Whole-genome HRD phenotyping as a predictor of PARP inhibitor benefit in first-line maintenance high-grade serous ovarian cancer.

Abstract

e17608 Background: Homologous recombination deficiency (HRD) in ovarian cancer identifies patients who may benefit from PARP inhibitor (PARPi) therapy beyond pathogenic germline or somatic BRCA1/2 mutations (g/s BRCA1/2 ). Despite increasing availability of whole-genome sequencing (WGS), the association between WGS-defined HRD (WGS-HRD) and PARPi maintenance outcomes in treated patients remains incompletely defined. Here, we evaluated PARPi outcomes in association with WGS-HRD and g/s BRCA1/2 profiling. Methods: We studied a prospective cohort of 84 patients with high-grade serous ovarian carcinoma. Each patient underwent matched tumor-normal WGS from FF (57%) or FFPE (43%) tissue and received PARPi maintenance as first-line (1LM) or second-line (2LM). Variant profiling and HRD phenotyping were performed by CancerVision (Inocras Inc.). Progression-free survival (PFS) was measured from PARPi start to clinical progression or censoring. Panel-based BRCA testing was available for comparison (n=50), and vendor scarHRD (n=35) was used to calibrate WGS-derived scarHRD. Results: Forty-seven patients (56%) received 1LM and 37 (44%) received 2LM. Median follow up was 46.6 months (range, 7.5-71.2) In the overall cohort, WGS-HRD(+) was associated with longer median PFS (mPFS) than WGS-HRD(-) (27.5 vs 12 months; HR=0.45, 95%CI 0.27-0.76; P<0.01). In 1LM, WGS-HRD(+) showed marked benefit (mPFS 44.2 vs 10.0 months; HR=0.36, 95%CI 0.16-0.76; P<0.01). In 2LM, PFS did not differ by WGS-HRD status (mPFS 17.4 vs 18.6 months; HR=0.68, 95%CI 0.32-1.44; P=0.31), suggesting attenuated predictiveness of baseline HRD phenotyping in later-line maintenance (potentially reflecting acquired resistance). WGS-based g/s BRCA1/2 calls were highly concordant with panel BRCA testing (accuracy 0.92, 95%CI 0.80-0.98; κ =0.83). g/sBRCA1/2 status alone was not significantly associated with PFS (mPFS 23.0 vs 20.2 months; HR=0.77, 95%CI 0.46-1.29; P=0.31); notably, 18/84 (21.4%) patients were BRCA-wildtype yet WGS-HRD(+), indicating that BRCA-only testing may not identify a clinically relevant subset of HRD-positive tumors. WGS-derived scarHRD was calibrated to third-party scarHRD calls, yielding a cutoff of 27 and moderate agreement (accuracy 0.74, 95%CI 0.57–0.88; κ =0.48). Unlike WGS-HRD, scarHRD was not significant in the overall cohort, but showed borderline stratification in 1LM (mPFS 33.8 vs 8.4 months; HR=0.45, 95%CI 0.18-1.15; P=0.09). In 2LM, scarHRD showed an opposite trend (HR=1.39, 95%CI 0.64-2.97; P=0.4), suggesting scarHRD is a less specific proxy for HRD than WGS-HRD. Conclusions: WGS-based HRD phenotyping stratifies PARPi maintenance benefit most strongly in the 1LM setting, with attenuated predictive value in 2LM. These findings support added clinical utility of WGS-HRD beyond BRCA-only testing to identify patients most likely to benefit from first-line PARPi maintenance.