Polygenic risk scores and breast and epithelial ovarian cancer risks for carriers of BRCA1 and BRCA2 pathogenic variants

Abstract

Purpose We assessed the associations between population-based polygenic risk scores (PRS) for breast (BC) or epithelial ovarian cancer (EOC) with cancer risks for BRCA1 and BRCA2 pathogenic variant carriers. Methods Retrospective cohort data on 18,935 BRCA1 and 12,339 BRCA2 female pathogenic variant carriers of European ancestry were available. Three versions of a 313 single-nucleotide polymorphism (SNP) BC PRS were evaluated based on whether they predict overall, estrogen receptor (ER)–negative, or ER-positive BC, and two PRS for overall or high-grade serous EOC. Associations were validated in a prospective cohort. Results The ER-negative PRS showed the strongest association with BC risk for BRCA1 carriers (hazard ratio [HR] per standard deviation = 1.29 [95% CI 1.25–1.33], P = 3×10−72). For BRCA2, the strongest association was with overall BC PRS (HR = 1.31 [95% CI 1.27–1.36], P = 7×10−50). HR estimates decreased significantly with age and there was evidence for differences in associations by predicted variant effects on protein expression. The HR estimates were smaller than general population estimates. The high-grade serous PRS yielded the strongest associations with EOC risk for BRCA1 (HR = 1.32 [95% CI 1.25–1.40], P = 3×10−22) and BRCA2 (HR = 1.44 [95% CI 1.30–1.60], P = 4×10−12) carriers. The associations in the prospective cohort were similar. Conclusion Population-based PRS are strongly associated with BC and EOC risks for BRCA1/2 carriers and predict substantial absolute risk differences for women at PRS distribution extremes.

Polygenic Risk Modelling for Prediction of Epithelial Ovarian Cancer Risk

“…al derived a PRS score using 22 SNPs that were significantly associated with high-grade serous EOC risk in GWAS (PRS HGS ) to predict EOC risk in BRCA1/BRCA2 pathogenic variant carriers (5). To make effect estimates obtained in this analysis comparable to the effect estimates obtained from the PRS HGS , we standardized all PRSs using the standard deviation from unaffected BRCA1/BRCA2 carriers; all PRS models in this analysis except the Stepwise (OCAC only) had higher effect estimates (5). However, the corresponding AUCs were higher for the PRS HGS model (0.604 for BRCA1 carriers and 0.667 for BRCA2 carriers), most likely as a result of inclusion of other predictors (birth cohort and principal components) in the model.…”

“…In the UK Biobank data, we evaluated discriminatory performance of the models using the AUC and examined the association between standardized PRS and risk of non-mucinous EOC using logistic regression analysis. For the CIMBA data, we assessed associations for each version of the PRS and invasive non-mucinous EOC risk using weighted Cox regression methods previously described (5). PRSs in the CIMBA data were scaled to the same PRS standard deviations as the OCAC data, meaning that per standard deviation hazard ratios estimated on CIMBA data are comparable to PRS associations in the OCAC and UK Biobank data.…”

“…(1,3), have identified 39 common (minor allele frequency [MAF] > 0.05) susceptibility variants which together explain about 6% of the heritability of EOC. Polygenic risk scores (PRS) provide an opportunity for refined risk stratification in the general population as well as in carriers of rare moderate or high risk alleles (4,5).…”

A Comprehensive Epithelial Tubo-Ovarian Cancer Risk Prediction Model Incorporating Genetic and Epidemiological Risk Factors

“…The EOC risk prediction model presented here combines the effects of FH of cancer, the 36,37 . The current evidence also suggests that known epidemiological RFs have similar effect sizes for EOC in BRCA1 and BRCA2 PV harbourers as in non-harbourers 38,39 .…”

“…Similarly, the model assumes that the relative effect-sizes of RFs and the PRS are similar in females harbouring PVs in BRCA1, BRCA2, RAD51C, RAD51D and BRIP1 and those without PVs in these genes. Evidence from studies of BRCA1 and BRCA2 PV harbourers suggests that this assumption is plausible: PRSs for EOC have been shown to be associated with similar relative risks of EOC in the general population and in BRCA1 and BRCA2 PV harbourers 36,37 . The current evidence also suggests that known epidemiological RFs have similar effect sizes for EOC in BRCA1 and BRCA2 PV harbourers as in non-harbourers 38,39 .…”