Identifying responders to gabapentin for the treatment of alcohol use disorder: an exploratory machine learning approach.

This article focuses on applying the Qualitative Interaction Tree (QUINT) machine learning method to identify subgroups of individuals with alcohol use disorder (AUD) who respond differently to gabapentin enacarbil extended-release (GE-XR), a prodrug formulation tested in a multisite clinical trial. The analysis of baseline clinical and demographic data from 338 participants revealed that treatment responders to GE-XR were characterized by factors including higher motivation for change, lower confidence in achieving drinking goals (self-efficacy), baseline drinking levels, cognitive impulsivity, and anxiety. Conversely, some subgroups showed iatrogenic responses, meaning they responded better to placebo than to GE-XR, often linked to differing motivation and drinking patterns. These findings suggest that motivation and self-efficacy, alongside withdrawal-related factors, may be important predictors of clinical response to gabapentin in AUD, highlighting the potential for personalized treatment approaches and the value of machine learning in medication development for AUD.