An empirical basis for the treatment of alcohol problems

This study summarizes over 50 years of randomized controlled trials testing treatment for alcohol problems in a systematic review of the outcome literature. This review provides a summary of the most effective treatments for alcohol problems and examines the research methodology used in testing these treatments. Methods for assigning empirical support were compared for their agreement as to the empirical support different treatment interventions. A review of the literature identified 451 outcome trials were analyzed by two independent coders. Treatments tested in these trials were compared using the Mesa Grande coding system, the criteria for empirical support defined by the Division 12 task of the American Psychological Association and traditional meta-analysis. The Mesa Grande system had moderate agreement with effect size in the identification of empirically supported treatments. The APA division 12 criteria supported the use of treatments that were not successful in the majority of trials where they were tested. Several empirically supported treatments were distinguished from other less effective options. Brief interventions demonstrated efficacy similar to longer interventions. Abstinence, harm reduction and controlled drinking goals were all empirically supported. A curvilinear relationship between the severity of alcohol dependence and positive outcome was observed. Many treatments performing well overall earned empirical support in populations with moderate to severe alcohol dependence, including acamprosate, naltrexone, social skills training, community reinforcement, marital behavioral therapy, and motivational interviewing. Most, treatments were tested in studies using one follow-up period. Among commonly utilized outcome measures, percent days abstinent was more sensitive than measures of negative consequences. Other measures, less sensitive in detecting group differences, included craving and ASI scores. Studies using additive designs and inactive control groups were more likely to detect group differences than other designs. These findings should be considered in the implementation of treatment and outcome research. (PsycINFO Database Record (c) 2010 APA, all rights reserved)