Population stratification or other underlying differences (e.g., genotyping batch effects) between the different cases and controls cohorts may give rise to false positive associations. To minimize this possibility, we followed a carefully constructed pipeline (Figure 1) that included stringent quality control, pairwise comparisons of control cohorts to ensure that λgc values were not inflated, use of only SNPs genotyped across all cohorts as inputs for imputation,(Johnson et al., 2013) adjustment for eigenvectors to circumvent population stratification, and application of genomic control to reduce residual inflation. The λgc values from our GWAS analyses (e.g., λgc=1.02 for the GWAS meta-analysis) suggest that our pipeline minimized the potential for false positive associations.
