CI 1.20-1.76, p < 0.001). In patients with lower GC scores the 10-year distant metastasis rate difference between the 70.2 Gy and 79.2 Gy was 5%, as compared with 26% for higher GC patients. Conclusions: This study represents the first validation of any biopsy-based gene expression classifier in intermediate-risk prostate cancer. Decipher is independently prognostic and can identify patients that have low rates of metastatic events despite not receiving concurrent hormone therapy, and can be used to help personalize therapy in this setting. Clinical trial information: NCT00033631.

277 Background: The prognostic importance of local failure (LF) after definitive radiotherapy (RT) in patients with NCCN intermediate- (IRPCa) and high-risk prostate cancer (HRPCa) remains unclear, particularly given the likelihood that occult distant metastases (DMs) at presentation may be the true driver of natural history. Here, we leverage individual patient data (IPD) from 18 randomized control trials (RCTs) to evaluate the prognostic impact of LF and the kinetics of DM after RT. Methods: IPD for 18 RCTs were obtained from the Meta-Analysis of Randomized trials in Cancer of the Prostate (MARCAP) Consortium, comprising a total of 12533 patients (6288 HRPCa & 6245 IRPCa). Multivariable Cox proportional hazards (PH) models were developed to evaluate the relationship between overall survival (OS), PCa-specific survival (PCSS), DM-free survival (DMFS) & LF as a time-dependent covariate, adjusted for clinicodemographic parameters. Markov PH models, defined via transitions between 4 states, were developed to evaluate the aforementioned relationship. Proportional hazards assumption was imposed and examined for both models. Time is from randomization. Results: Median follow-up was 9.1 years. There were 795 (13%) LF & 1288 (21%) DM events for patients with HRPCa; these numbers were 449 (7%) & 451 (7%) for IRPCa. For HRPCa & IRPCa, 81% and 81% of DMs developed from a clinically relapse-free state (cRFS), with a median time of 46 and 60 months, respectively (p < 0.0001). 39% & 13% of DM events occurred within 2 years after RT for HRPCa & IRPCa, respectively. At later time points, DM events were more likely to emerge after an LF event for both HRPCa (9% vs. 34% between 0-2 vs. 8-10 years post-RT, p = 0.001) and IRPCa (10% vs. 34% between 0-2 vs. 8-10 years post-RT, p = 0.008). LF was significantly associated with OS (hazard ratio [HR] 1.17, 95% confidence interval [CI] 1.06–1.30), PCSS (HR 2.02, 95% CI 1.75-2.33) & DMFS (HR 1.94, 95% CI 1.75–2.15) (p < 0.01 for all) in patients with HRPCa. LF was also significantly associated with DMFS (HR 1.57, 95% CI 1.36–1.81) but not OS in patients with IRPCa. Patients who had not transitioned to the LF state had a significantly lower HR of transitioning to a PCa-specific death state than those who transitioned to the LF state (HR 0.32, 95% CI 0.21–0.50, p < 0.001). Conclusions: LF is an independent prognosticator of OS, PCSS & DMFS in HRPCa and of DMFS in IRPCa. The predominant mode of DM development is from the cRFS state, underscoring the importance of accurate upfront staging & systemic therapy. However, particularly at late time points, an increasing proportion of DM events originated after diagnosis of a LF, constituting a “second wave” of DM events. This suggests that optimizing local control is also important, though the majority of DM events appear prior to a clinically-detected LF.

222 Background: Prognostication in localized prostate cancer is reliant on non-specific tools, an issue that leads to the over- and under-treatment of patients. Various tissue-based molecular biomarkers have attempted to fill this unmet need, but most lack prospective randomized trial validation. Herein, we train and validate prognostic biomarkers in localized prostate cancer using five phase III randomized trials, by leveraging multi-modal deep learning on digital histopathology. Methods: Histopathology image data was generated from pre-treatment biopsy slides in five NRG Oncology phase III randomized radiotherapy prostate cancer trials (RTOG 9202, 9408, 9413, 9910, and 0126). The trials were randomly split into training (80%) and validation (20%) cohorts. A multi-modal artificial intelligence (MMAI) architecture was developed to take clinicopathologic and image-based (histopathology) data as input and predict binary outcomes. Using this architecture, various models were trained to predict relevant clinical endpoints: biochemical recurrence (BCR), distant metastasis (DM), prostate cancer-specific survival (PCaSS), and overall survival (OS). These models were then validated for measures of prognostic discrimination using the time-based area under the curve (AUC) method. Results: Clinicopathologic and histopathology image data was available for 5,654 of 7,957 eligible patients (71.1%), yielding 16.1 TB of data from 16,204 histopathology slides of pretreatment biopsy samples. After training the models, locking them, and evaluating them on the validation cohort, we found that the MMAI prognostic model had superior discrimination compared to the NCCN model (PSA, T-stage, and Gleason score) for 5-year DM (AUC of 0.84 vs 0.73), 5-year BCR (AUC of 0.69 vs 0.58), 10-year PCaSS (AUC of 0.79 vs 0.66), and 10-year OS (AUC of 0.65 vs 0.58). Within each of the individual trials in the validation cohort, the MMAI-model had superior performance compared to NCCN risk groups for all clinical endpoints. Conclusions: This represents the first ever development and validation of prognostic biomarkers in localized prostate cancer using multiple large phase III clinical trials. We have successfully validated that our MMAI-prognostic biomarkers are superior to standard clinical and pathologic variables in identifying future BCR, DM, PCaSS, and OS. This massively scalable technology is feasible and can help personalize the management of prostate cancer patients. Funding: This project was supported by grants U10CA180868 (NRG Oncology Operations), U10CA180822 (NRG Oncology SDMC), UG1CA189867 (NCORP), U24CA196067 (NRG Specimen Bank) from the National Cancer Institute (NCI).

223 Background: The current standard of care for men with intermediate- and high-risk localized prostate cancer treated with radiotherapy (RT) is the addition of androgen deprivation therapy (ADT). Presently, there are no validated predictive biomarkers to guide ADT use or duration in such men. Herein, we train and validate the first predictive biomarker for ADT use in prostate cancer using multiple phase III NRG Oncology randomized trials. Methods: Pre-treatment biopsy slides were digitized from five phase III NRG Oncology randomized trials of men receiving RT with or without ADT. The training set to develop the artificial intelligence (AI)-derived predictive biomarker included NRG/RTOG 9202, 9413, 9910, and 0126, and was trained to predict distant metastasis (DM). A multimodal deep learning architecture was developed to learn from both clinicopathologic and digital imaging histopathology data and identify differential outcomes by treatment type. After the model was locked, an independent biostatistician performed validation on NRG/RTOG 9408, a phase III randomized trial of RT +/- 4 months of ADT. The DM rates were calculated using cumulative incidence functions in biomarker positive and negative groups, and biomarker-treatment interaction was assessed using Fine-Gray regression such that death without DM was treated as a competing event. Results: Clinical and histopathological data was available for 5,654 of 7,957 eligible patients (71.1%). The training cohort included 3,935 patients and had a median follow-up of 13.6 years (IQR [10.2, 17.7]). After the AI-derived predictive ADT classifier was trained, it was validated in NRG/RTOG 9408 (n = 1719, median follow-up 17.6 years, IQR [15.0, 19.7]). In the NRG/RTOG 9408 validation cohort that had digital histopathology data, ADT significantly improved DM (HR 0.62, 95% CI [0.44, 0.87], p = 0.006), consistent with the published trial results. The biomarker-treatment interaction was significant (p-value = 0.0021). In patients with AI-biomarker positive disease (n = 673, 39%), ADT had a greater benefit compared to RT alone (HR 0.33, 95% CI [0.19, 0.57], p < 0.001). In the biomarker negative subgroup (n = 1046, 61%), the addition of ADT did not improve outcomes over RT alone (HR 1.00, 95% CI [0.64, 1.57], p = 0.99). The 15-year DM rate difference between RT versus RT+ADT in the biomarker negative group was 0.3%, vs biomarker positive group 9.4%. Conclusions: We have successfully validated in a phase III randomized trial the first predictive biomarker of ADT benefit with RT in localized intermediate risk prostate cancer using a novel AI-derived digital pathology-based platform. This AI-derived predictive biomarker demonstrates that a majority of patients treated with RT on NRG/RTOG 9408 did not require ADT and could have avoided the associated costs and side effects of this treatment.

133 Background: The genomic classifier, Decipher, has been validated to predict risk of metastasis after radical prostatectomy (RP). However, the cohort size and event rate in the previous studies did not allow for a thorough investigation into performance within individual clinicopathologic or treatment subgroups. In this study, we present the first meta-analysis of the performance of the 22-marker genomic classifier in men with prostate cancer (PCa) post-RP. Methods: MEDLINE, EMBASE, and the Decipher genomic resource information database were searched for published reports of men with PCa treated by RP between 2010 and 2016 where the benefit of the Decipher genomic classifier test was assessed. The primary end point was the ability of Decipher to independently improve prognostication of regional or distant metastasis over routine clinicopathologic factors. Meta-analysis was performed with random-effects modeling, and extent of heterogeneity between studies was determined with the I2 test. Results: Five studies (975 total patients, and 855 with individual patient genomic and clinicopathologic data) were eligible for analysis. The median follow-up was 8 years. All patients had clinical high-risk disease, yet 60.9%, 22.6%, and 16.5% of patients were classified as low, intermediate, and high-risk, respectively by Decipher and had 10-year cumulative incidence rates of metastases of 5.5%, 15.0% and 26.7% (p < 0.001), respectively. Adjusting for standard clinicopathologic variables, on multivariable analysis Decipher remained a statistically significant predictor of metastasis (hazard ratio [HR] 1.30 per 0.1 unit, 95% confidence interval [CI] 1.14-1.47, p < 0.001), and the summary HR for metastasis of Decipher across the 5 studies was 1.52 (95% CI 1.39-1.67) per 0.1 unit. Conclusions: The genomic classifier test, Decipher, has the ability to independently improve prognostication of men post-RP, as well as within nearly all clinicopathologic and treatment subgroups. Strong consideration should be given to incorporating the use of genomic testing in clinical decision making and clinical trials to better individualize treatment.