Percentage of subjects with no heavy drinking days: evaluation as an efficacy endpoint for alcohol clinical trials

Background: Percent subjects with no heavy drinking days (PSNHDDs), an efficacy end point recommended by the Food and Drug Administration, considers abstinent individuals or those engaging in low-risk drinking behavior as successful responders to treatment. As PSNHDD has been used infrequently in previous alcohol clinical trials, we evaluated the utility and validity of the PSNHDD outcome measure in 2 large alcohol clinical trials.

Methods: Data sets from 2 alcohol trials, COMBINE and a multisite topiramate trial, were used to analyze PSNHDDs and other traditional end points for the topiramate, naltrexone, acamprosate, and placebo groups. Effect sizes of PSNHDDs were determined for each month of active treatment and by varying grace periods-early periods in the trial where outcome is not considered in the analysis-and were compared with that of other traditional outcome measures. Long-term outcomes were compared for groups that had no heavy drinking days versus those that had heavy drinking days during active treatment.

Results: PSNHDD effect sizes were significant for both topiramate (0.34 and 0.25 at months 2 and 3, respectively) and naltrexone (0.24 and 0.26 at months 3 and 4, respectively). Given a 2-month grace period for naltrexone, the effect size of PSNHDDs was comparable to the effect sizes using traditional outcome measures. With a 1-month grace period for topiramate, it was greater than the majority of traditional outcome measures. Little is gained by allowing up to 1, 2, or 3 heavy drinking days as an end point. Subjects with no HDDs during treatment fared better than those with some HDDs on drinking outcomes and alcohol-related consequences during a 1-year follow-up.

Conclusions: PSNHDD appears to be a clinically informative end point measure, especially when used with a grace period, and is as sensitive as most traditional outcome measures in detecting differences between the medication and placebo groups. Nonetheless, these findings should be replicated in other clinical data sets, particularly with medications that work via different mechanisms.