Real-world effectiveness of PARP inhibitors (PARPi) in metastatic castration-resistant prostate cancer (mCRPC) by genomic homologous recombination repair (HRR) alterations and homologous recombination deficiency signature (HRDsig).

Abstract

186 Background: Multiple randomized controlled trials have demonstrated efficacy of poly ADP-ribose polymerase (PARP) inhibitors in patients with metastatic castration resistant prostate cancer (mCRPC) harboring deleterious BRCA1 or BRCA2 alterations (BRCAalt). Alterations in other DNA HRR gene pathways are common and the efficacy of PARPi in these populations is less clear. We sought to evaluate the real-world effectiveness of PARPi in commonly defined HRR groups and explore clinical validity of HRDsig to additionally predict outcomes on PARPi. Methods: This study used the nationwide (US-based) de-identified Flatiron Health-Foundation Medicine Prostate Cancer clinico-genomic database (FH-FMI CGDB), originating from approximately 280 US cancer clinics (~800 sites of care). Patients with mCRPC and tumor genomic profiling (tissue or liquid) who underwent single agent PARPi treatment were included, grouped by biomarkers with deleterious alterations detected: BRCA, ATM, other HRR ( ATR, ATRX, BAP1, BARD1, BRIP1, CHEK1/2, CDK12, FANCA, FANCL, MRE11, RAD51B/C/D, RAD54L, PALB2), or no HRR. Separately, patients with mCRPC and tumor tissue profiling were grouped by HRDsig status (+ vs -) using a machine learning- based algorithm built using a diverse set of copy number features. Kaplan Meier estimates and multivariable Cox PH models assessed time to next therapy (TTNT), time to treatment discontinuation (TTD), and real-world overall survival (rwOS). Results: 313 patients with mCRPC met inclusion criteria, of which 162 had tissue specimens for genomics evaluations and 151 had liquid profiling. In the tissue cohort, BRCAalt had more favorable TTNT (median 10.0 months, 95% CI 8.0 – 14.1) relative to ATM (median 3.8 months, 95% CI 2.7 – 6.9), other HRR (median 4.9 months, 95% CI 3.6 – 7.2) and no HRR (median 4.9, 95% CI 3.6 – 7.2 months). Other proxies for drug effectiveness (TTD, PSA kinetics) and rwOS had consistent results. More granular mutational analyses and liquid profiling data will be presented at the conference. Among evaluable patients without BRCAalt on tissue biopsy receiving PARPi, 5 of 75 (7%) were HRDsig(+) and had directionally more favorable (vs. HRDsig(-)) TTNT (HR: 0.39, 95%CI 0.14 – 1.12), TTD (0.46, 95%CI 0.16 – 1.31) and rwOS (HR: 0.27, 95%CI 0.07 – 1.14). Conclusions: We observe no significant outcomes difference between non- BRCAalt groups (defined as ATM, other HRR, and no HRR) in this cohort with respect to proxies of drug effectiveness. These results are largely consistent with biomarker-defined subgroup analyses from completed clinical trials. However, HRDsig may be able to identify a small non-BRCAalt subgroup with enhanced benefit. Associations of BRCAalt(-)/HRDsig(+) and PARPi performance deserve further attention in additional cohorts.