We conducted the fixed effects meta-analysis with the inverse variance weighting and random effects meta-analysis from the DerSimonian-Laird method31. All meta-analysis and calculations were performed using SAS version 9.4 (SAS Institute Inc., Cary, NC, USA). As the same referent panel was used for all studies, all SNPs showed the same forward alignment profiles. We excluded poorly imputed SNPs defined by imputation quality R2 < 0.3 or Info < 0.4 for each meta-analysis component and SNPs with a Minor allele frequency (MAF) >0.01 (except for CHEK2 rs17879961 and BRCA2 rs11571833 which we have validated extensively previously4. We generated the index of heterogeneity(I2) and P-value of Cochran’s Q statistic to assess heterogeneity in meta-analyses and considered only variants with little evidence for heterogeneity in effect between the studies (P-value of Cochran’s Q statistic >0.05). SNPs were retained for study provided the average imputation R-square was at least 0.4. For SNPs in the 0.4–0.8 range that reached genome wide significance results were evaluated for consistency with neighboring SNPs to assure a reliable inference. Due to the smaller sample size and fewer sites contributing
