To test whether the association between OPRM1 and the development of an AUD in adolescence was moderated by two environmental influences we used logistic regression with a hierarchical backward elimination procedure to determine the most parsimonious model. Interaction effects of the OPRM1 genotype with environmental variables were calculated by multiplying the OPRM1variable by each centered environmental variable. Potential confounding effects of sex and age were controlled by entering these variables as covariates. The main and interaction effects of OPRM1 genotype, deviant peer affiliation, and parental monitoring were entered into the logistic regression model simultaneously. The final backward elimination logistic regression model was significant (χ2 = 64.41, df = 3, p = .000, NagelkerkeR2 = .77, see Table 3) and included sex, the OPRM1 × parental monitoring interaction, and the OPRM1 × deviant peer affiliation interaction. These predictors remained significant after controlling for multiple testing using the Bonferroni correction (adjusted α = 0.025). As hypothesized, decomposition of theOPRM1 × parental monitoring interaction indicated that carriers of the G allele who reported relatively high levels parental monitoring were least likely to develop
