analysis, namely, pooled regression analysis. Heuristically, all of the studies contributed to the estimation of the regression coefficients for each demographic and dependence variable, which were assumed to have the same value across arms; each individual arm contributed to estimation of an arm-specific level variable, allowing for different abstinence rates across arms, and each individual arm contributed to estimation of a pharmacotherapy-specific coefficient for the SNP variable, and, if present, pharmacotherapy-specific coefficients for covariate SNPs. This approach was implemented because many of the arms had insufficient observations for reliable estimation of SNP effects if all of the covariates had been included and regressions were performed separately by arm.
