There were limitations in the study. In the GWAS study, the CRIME sample was genotyped with a different array than the control samples (GenMets, Corogene), although in the same facility of Welcome Trust Sanger Institute. Therefore, only the SNPs that overlapped in all of the three genotyping data sets were included in the analyses and even unusually stringent thresholds were used in the quality control. Furthermore, the regenotyping of the eight replication variants in the CRIME subjects and GenMets controls that were included in the GWAS, revealed practically identical genotypes as the GWAS genotyping with >99% concordance between the two methods. In a post hoc repeat GWAS analysis, including the CRIME subjects and GenMets controls, the results were consistent with the original analysis. This also indicated that had there been discrepancy with the Corogene control sample genotypes of the eight SNPs in the GWAS, this would have shown in the repeat analysis results. In spite of intensive QC, spurious associations may occur. We considered the top variant (rs6462756, near SDK1 gene) of the analysis of males and females combined as
