In order to investigate the possibility of secondary signals at loci that met the criteria for genome-wide significance (defined as P < 5 × 10−8), conditional regression analyses were performed conditioning on the most strongly associated SNP in each region. We then applied the Nyholt method for multiple testing correction to derive a threshold for determining statistical significance based on the number of SNPs tested and taking into account LD across the region (10). These regions were defined based on locations of nearby recombination hot spots. In absence of these, we defined a region as ±250 kb from the top SNP.
