Introduction/Background Various biomarkers have been investigated to identify patients likely to respond to PARP inhibition. PARP inhibitor olaparib plus bevacizumab is approved by the US FDA as maintenance therapy for homologous recombination deficiency (HRD)–positive advanced ovarian cancer; the FDA contemporaneously approved a commercial assay as a companion diagnostic for HRD assessment that includes a genomic instability biomarker. Other genomic platforms measuring HRD are available or in development, including single-nucleotide polymorphism (SNP) genotyping arrays designed to measure tumour-related copy number changes. We evaluated the performance of OncoScan™ (ThermoFisher) and Infinium CytoSNP-850K (CytoSNP; Illumina) for assessing HRD genomic instability.

Methodology DNA extracted from pretreatment archival tumour samples (N=126 across 20 indications) was evaluated with Oncoscan™, CytoSNP and an FDA-approved HRD test. ASCAT (v2.5.1), using log R ratio and B-allele frequency of autosomal markers with GC wave correction, was used to evaluate copy number variation (CNV) and loss of heterozygosity (LOH). The genomic metrics were further generated with default parameters using previously reported algorithms1 for LOH,2 number of telomeric-allelic imbalance (NTAI)3 and large-scale state transition (LST)4; the aggregate HRD metric was the sum of the three components. The association between genomic metrics (with BRCA deleterious alterations) and an FDA-approved HRD test metric (dichotomised at clinical cutoff) was calculated using AUROC. Correlations among continuous metrics were assessed using Spearman rank correlation coefficients.

Results CNV segmentation and genomic metrics were successfully calculated for 120 (Oncoscan™), 106 (CytoSNP) and 126 (FDA-approved test) samples. Assessed by SNP genotyping arrays, the genomic metric as a continuous variable demonstrated good association with deleterious BRCA alterations (AUROC of HRD: Oncoscan™, 0.87; CytoSNP, 0.75) (table 1) and the FDA-approved test at cutoff 42 (AUROC of HRD: Oncoscan™, 0.92; CytoSNP, 0.91) (table 2). The genomic metric as a continuous variable showed good correlation with the FDA-approved HRD test metric (Spearman correlation of HRD: Oncoscan™, 0.82; CytoSNP, 0.81). The Spearman correlation of genomic metrics with the FDA-approved HRD test metric was 0.68 (LOH), 0.76 (TAI), 0.78 (LST) and 0.82 (HRD) for Oncoscan™ and 0.59 (LOH), 0.77 (TAI), 0.82 (LST) and 0.81 (HRD) for CytoSNP.

Conclusion HRD as a continuous variable assessed by SNP genotyping arrays showed good correlation with an FDA-approved HRD test metric; SNP assays may potentially be able to identify most HRD-positive tumours if appropriate clinically relevant cutoffs can be determined.

Disclosures This study was sponsored by Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.

Razvan Cristescu is an employee of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, and has stock ownership interests in Merck & Co., Inc., Kenilworth, NJ, USA.

Xiao Qiao Liu is an employee of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.

Gladys Arreaza is an employee of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, and has stock ownership interests in Merck & Co., Inc., Kenilworth, NJ, USA.

Cai Chen is an employee of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA.

Andrew Albright is an employee of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, and has stock ownership interests in Merck & Co., Inc., Kenilworth, NJ, USA.

Ping Qiu is an employee of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, and has stock ownership interests in Merck & Co., Inc., Kenilworth, NJ, USA.

Matthew Marton is an employee of Merck Sharp & Dohme Corp., a subsidiary of Merck & Co., Inc., Kenilworth, NJ, USA, and has stock ownership interests in Merck & Co., Inc., Kenilworth, NJ, USA.