Background <p>The relative efficacies of nalmefene versus naltrexone for alcohol use disorder is the subject of intense and ongoing debate. The two pan-opioid receptor ligands differ primarily in actions at the kappa opioid receptor, where naltrexone acts as an antagonist and nalmefene acts as a partial agonist. Parallel clinical trials for nalmefene or naltrexone have produced widely disparate outcomes and a marked lack of consensus regarding which of the compounds should be used for the treatment of alcohol use disorder.</p> Methods <p>Here we leveraged a mouse model (n = 56 male C57BL/6 J) to directly compare the efficacy of nalmefene and naltrexone within-subject. After acquiring operant responding for ethanol, each subject underwent four treatment block conditions: nalmefene (0.1 mg/kg i.p.), naltrexone (1.0 mg/kg i.p.), the selective kappa opioid receptor agonist U50,488 (1.0 mg/kg i.p.) and placebo (saline 10 ml/kg i.p.). Each treatment block consisted of an ethanol self-administration session followed by two subsequent sessions of punished (quinine adulterated) ethanol self-administration sessions with treatment given 30 min prior to each session.</p> Results <p>We show that nalmefene and naltrexone have similar efficacy in reducing ethanol consumption, whereas U50,488 increases ethanol consumption. Despite similar effects in aggregate analyses, nalmefene- and naltrexone-induced reductions in drinking are driven by fully separate subpopulations which do not show any beneficial response to the non-preferred compound and display markedly different behavioral phenotypes prior to treatment. A predictive model based on circulating biogenic amines allows for high accuracy classification of nalmefene- versus naltrexone-responders.</p> Conclusion <p>Together, these results provide a roadmap for improving alcohol use disorder treatment outcomes via precision application of existing compounds.</p>