Top 10 PubMed Studies on GLP-1 Receptor Agonists and Substance Use (2026)

Hendershot 2025 is the first published randomized trial of semaglutide for alcohol use disorder. Forty-eight non-treatment-seeking adults with AUD were randomized to semaglutide (titrated to 1.0 mg weekly) or placebo for 9 weeks. Semaglutide reduced laboratory alcohol self-administration, weekly alcohol craving, and drinks per drinking day relative to placebo; heavy drinking days fell numerically but did not reach significance. The trial is small, short, and used a lower dose than weight-loss protocols — but it is the strongest single piece of evidence in the field and the trial most often cited in 2025 coverage of the topic.

PMID 39937469 ↗DOI 10.1001/jamapsychiatry.2024.4789 ↗

Klausen 2022 is the earlier and larger GLP-1 randomized trial in alcohol use disorder. 127 adults with AUD were randomized to exenatide 2 mg weekly or placebo for 26 weeks alongside cognitive-behavioral therapy. The overall heavy-drinking-days endpoint was not significantly reduced. A pre-specified subgroup analysis in participants with BMI greater than 30 showed exenatide cut heavy drinking days substantially and reduced fMRI cue-induced ventral striatum activation. Negative top-line but mechanistically positive — and the obesity-subgroup signal drove much of the subsequent semaglutide and tirzepatide AUD research.

PMID 36066977 ↗DOI 10.1172/jci.insight.159863 ↗

Wang/Volkow 2024 used the TriNetX electronic-health-record network to compare cannabis-use-disorder outcomes in patients with type 2 diabetes or obesity who were prescribed semaglutide versus other antidiabetic or anti-obesity drugs. After propensity-score matching, semaglutide was associated with a 32 to 56 percent lower risk of incident cannabis use disorder and similar reductions in recurrence among patients with prior CUD. The study is observational, vulnerable to residual confounding, and does not establish causation — but it is the largest real-world signal that GLP-1s may reduce cannabis-related outcomes.

PMID 38486046 ↗DOI 10.1038/s41380-024-02498-5 ↗

The Wang/Volkow tobacco paper applied target-trial-emulation methods to TriNetX records of type 2 diabetes patients with tobacco use disorder. Semaglutide initiation was associated with substantially lower rates of medical encounters for tobacco use disorder and lower rates of smoking-cessation-drug prescriptions versus other antidiabetic comparators (insulins, metformin, dipeptidyl-peptidase-4 inhibitors, SGLT2 inhibitors, and other GLP-1s). Effect sizes were largest versus insulins. Observational and registry-based, but the target-trial design substantially reduces some forms of confounding.

PMID 39074369 ↗DOI 10.7326/M23-2718 ↗

The third paper in the Wang/Volkow TriNetX series compared opioid-overdose incidence in patients with type 2 diabetes and opioid use disorder who were prescribed semaglutide versus seven other antidiabetic drug classes. Semaglutide use was associated with a 42 to 68 percent lower hazard of incident opioid overdose across comparators after propensity-score matching. As with the cannabis and tobacco papers, the design cannot prove causation — patients prescribed semaglutide differ systematically from those prescribed insulins — but the magnitude and consistency of the signal have driven calls for randomized trials in OUD.

PMID 39320894 ↗DOI 10.1001/jamanetworkopen.2024.35247 ↗

Lahteenvuo 2025 leveraged Finnish nationwide registries covering 227,866 individuals with AUD, applying a within-individual design that uses each person as their own control across treatment exposure periods. Semaglutide use was associated with a 36 percent lower risk of AUD hospitalization and liraglutide with a 28 percent lower risk versus non-use periods. Naltrexone, the standard pharmacotherapy, produced a 14 percent reduction in the same model. The within-individual design controls for time-stable confounding but cannot rule out reverse causation or indication channeling. Strongest real-world AUD signal published to date.

PMID 39535805 ↗DOI 10.1001/jamapsychiatry.2024.3599 ↗

Quddos 2023 surveyed 153 adults with obesity who had initiated semaglutide or tirzepatide for weight loss. Roughly two-thirds reported reduced alcohol consumption after starting therapy and significantly fewer binge-drinking episodes versus a matched non-GLP-1 obesity cohort. Effect sizes were larger for tirzepatide than semaglutide in unadjusted comparisons. The study is small, self-reported, retrospective, and not randomized — but it is one of the earliest signals that the alcohol-reducing effect extends to dual GIP/GLP-1 agonists and is detectable at weight-loss doses outside the AUD-diagnosed population.

PMID 38017205 ↗DOI 10.1038/s41598-023-47934-8 ↗

Farokhnia/Leggio 2025 used administrative pharmacy and healthcare-claims data to compare alcohol-related outcomes between GLP-1 receptor agonist users and dipeptidyl-peptidase-4 inhibitor users — a methodologically important active-comparator design because both drug classes share diabetes indications but only GLP-1s engage central reward circuits. GLP-1 use was associated with significantly lower alcohol-related healthcare utilization than DPP-4 use. The DPP-4 active comparator substantially reduces healthy-user bias that affects studies using insulin or metformin as controls.

PMID 40048376 ↗DOI 10.1172/JCI188314 ↗

Martinelli 2024 systematically reviewed randomized controlled trials of GLP-1 receptor agonists in substance use disorder published through mid-2024. The review identified a small number of completed trials — chiefly Klausen 2022 exenatide-AUD plus smaller exenatide and liraglutide studies in smoking cessation and cocaine use disorder — and concluded that signals are heterogeneous, subgroup-driven (particularly in obesity), and insufficient to support clinical use. The review pre-dates Hendershot 2025 but remains the most comprehensive accounting of the randomized evidence base and the cleanest enumeration of which RCTs have actually been completed.

PMID 39288591 ↗DOI 10.1016/j.drugalcdep.2024.112424 ↗

Leggio 2023 is the field's leading honest framing of the evidence. Written by the senior NIH/NIDA team running much of the underlying preclinical and translational work — and by Hendershot and Farokhnia, lead authors of the two key 2025 papers in this list — the commentary lays out why the preclinical and observational signals are compelling, why the existing randomized data are not yet conclusive, and what trials would be needed to change clinical practice. It is the single most-cited piece in 2024 and 2025 reviews on the topic and the recommended starting point for clinicians fielding patient questions about prescribing GLP-1s off-label for addiction.

PMID 38001271 ↗DOI 10.1038/s41591-023-02634-8 ↗