A series of mixed models (PROC MIXED) in SAS (V9.2 Cary, NC) were conducted for each tSNP on the three dependent variables of interest: alcohol-induced sedation, stimulation, and craving. Specifically, Medication was modeled as a two-level within subjects factor (i.e., Placebo vs. Naltrexone, coded 0 and 1 respectively), Genotype as a two-level between subjects factor (e.g., OPRD1^4: GG = 0 vs. AG/AA = 1 and OPRK^2: TT = 0 vs. CT/CC = 1), and BrAC as a four-level within subjects factor (i.e., BrAC = 0.00, 0.02, 0.04, and 0.06 g/dl, coded 0 to 3). The model also included interactions between genotype and medication, as well as between BrAC and medication. If a non-significant interaction was found (e.g., BrAC × medication), the model was re-run without the interaction term. In addition, models producing significant genotype × medication interactions were followed up by a simple effects analysis. An asterisk (*) in the figures indicates a significant simple effect (p < 0.05). Finally, OPRM1 genotype and the interaction between OPRM1 genotype and medication were used as covariates in subsequent models. This added step
