Genetic associations for continuous abstinence rates and nausea incidence were assessed using logistic regression, assuming an additive genetic model. A survival model was used to assess the relapse data; specifically, a proportional hazards model was fitted to determine whether the time to relapse was affected by genotype. The proportional hazards assumption was checked using plots of the log-log survival (relapse) curves. The relapse analysis also assumed an additive genetic model. Our analysis was limited to SNPs whose minor allele frequency was >5% in our overall genotyped population, reducing the total number of candidate gene SNPs analyzed to 785. For each analysis, a treatment × genotype interaction term was initially included in the model. For markers with at least marginally significant interaction with treatment (p<0.20), the analysis was performed for each treatment group separately, as well as for all subjects together and adjusted for treatment. When the interaction was not significant, the data were analyzed as a single pooled sample, adjusted for treatment. To correct for multiple testing in which many markers were in strong linkage disequilibrium (LD), we selected individual
