Circulating tumor DNA fraction (ctDNA%) to independently predict for clinical outcomes in patients (pts) with metastatic castration-resistant prostate cancer (mCRPC).

Abstract

5049 Background: CtDNA% (the tumour-derived proportion of cell-free DNA (cfDNA)) is abundant in >60% of mCRPC pts and associates with adverse clinical prognostic factors. However, prognostic associations have not been comprehensively tested across clinical contexts. We evaluated the utility of ctDNA% as an independent prognostic biomarker in patients with mCRPC prior to first-line (1L) therapy. Methods: 410 treatment-naïve mCRPC pts had blood samples drawn prior to 1L therapy and followed prospectively for outcomes. Plasma cfDNA was subjected to deep targeted sequencing and ctDNA% was calculated using validated methods ( Annala, Cancer Discov, 2018 ). Overall survival (OS), PSA progression free survival (PSA PFS) and PSA declines ≥50% from baseline (PSA50 response rate (RR)) were stratified by ctDNA% and compared using Kaplan-Meier and Cox proportional hazards analysis. Results: Median age was 73 yrs. (range 45-98), the majority of pts had ECOG PS 0-1 (78%) and 9.5% had liver metastases at baseline. The most common 1L therapy employed was androgen receptor pathway inhibitors (90%). Median follow-up was 21 mo. (range 1-75) and median ctDNA% was 4.9% (range: 0-89%). Stratifying patients into high ctDNA (>30%) and Low ctDNA (≤2%) groups showed stronger association with OS and PSA PFS than grouping by median (Table). In a univariate comparison to pts with low ctDNA (≤2%), pts with high ctDNA% (>30%) had significantly shorter median PSA PFS, median OS and a lower PSA50 RR (Table). In a multivariable adjustment for clinical prognostic factors and cfDNA concentration, high ctDNA% remained strongly associated with OS (HR= 3.3, 95%CI: 2.1-5.3, p<0.001) and PSA PFS (HR: 3.7, 95%CI: 2.4-5.9, p<0.001). Although ctDNA% and total cfDNA concentration were correlated (R2=0.55), association with OS was stronger for ctDNA% than cfDNA concentration (stratified at median; HR: 2.9 (2.3-3.7), p<0.001 vs HR: 2.1 (1.7-2.6), p<0.001). Conclusions: In a large cohort of treatment-naïve mCRPC pts, ctDNA% prior to 1L treatment provided strong prognostic information independent of known clinical factors. These data further demonstrate the multipronged clinical utility of ctDNA-based profiling for actionable genomic alterations.[Table: see text]