Several post-hoc analyses were undertaken to provide a better understanding of our primary findings. An additional logistic regression comparing cases and non-dependent controls was performed to control for the allelic dosage of the most significant OPRD1 SNP using the condition option in PLINK to examine whether evidence of more than a single association signal was found. Based on the results of this analysis, the haplotype command in PLINK was used to assess risk associated with a single 2 SNP OPRD1 haplotype. Finally, consistent with the exclusion criteria for the non-dependent controls, we divided neighborhood controls into those who did not meet criteria for dependence on alcohol or a non-opioid illicit (i.e., cannabis, stimulants, sedatives, or cocaine) drug (N=275) and those who did (N=256). We then separately compared cases to each of these groups to estimate the risk for heroin dependence risk associated with the most significant OPRD1 SNP.
