Cumulative proportion of responders analysis (CPRA) as a tool to assess treatment outcome in alcohol clinical trials

Objective: Several definitions of treatment response have been proposed for alcohol clinical trials (e.g., abstinence and no heavy drinking). However, each of these outcomes allows only one definition of successful response. In contrast, the cumulative proportion of responders analysis (CPRA) includes all of the possible drinking response cutoff points, providing a more complete picture of the therapeutic effects of a treatment. CPRA has been used to examine the efficacy of analgesics but not alcohol pharmacotherapy. To demonstrate its potential utility, we conducted CPRA in two large alcohol treatment trials: the COMBINE (Combined Pharmacotherapies and Behavioral Interventions for Alcohol Dependence) trial (naltrexone) and a multisite topiramate trial. CPRA was used to demonstrate the efficacy of naltrexone and topiramate on continuous measures of in-treatment drinking-heavy drinking days and drinks per day-and their reductions from pretreatment. Method: All possible cutoff points were portrayed for each measure. We provide graphs to illustrate the effects of the active medications compared with placebo and examined them statistically over a number of salient drinking outcomes to evaluate their efficacy. Results: Treatment group responder curves were not parallel across the entire range of cutoff points; rather, they separated only at lower levels of drinking. In general, effect sizes increased by 0.10-0.15 when going from the lowest drinking level cutoff (i.e., abstinence and no heavy drinking) to the cutoff associated with the maximal treatment effect. Conclusions: CPRA may be useful in designing subsequent trials and helping to illustrate for treatment providers the likelihood of treatment success given various definitions of a positive response.