Leveraging tumor multigene panel testing to identify germline variants in gynecologic cancers: A retrospective evaluation of an algorithm-based approach

Abstract

Objective. To evaluate the recall rate and positive predictive value of a novel algorithm for recommending germline testing in women with variants identified through tumor gene panel testing. Methods. A retrospective analysis was performed in patients with gynecological cancers who underwent TruSight Oncology 500 tumor gene panel testing (Illumina, San Diego, CA, USA) between September 2020 and July 2023 at two tertiary institutions in Korea. The algorithm recommends germline testing for patients who meet the following criteria: (1) presence of Tier 1 or 2 variants in any of 11 oncogenes (BRCA1, BRCA2, PALB2, MLH1, MSH2, MSH6, BRIP1, RAD51C, RAD51D, ATM, and CHEK2) with a variant allele frequency (VAF) of >= 40 %; and (2) exclusion of epithelial ovarian cancers with BRCA variants. Medical records were reviewed to assess germline testing results and their concordance with the somatic findings. Results. The algorithm recommended germline testing for 19 of the 702 patients (recall rate: 2.7 %). Of these, four patients underwent germline testing based on tumor-detected variants. All four patients were confirmed to have germline variants identical to their somatic findings (positive predictive value: 100 %). Specifically, germline mutations were confirmed in one endometrial cancer (endometroid) patient with a BRCA variant, one tuboovarian cancer (high grade serous) patient with a RAD51D variant, and two tubo-ovarian cancers (high grade serous) patients with BRIP1 variants. Conclusion. A conservative (specificity-focused) tumor-guided algorithm demonstrated high positive predictive value for identifying germline pathogenic variants and may provide a cost-effective means of prioritizing genetic counselling and germline testing in resource-limited settings. (c) 2026 Elsevier Inc. All rights are reserved, including those for text and data mining, AI training, and similar technologies.