We performed post-hoc ROC analyses to evaluate the contributions of demographic, dependence, and genetic variables to predict abstinence at 6MO. We evaluated ROC models for the association of rs1051730 with abstinence in the PLA PG at 6MO (N~467), in the NRT PG at 6MO (~740), and in all PGs at 6MO (sample size ~2592) (Figure 3 and see Table, Supplemental Digital Content 10: Area Under the Curve (AUC) mean and 95%CI estimates from PLA, NRT and All PG models). The ROC AUC values increase when pharmacotherapy is added, e.g., with the addition of NRT or all PGs, compared to PLA, and, within each set of ROC models, the ROC AUC increases when including additional variables in the model. For the PLA models, the AUC of the full model is significantly greater that PLA models with demographic variables or demographic variables and rs1051730. For NRT or all PG models, the inclusion of dependence variables, dependence variables and rs1051730, or dependence variables, rs1051730 and covariate SNPs (rs588765 and rs578776), results in ROC curves with significantly greater AUC than the models with only
