Results in the ordinal model compared with the ordinal model adjusted for dependence on other substances (which was evaluated only on the SNPs with p < 1.0 × 10−4 in the unadjusted model) were somewhat different. The only GWS region in EAs was observed under the adjusted model, with findings about an order of magnitude less significant under the nonadjusted model. In AAs, DLC1 was significant only under the ordinal model and the more distal intergenic region, only under the adjusted model. The extended chromosome 14 region was GWS in AAs under the ordinal model, with only one SNP in the region meeting significance criteria under the adjusted model as well. Regions significant under the adjusted model should be considered more specific to ND than those significant under the unadjusted ordinal model, and the use of an adjusted model could account for some of our novel association findings. Although ND frequently co-occurs with dependence on other substances, especially the three SD disorders on which our sample was ascertained, this is rarely adjusted analytically and sometimes not even measured.
