Using methylation of OXTR as the outcome, we again used regression analyses to examine main and interaction effects, beginning by checking for multicollinearity. The VIF ranged from 1.01 to 1.15, and all measures of tolerance were above .85, indicating no multicollinearity. As can be seen in Model 2a of Table 2, neither the main effect of early initiation of substance use (age 13), nor genotype was significant. In Model 2b we added the multiplicative interaction term for substance use by genotype. As hypothesized, the interaction was significant even after controlling for ancestry and cell-type variation. Exclusion of outliers greater than three standard deviations from the mean on OXTR methylation (n = 8) left the pattern of significant results unchanged (See supplemental Table S3) and so all participants were retained.
