Linear dose effects for supportive and secondary outcomes were determined using the MIXED TEST subcommand for Linear Trend Contrasts and Multiple Event Models (MEMs) using PASW 17.0 software (IBM Corp., Armonk, NY).40,41 All MEMs were repeated measures centered at Week 12 and included the baseline value of the dependent variable as a covariate. Week, treatment and week x treatment were evaluated as fixed effects in each model. Missing values were assumed missing at random and treatment effects were estimated by the restricted maximum likelihood method. Supportive primary outcomes of drinking quantity and frequency were reported as estimates of drug effects (regression coefficients) with associated 95% confidence intervals (95% CI). Outcomes involving craving, mood and sleep, and log-transformed GGT values were reported as F values from Type III tests of fixed effects.
