With regard to our initial GWAS conducted as a consecutive three-stage analysis, the lowest combined P-value for the entire sample was P=8.044 × 10−7 (Supplementary Table S2), which would have been deemed genome-wide nonsignificant if only a single-stage analysis was used to calculate conventional Bonferroni- or false discovery rate-corrected P-values for the total samples to determine statistical significance. However, ‘significant' results obtained as conventionally corrected P-values will not always represent true associations, meaning that the results may not be necessarily replicated in other studies, and vice versa. For example, data from the National Human Genome Research Institute GWAS catalog (as of 31 January 2009), show 1321 entries of discovered associations with a P-value of <10–5, but only 550 of these entries have a P-value of <5 × 10–8,44 which is a conventionally corrected conservative threshold for declaring a significant association in a GWAS.45, 46 In both cases, truly potent candidate SNPs may be included in the outcome of the studies. Furthermore, GWASs in pharmacogenomics, such as this study, would tend not to yield ‘significant' results obtained as conventionally corrected P-values
