A meta-regression of methodological features that predict the effects of medications on the subjective response to alcohol.

Abstract

Alcohol administration paradigms have been used for early efficacy testing of novel compounds for alcohol use disorder (AUD). There has been an ongoing debate about sample characteristics and methodological features that affect the likelihood of detecting an early efficacy signal for AUD medications. We conducted a meta-regression to test whether the drinking level of the study sample and the peak breath alcohol concentration (BrAC) in the alcohol administration study predict the efficacy of AUD pharmacotherapies on the subjective responses to alcohol. We computed the effects of 21 medications on alcohol-induced stimulation, sedation, negative mood, and craving during alcohol administration in49studies. Meta-regression analyses indicated a significant and positive effect of pre-study drinks per month on alcohol-induced stimulation (β=0.142, p<0.0001), such that as drinking increases, the benefit of medication over placebo decreases. There was an effect of drinks per month on negative mood (β=-0.164, p=0.0248), such that at higher levels of drinks per month, the effects of medications on negative mood are stronger. For sedation, there was an effect of peak BrAC (β=0.119, p=0.0002), such that at low levels of peak BrAC, the effects of medication on sedation were null. For craving, there was a peak BrAC×drinks per month interaction such that at low levels of BrAC, a heavier drinking sample is required to detect the effects of medication on craving. Sensitivity analyses comparing naltrexone studies and non-naltrexone studies suggested that naltrexone was less sensitive to drinks per month across subjective response domains. These analyses show that design features are critical in studies that test the effects of medications on the subjective responses to alcohol. By specifying the significance and directionality of these effects, as well as the specific points in BrAC or drinks per month at which medication effects are detectable, the study offers recommendations for design features of alcohol administration studies that aim to inform AUD medication development.