EP061: Genetic risk for breast and ovarian cancer in a diverse and unselected population

Abstract

A clear picture of the genetic architecture of breast cancer risk is taking shape since the advent of multi-gene panel testing, with damaging variants in a set of DNA repair genes now firmly established as having high (>5-fold risk) or moderate (2-5-fold risk) penetrance for this cancer. A subset of these variants also increase the risk for ovarian cancer. Identification of individuals at increased genetic risk for these cancers creates an opportunity to improve health outcomes through interventions aimed at prevention or early diagnosis and treatment. Here, we evaluated the prevalence and impact on time to cancer diagnosis of expected pathogenic variants in 14 high and moderate penetrance breast and ovarian cancer risk genes in an ancestrally diverse patient population. The BioMe Biobank is an electronic health record (EHR)-linked biobank with over 60,000 participants enrolled non-selectively from ambulatory care practices across the Mount Sinai Health System in New York, NY. From exome sequence data available for 30,223 adult BioMe participants, we identified expected pathogenic variants in ATM, BARD1, BRCA1, BRCA2, BRIP1, CDH1, CHEK2, NF1, PALB2, PTEN, RAD51C, RAD51D, STK11, and TP53. These included ClinVar pathogenic and likely pathogenic (P/LP) variants, conflicting ClinVar variants with majority P/LP assertions, and additional predicted loss-of-function variants (frameshift, stop-gain, start-loss, or variants at canonical splice acceptor or donor sites) in these genes. We estimated the prevalence of high penetrance breast (BRCA1, BRCA2, CDH1, PTEN, STK11, TP53), moderate penetrance breast (ATM, BARD1, BRIP1, CHEK2, NF1, PALB2, RAD51C, RAD51D), and ovarian (BRCA1, BRCA2, BRIP1, PALB2, RAD51C, and RAD51D) cancer gene variants in BioMe overall, and across the 8 largest population groups defined by shared recent genetic ancestry (African American and African, Ashkenazi Jewish, Non-Ashkenazi Jewish European, Filipino and Other Southeast Asian, Dominican, Ecuadorian, Puerto Rican, and Colombian and Other Central and South American). We evaluated the clinical consequences of expected pathogenic variants in these genes in a subset of unrelated adult female participants (N=15,919) using International Classification of Disease-9/10 codes extracted from EHRs and self-reported data from BioMe enrollment questionnaires. Kaplan-Meier curves were generated to estimate time to diagnosis of breast or ovarian cancer, with the age at first record (EHR or self-reported) of breast or ovarian cancer diagnosis considered the event, and censoring at current age or age 90. Hazard ratios (HRs) were generated using Cox proportional hazards models, adjusting for current age, self-reported race and ethnicity, and the first 5 principal components of ancestry. We identified 667 variant-positive individuals with 292 unique expected pathogenic variants in high or moderate penetrance breast cancer genes. The estimated prevalence of expected pathogenic variants in these genes in a subset of sequenced individuals that excluded second degree relatives or closer (N = 27,816) was 1 in 44 (Table). Across population groups defined by shared recent genetic ancestry, prevalence ranged from 1 in 21 in individuals with Ashkenazi Jewish (AJ) ancestry to 1 in 84 in those with ancestry from Puerto Rico or Ecuador (Table). At an individual gene level, CHEK2 variants had the highest overall prevalence (1 in 140). This varied widely by genetic ancestry, with the highest prevalence observed in AJ individuals (1 in 44), largely due to overrepresentation of two low penetrance missense variants that are common in individuals of European and AJ descent (ie, c.1283C>T and c.470T>C, NM_007194.4). We conducted survival analyses to determine if age at diagnosis differed by high or moderate penetrance breast cancer gene variant status. These analyses revealed that, by age 60, 30.7% of high penetrance variant-positive, 13.0% of moderate penetrance variant-positive, and 5.4% of variant-negative women had breast cancer. Both high and moderate penetrance variants were associated with earlier time to breast cancer diagnosis (HR = 5.8, p = 1.2 x 10-31; HR = 1.6, p = 5.7 x 10-3, respectively). The common CHEK2 variants c.1283C>T and c.470T>C were not associated with time to breast cancer diagnosis (HR 0.81, p = 0.6), whereas protein-truncating and rare missense variants in this gene were associated with earlier time to breast cancer diagnosis (HR 3.2; p = 5.6 x 10-4). Similarly, expected pathogenic variants in the subset of ovarian cancer genes were associated with earlier time to ovarian cancer diagnosis (HR 8.3, p = 2.5 x 10-10), with 2.3% of variant-positive women estimated to develop ovarian cancer by age 60 compared to 0.4% of variant-negative women. Finally, among 112 women with a documented diagnosis of ovarian cancer, we observed low rates of prior clinical genetic testing in their EHRs (18.8%). We looked at rates of genetic testing across population groups defined by self-reported race and ethnicity. The highest rates of genetic testing were in European American women at 28.6%, and lower rates were seen in Hispanic and Latina and African American-African women (17.8% and 4.2%, respectively), suggesting disparities in clinical genetic testing across population groups in women who meet clinical criteria for testing. These findings highlight the relevance of breast cancer genomic risk stratification to a diverse urban health care system. In aggregate, expected pathogenic variants in high and moderate penetrance breast and ovarian cancer risk genes are highly prevalent. These variants significantly reduced time to breast cancer diagnosis, with rates of diagnosis by age 60 approximately 6- and 2-fold greater in high and moderate penetrance variant-positive women respectively than in variant-negative women. However, we found no difference in time to diagnosis with common CHEK2 missense variants, suggesting that these are of questionable clinical relevance despite routinely being reported as P/LP in clinical care. It is imperative that clinicians and patients alike be aware of the nuanced range of genomic risk to implement personalized cancer risk assessment effectively and equitably across populations.TableEstimated prevalence of expected pathogenic variants in high and moderate penetrance breast and ovarian cancer genes in unrelated adults in BioMe.Population, NPrevalenceAll genesHigh penetrance breast cancer genesModerate penetrance breast cancer genesOvarian cancer genesTotal, 27,8161:441:125a1:67b1:89Genetic Ancestry GroupsAfrican American and African, 6,8741:661:2151:961:107Dominican, 1,8761:721:3751:891:156Ecuadorian, 4181:841:2091:1391:139Filipino and Other Southeast Asian, 5661:381:631:941:63Ashkenazi Jewish, 3,8891:211:471:371:45Non-Ashkenazi Jewish European, 5,4741:301:961:431:71Puerto Rican, 51041:841:3001:1161:150Colombian and Other Central and South American, 1,1161:701:1241:1591:112aIncludes 5 individuals who are also heterozygous for a moderate penetrance variant.bIncludes 5 individuals who are also heterozygous for a high penetrance variant, and 3 individuals who are heterozygous for two moderate penetrance variants. Open table in a new tab