Conditional analyses using GCTA-COJO37,38 to evaluate the independence of rs9478500 and rs1799971 suggested that each association was independent of the other. However, GCTA-COJO only accounts for LD in terms of r2 which is low between these variants in EA (0.035), whereas D’ is 1.0. Thus, following prior candidate gene studies39,40, we examined the associations of specific OPRM1 haplotypes with OA. The OA association was strongest with the haplotype consisting of the major rs1799971-A allele and minor alleles at all GWS OPRM1 variants from our study, which was associated with increased risk compared to the other haplotypes. The association with the haplotype consisting of the minor rs1799971-A allele and major alleles at all other GWS OPRM1 variants was not significant. In the subset of cohorts used in the haplotype analysis, the variant level association for rs1799971 was also not significant, which limits the strength of our haplotype-based conclusions. However, our earlier haplotype analyses of OPRM1 intron 1 variants and rs1799971 came to similar conclusions, albeit in more limited datasets39,40. This relationship between the underlying EA haplotype structure and risk for OA
