Equity in cancer genomics in the UK: a cross-sectional analysis of a national cancer cohort

Background: Most research on genetic screening and precision oncology is based on individuals of European ancestry. We applied the National Health Service (NHS) England's cancer variant prioritisation workflow to evaluate the performance of these approaches in ethinically and ancestrally diverse populations. The second aim of the study was to assess the representativeness of the 100 000 Genomes Project cancer cohort of the population of England. Methods: In this cross-sectional analysis, whole-genome sequencing data from patients with cancer recruited into the 100 000 Genomes Project between February 2015 to December 2018 were analysed. Clinical information, including tumour stage and grade, was gathered from the NHS England National Cancer Registration and Analysis Service. Patients with cancer types with fewer than five individuals, haematological cancers, childhood cancers, unknown primary carcinomas, patients with indeterminate sex, and patients missing somatic mutations in genes were excluded. To assess ethnicity representation in the 100 000 Genomes Project, we calculated the recruitment ratios for self-reported ethnicities for patients with cancer recruited to the 100 000 Genomes Project and patients with cancer in England. We also analysed differences in classification rates for potentially pathogenic variants to assess ancestry-related differences in germline and somatic mutations of different ancestry groups. Findings: 14 775 patients with cancer were recruited between February, 2015, and December, 2018, into the 100 000 Genomes Project. There was no evidence of under-representation of diverse ethnic groups in the 100 000 Genomes Project when compared with the national statistics. The recruitment rate ratio for breast cancer was 2·2 (95% CI 1·6–3·0) for Black versus White women in the 100 000 Genomes Project compared with 0·81 (0·79–0·83) for Black versus White women in the national data (fold-change in rate ratios 2·7; 95% CI 2·0–3·7, p<0·0001), suggesting higher representation of Black women in the 100 000 Genomes Project than expected given the ethnicity-specific incidence rates in England. Compared with national rates, the 100 000 Genomes Project also had higher recruitment rates of Black versus White men with prostate cancer (fold-change in rate ratios 3·7; 1·8–7·5, p=0·0004), Black versus White men with bladder cancer (fold change in rate ratios 6·1; 2·0–18·8, p=0·0016), and Asian versus White women with breast cancer (fold change in rate ratios 1·4; 1·2–1·7, p=0·0008). Ancestry had a significant association with the likelihood of carrying a variant classified as a potentially pathogenic (likelihood ratio test p=0·0011). Potentially pathogenic variants were identified in 23 (4·6%) of 500 South Asian (adjusted model odds ratio [OR] 1·88, 95% CI 1·21–2·93, p=0·0052) and 24 (5·3%) of 453 African ancestry patients (OR 2·24, 1·44–3·48, p=0·0003) compared with 263 (2·2%) of 11 955 in European-ancestry patients. However, we found that fewer tumour mutations in actionable genes were identified for patients of non-European ancestry compared with patients of European ancestry when adjusting for sex and cancer type (likelihood ratio test p<0·0001). Interpretation: The was an excess of germline variants classified as potentially pathogenic variants in patients with non-European ancestry, which might impede the diagnostic process. Improved variant prioritisation workflows and more research in diverse groups are needed to ensure equitable implementation of genomics in cancer care.