Interpretation of constitutional cancer predisposition gene variants in 14 765 individuals in the 100 000 Genomes Project cancer arm: a retrospective cohort analysis

BACKGROUND: Cancer predisposition due to constitutional (germline) genetic variants in high-risk or moderate-risk cancer predisposition genes presents clinical opportunities for risk mitigation. Focusing genetic testing only on patients who are most likely to be positive for germline variants might enhance the clinical utility of positive results, but this approach could fail to assess the rate and pattern of such variants in patients with cancer overall. We aimed to assess the frequency and nature of constitutional variants in cancer predisposition genes in patients with cancer in the UK health-care system. METHODS: In this retrospective cohort study, we analysed data from participants with cancer from the 100 000 Genomes Project research environment. We included participants recruited from 14 genomic medicine centres operating within the UK National Health Service. Cancer predisposition genes (n=109) were curated to include those for which pathogenic variants were consistent with neoplasia as a primary presenting feature. We developed a detailed, semi-automated workflow for variant interpretation based on American College of Medical Genetics and Genomics guidance. This workflow used cancer and gene-specific adaptions developed by the Cancer Variant Interpretation Group UK and the US-based resource ClinGen. Multiple external reference sources (including the databases ClinVar and Gnomad, as well as functional studies) were used to inform the application of the guidance in the assessment of variants as pathogenic or likely pathogenic. Tumour type was assessed to ascertain associations between tumours and pathogenic or likely pathogenic variants. Fisher's exact tests were used for some elements of variant assessment. FINDINGS: 14 765 participants recruited into the 100 000 Genomes Project between Nov 29, 2016, and Feb 26, 2020, were included in the study (8315 [56%] female and 6450 [44%] male). Ancestry designation was European in 13 219 (90%) participants. The most frequent cancer types were breast (2908 [20%]), colorectal (2605 [18%]), and lung (1522 [10%]). Mean age at diagnosis was 62·9 years (SD 16·1). 711 (5%) participants were identified with a cancer predisposition gene variant assessed as pathogenic or likely pathogenic, with a total of 727 pathogenic or likely pathogenic variants. The genes with the most frequently detected pathogenic or likely pathogenic variants were CHEK2 (121 [0·82%] participants) and BRCA2 (110 [0·75%]), which are both associated with breast cancer. The rate of pathogenic or likely pathogenic variants varied widely between tumour types, with the highest proportion (for which >50 diagnoses were observed in the cohort) found in ovarian cancer (53 [9%] of 610 cases). 326 (45%) of 727 pathogenic or likely pathogenic variants had a known association with the tumour diagnosed in that participant, with proportions varying considerably between genes. INTERPRETATION: Understanding the frequency and nature of variants in cancer predisposition genes is important for planning of clinical services. Our analysis highlights the implications of more expansive genetic testing and the variant interpretation considerations necessary to yield benefit and avoid harm (eg, unnecessary surveillance) for patients. FUNDING: UK National Institute for Health and Care Research Cambridge Biomedical Research Centre.