Next-generation sequencing of circulating tumor DNA (ctDNA) is a non-invasive method to guide therapy selection for cancer patients. ctDNA variant allele frequency (VAF) is commonly reported and may aid in discerning whether a variant is germline or somatic. We report on the fidelity of VAF in ctDNA as a predictor for germline variant carriage. Two patient cohorts were studied. Cohort 1 included patients with known germline variants. Cohort 2 included patients with any variant detected by the ctDNA assay with VAF of 40–60%. In cohort 1, 36 of 91 (40%) known germline variants were identified through ctDNA analysis with a VAF of 39–87.6%. In cohort 2, 111 of 160 (69%) variants identified by ctDNA analysis with a VAF between 40 and 60% were found to be germline. Therefore, variants with a VAF between 40 and 60% should induce suspicion for germline status but should not be used as a replacement for germline testing.

PURPOSE Identification of incidental germline mutations in the context of next-generation sequencing is an unintended consequence of advancing technologies. These data are critical for family members to understand disease risks and take action. PATIENTS AND METHODS A retrospective cohort analysis was conducted of 1,028 adult patients with metastatic cancer who were sequenced with tumor and germline whole exome sequencing (WES). Germline variant call files were mined for pathogenic/likely pathogenic (P/LP) variants using the ClinVar database and narrowed to high-quality submitters. RESULTS Median age was 59 years, with 16% of patients ≤ 45 years old. The most common tumor types were breast cancer (12.5%), colorectal cancer (11.5%), sarcoma (9.3%), prostate cancer (8.4%), and lung cancer (6.6%). We identified 3,427 P/LP variants in 471 genes, and 84% of patients harbored one or more variant. One hundred thirty-two patients (12.8%) carried a P/LP variant in a cancer predisposition gene, with BRCA2 being the most common (1.6%). Patients with breast cancer were most likely to carry a P/LP variant (19.2%). One hundred ten patients (10.7%) carried a P/LP variant in a gene that would be recommended by the American College of Medical Genetics and Genomics to be reported as a result of clinical actionability, with the most common being ATP7B (2.7%), BRCA2 (1.6%), MUTYH (1.4%), and BRCA1 (1%). Of patients who carried a P/LP variant in a cancer predisposition gene, only 53% would have been offered correct testing based on current clinical practice guidelines. Of 471 mutated genes, 231 genes had a P/LP variant identified in one patient, demonstrating significant genetic heterogeneity. CONCLUSION The majority of patients undergoing clinical cancer WES harbor a pathogenic germline variation. Identification of clinically actionable germline findings will create additional burden on oncology clinics as broader WES becomes common.

e15555 Background: Next-generation sequencing (NGS) of circulating tumor DNA (ctDNA) is a non-invasive method to guide therapy selection for cancer patients. Identification of inherited germline cancer predisposition mutations that have significant implications for at-risk relatives may be missed during routine ctDNA testing. Allele frequency has the potential to enhance the likelihood that a mutation is germline; and is often reported in many NGS tests from ctDNA. Here, we report on the fidelity of allele frequency in ctDNA as a predictor for pathogenic germline variant carriage. Methods: ctDNA sequencing of patients with metastatic cancer from the Indiana University Health Precision Genomics Program was performed using the FoundationOne Liquid assay. All variants detected by the ctDNA assay report were considered. All patients also had germline testing information and pathogenicity of germline variants were determined using ClinVar. Germline variants with conflicting interpretations were manually reviewed to determine pathogenicity. Comparisons between ctDNA results with known germline status were performed. Results: Of 91 previously identified germline cancer predisposition variants, 36 (40%) were also identified by ctDNA analysis. All germline variants that were tested for in the ctDNA assay (n = 36, 100%) were identified. When detected, the allele frequencies of detected germline variants in the ctDNA ranged from 39-87.6% with an average of 52.1%. Conversely, 111 of 160 (69%) variants identified by ctDNA analysis with allele frequency between 40-60% in a cancer predisposition gene were found to be germline in origin (regardless of pathogenicity). Variants in the BRCA2, BRCA1, and CDH1 genes were most likely to be germline in origin (26/27 [96%], 20/22 [91%], 13/15 [87%], respectively). Variants in the TP53 and APC genes were least likely to be germline in origin (9/36 [25%] and 1/6 [17%], respectively). There was an 85% (95/111) concordance in actionability between the somatic testing lab and ClinVar germline classifications. Of the 16 discordant variants, 100% were determined to be actionable by the somatic testing lab but not actionable in ClinVar. Conclusions: ctDNA allele frequency can alter the likelihood that a variant is germline. Importantly, however, this testing is far from comprehensive and should not be used as a replacement for germline testing. Variants with allele frequency between 40-60% in cancer predisposition genes should induce a high level of suspicion for germline status.