Circulating tumor DNA–based clinical-genetic (CG) prognostic model for radiographic progression-free survival (rPFS) in patients (pts) with metastatic castration-resistant prostate cancer (mCRPC): Analysis of Alliance A031201.

Abstract

5018 Background: mCRPC is characterized by marked molecular heterogeneity and variable clinical outcomes. Circulating tumor DNA (ctDNA) profiling provides a noninvasive approach to capture tumor genomic alterations with potential prognostic value. We developed and validated a ctDNA-based prognostic model for rPFS using data from the Alliance A031201 (NCT01949337) trial, distinct from our published CG model of overall survival (Halabi et al. Eur Urol 2025). Methods: We analyzed ctDNA from 776 pts enrolled in the A031201 trial that randomized men with chemotherapy-naïve mCRPC to enzalutamide +/- abiraterone acetate and prednisone. Baseline cell free DNA samples from 776 pts were sequenced using the AR-ctDETECT assay. The primary endpoint of this analysis was rPFS. Proportional hazards model was used to assess the association of each genetic alterations and clinical variables with rPFS. Random survival forest (RSF) incorporating CG and clinical (C) variables only were trained and evaluated using 10-fold cross-validation. Model performance was assessed using integrated time-dependent area under the ROC curve (itAUC) and net reclassification improvement (NRI). Results: Higher ctDNA aneuploidy fraction, ctDNA positivity, and multiple pathogenic genomic alterations were associated with worse rPFS. Among prevalent alterations, AR enhancer gain and AR gain were strongly prognostic, with median rPFS of 13.7 vs 27.3 months (mos) and 13.5 vs 27.0 mos for pts with and without alterations, respectively. The rPFS prognostic model included gains in AR, AR enhancer, MYC, CCND1, and FOXA1 , and losses in PTEN, TP53, and LRP1B based on RSF . PTEN loss, gains in AR enhancer, MYC, along with hemoglobin, PSA and alkaline phosphatase showed the largest contribution to risk prediction. Mean itAUC for C model was 0.66 (95% confidence interval [CI] 0.62-0.70), mean itAUC for CG model was 0.73 (95% CI: 0.69–0.77). At 22 mos (median rPFS), NRI comparing CG model with C model was 0.30 (95% CI: 0.19-0.36). Pts were stratified into low-, intermediate-, and poor-risk groups and showed markedly distinct rPFS outcomes, with median rPFS of 39.2, 25.2, and 12.8 mos, respectively. The hazard ratios for low vs poor risk was 0.25 (95% CI: 0.20–0.31) and for intermediate vs poor risk 0.46 (95% CI: 0.38-0.55). Conclusions: A ctDNA-based prognostic model integrating CG features robustly stratifies rPFS risk in pts with mCRPC treated with first line AR pathway inhibitor therapy. This approach may support individualized risk assessment and inform trial design around treatment intensification combinations and therapeutic decision-making. External validation in independent cohorts is warranted. Support: U10CA180821, U10CA180882; R01CA256157; R01CA174777; https://acknowledgments.alliancefound.org. Clinical trial information: NCT01949337 .