Effects of GLP-1 receptor agonists on craving and compulsive consumption in substance use and binge eating disorders: a systematic review

Introduction

Substance use disorders (SUD) and binge eating disorder (BED) share core neurobiological features, including mesocorticolimbic dopaminergic dysregulation and impaired inhibitory control. Current pharmacological options for both conditions remain limited in efficacy and scope. Glucagon-like peptide-1 receptor agonists (GLP-1 RAs), approved for type 2 diabetes and obesity, modulate reward circuitry through receptors expressed in the nucleus accumbens, ventral tegmental area, amygdala, and prefrontal cortex, providing a neurobiological rationale for their use across both conditions.

Aims

To synthesize evidence from randomized controlled trials (RCTs) on the efficacy of GLP-1 RAs across SUD and BED, organized by neurocircuit domain, and to identify candidate moderators of treatment response.

Methods

A systematic review following PRISMA guidelines was conducted. Six databases (PubMed, MEDLINE, EMBASE, CENTRAL, Scopus, ClinicalTrials.gov) were searched up to March 2026. Eligible studies were RCTs reporting behavioral outcomes in participants with SUD or BED. Outcomes were categorized into four domains derived from the Koob–Volkow addiction cycle: compulsive consumption/use, craving, withdrawal/negative reinforcement, and relapse/abstinence. Methodological quality was assessed using the Cochrane RoB2 tool.

Results

Eighteen studies from 13 parent RCTs were included (n=3 to 255 per trial). GLP-1 RAs examined were semaglutide, liraglutide, exenatide, and dulaglutide. Compulsive consumption was the most consistently affected domain. In alcohol use disorder (AUD), once-weekly semaglutide significantly reduced heavy drinking days (β=0.84; 95% CI 0.71–0.99; p=0.04) and laboratory alcohol self-administration (β=−0.48; 95% CI −0.85 to −0.11; p=0.01), while exenatide showed efficacy only in participants with comorbid obesity. In BED, significant reductions in disorder severity scores were observed across three trials, though binge episode frequency and food craving were unaffected. The most clinically significant finding was a simultaneous reduction in alcohol consumption and cigarette use within the same semaglutide RCT, indicating a transdiagnostic, non-substance-specific effect. Craving was significantly reduced only for alcohol under semaglutide; tobacco, cocaine, and food craving outcomes were null or below threshold. Abstinence findings in tobacco use disorder were divergent: exenatide added to nicotine replacement therapy significantly improved 7-day point-prevalence abstinence (46.3% vs 26.8%; RR=1.70; 95% CrI 0.96–3.27; PP=96.5%), whereas dulaglutide added to varenicline showed no effect (63% vs 65%; p=0.859). Withdrawal outcomes were assessed in one trial only. Of the 13 parent trials assessed using the Cochrane RoB2 tool, one was rated low risk of bias, six raised some concerns, and six were rated high risk, reflecting the early-stage and methodologically heterogeneous nature of the current evidence base.

Conclusions

GLP-1 RAs show a promising but inconsistent efficacy signal across SUD and BED. The evidence is strongest for semaglutide in AUD and points towards a transdiagnostic mechanism of action. Two candidate moderators - GLP-1 RA agent and metabolic phenotype of participants - may explain divergent results and require prospective and adequately powered evaluation. Current evidence is limited by small samples, methodological heterogeneity, and near-absence of withdrawal-domain outcomes. Larger, adequately powered trials with standardized outcome measurement and agent-specific designs are needed before clinical recommendations can be made.

References

No references