Abstract
BACKGROUND: Partially due to the complexity associated with conducting trials, there have been relatively few regulatory agency-approved medications for the treatment of alcohol use disorder (AUD). Heterogeneity of study samples, such as participant characteristics (including variation in alcohol consumption) as well as drinking efficacy endpoints, may lead to inconsistency in results and a potential increase in Type II errors. The aim of this study is to begin to fill in the knowledge gap of optimal clinical trial design by analyzing potential predictors of outcomes. METHODS: Five federally funded and publicly available multisite randomized pharmacotherapy clinical trials for the treatment of AUD using similar methodologies to assess drinking outcomes were included. Three FDA-guided drinking efficacy outcomes were calculated for each study independently and combined: abstinence (no drinking days), no heavy drinking days, WHO 2+ risk drinking level (RDL) reduction as well as a measure of study participant completion. These outcomes were analyzed by logistic regression models including a variety of predictors within the full-study sample as well as among only participants treated with placebo. RESULTS: The number of days abstinent prior to randomization was a strong positive predictor of all three predefined drinking outcomes. Further analysis indicated a cutoff of less than three to five abstinent days before randomization might be optimal. For the WHO 2+ RDL endpoint, those within the "high" or "very-high" risk categories were more likely to meet criteria for successful two or more risk level reductions than "medium" risk. Across various demographic variables, only age (older participants) was associated with better outcomes. CONCLUSIONS: These findings suggest that some important prestudy drinking characteristics (e.g., less abstinence and older age) as well as being in a higher risk drinking category should be considered for inclusion in future alcohol pharmacotherapy trials.