The current study had numerous limitations. First, we were limited to the available data across the three alcohol clinical trials, and due to this limitation, we did not include many other factors that have previously been identified as predictors of controlled drinking, such as the individual’s drinking goal[46, 49] and family history of alcohol dependence[11]. Preliminary model testing with the available family history data (COMBINE and MATCH) and available drinking goal data (COMBINE and UKATT) did not suggest these were robust predictors of low risk drinking. Preliminary models also tested education, income, and employment status, but none of these emerged as significant predictors. Finally, COMBINE, MATCH, and UKATT all excluded individuals with comorbid psychiatric disorders and COMBINE excluded individuals who could not maintain abstinence for 4 days prior to starting treatment, which may limit the generalizability of the present findings with respect to psychiatric symptoms predicting drinking during treatment and the severity of AUD in the sample (particularly in the COMBINE participants). Additionally, the treatment period for COMBINE was 16 weeks, however only the first 12 weeks of data were
