Can GLP-1 Medicines Curb Opioid and Cannabis Use? What the Addiction Research Shows

Can GLP-1 medicines curb opioid and cannabis use? Large health-record studies say people taking them develop fewer substance use disorders and have fewer overdoses. No completed randomized trial has yet shown that GLP-1 medicines treat opioid or cannabis use disorder, and the one human laboratory test with cocaine came back negative.

The idea grew out of patient reports. People on semaglutide described losing interest in drinking and smoking, and researchers started checking whether the effect reached other drugs. We have covered the alcohol and nicotine evidence separately, in our articles on GLP-1s and alcohol cravings and GLP-1s and quitting smoking.

This article looks at the substances with less coverage: opioids, cannabis and cocaine. It covers a 2026 BMJ study of more than half a million veterans, two overdose analyses, a cannabis cohort, a human laboratory test with cocaine, and the trials now under way. It ends with what this means, and does not mean, for low-dose protocols.

Why would a GLP-1 medicine affect drug use at all?

The reward system, mostly in rodents

GLP-1 receptors are found in brain regions that handle reward and motivation, so the same signal that dampens interest in food may dampen interest in drugs. Our article on GLP-1s and dopamine goes deeper into the reward pathway itself.

The strongest support for the idea comes from animals. Patricia Grigson and colleagues reviewed the opioid evidence in Biological Psychiatry in 2026. In rats, exendin-4, liraglutide, semaglutide and tirzepatide all reduced opioid taking or seeking, whether the seeking was triggered by drug cues, the drug itself, or stress. The review also flags a catch: tolerance developed with chronic administration of higher doses. Rodent results are a reason to run human trials, not evidence that the drugs work in people.

Do people on GLP-1 medicines develop fewer substance use disorders?

The 2026 BMJ veterans study

In the largest study so far, veterans who started a GLP-1 medicine were less likely to develop a new substance use disorder of every type measured. Miao Cai, Ziyad Al-Aly and colleagues published the analysis in the BMJ in 2026, drawing on records of 606,434 US veterans with type 2 diabetes. The main comparison set 124,001 people starting a GLP-1 against 400,816 starting an SGLT-2 inhibitor, a diabetes drug with no known effect on reward, and followed them for up to three years.

The hazard ratio for a new opioid use disorder was 0.75. For cocaine and nicotine it was 0.80, for alcohol 0.82 and for cannabis 0.86. Across all substance use disorders combined, the hazard ratio was 0.86, which the authors translated into 6.61 fewer cases per 1000 people over three years. Comparing against another diabetes drug, rather than against people taking nothing, removes some of the bias that comes from comparing sicker and healthier groups.

What about people who already have a substance use disorder?

Overdoses, emergency visits and deaths

In veterans who already had a substance use disorder, starting a GLP-1 was linked to fewer of the outcomes that matter most. The same BMJ study ran a second analysis of 81,617 people with an existing disorder, 16,768 of whom started a GLP-1. Compared with SGLT-2 starters, the hazard ratio was 0.61 for drug overdose, 0.50 for substance-related death, 0.69 for substance-related emergency visits and 0.74 for substance-related hospital admissions. Suicidal ideation or attempt was also lower, at 0.75. All participants had type 2 diabetes, so the results may not carry over to people without it.

Are GLP-1 medicines linked to fewer opioid overdoses?

Two more studies, and a warning about comparisons

Yes in several studies, but how strong the link looks depends on the comparison. Fares Qeadan and colleagues, in Addiction in 2025, analysed records of 503,747 patients with opioid use disorder across about 136 US health systems. Those with a GLP-1 or GIP/GLP-1 prescription had an adjusted incidence rate ratio of 0.60 for opioid overdose. In 817,309 patients with alcohol use disorder, the ratio for alcohol intoxication was 0.50.

A smaller 2026 study by Kevin Kennedy and colleagues in the Journal of Clinical Psychiatry shows why one number is never enough. Among veterans with type 2 diabetes and opioid use disorder who started semaglutide or tirzepatide between 2020 and 2024, overdose risk over 12 months was much lower than in people starting insulin (hazard ratio 0.32) or an SGLT2 inhibitor (0.25). Against metformin, sulfonylureas and DPP4 inhibitors, there was no significant difference. Insulin users tend to have more advanced disease, so a big gap against insulin can reflect who gets insulin as much as anything the GLP-1 does.

Can semaglutide help with cannabis use disorder?

One large cohort, no trials yet

Semaglutide users were less likely to receive a cannabis use disorder diagnosis in one large records study, but no clinical trial has tested it. Wang and colleagues, with Nora Volkow of the National Institute on Drug Abuse, reported in Molecular Psychiatry in 2024 that more than 45 million Americans use cannabis, about one-third of them with cannabis use disorder, and no medicine is FDA-approved to treat it.

The team compared 85,223 patients with obesity who were prescribed either semaglutide or a non-GLP-1 weight-loss medicine. Over 12 months, semaglutide was linked to a hazard ratio of 0.56 for a first cannabis use disorder diagnosis and 0.62 for a recurrent one. In a replication with 596,045 patients with type 2 diabetes, the figures were 0.40 and 0.66, though the recurrent result in that group did not reach statistical significance. A diagnosis in a medical record is a rough measure of use, because many people who use cannabis heavily are never diagnosed.

What happened when a GLP-1 was tested directly against cocaine?

A controlled human laboratory test

A controlled human laboratory test found no effect. Gustavo Angarita and colleagues at Yale, in Drug and Alcohol Dependence in 2021, gave 13 people with cocaine use disorder a single 5 mcg injection of exenatide or placebo 3 hours before a laboratory session in which they could self-administer cocaine.

Participants took 8.5 cocaine infusions after exenatide and 9.1 after placebo, a difference that was not significant. Ratings of euphoria and of wanting cocaine were also unchanged. The authors were clear about the limits: a single dose, one dose level, and an older, short-acting GLP-1. The result does not rule out an effect from weeks of treatment with a long-acting drug like semaglutide, but it is a reminder that the rodent results have not yet been reproduced in a human experiment.

What clinical trials are testing GLP-1 medicines for addiction?

Mostly alcohol and nicotine, a few for opioids

Most registered trials target alcohol and nicotine, with only a handful for opioids and none for cannabis. Sanjana Patil and colleagues searched ClinicalTrials.gov up to July 2, 2025 and found 33 GLP-1 trials in substance use disorders, published in Addictive Behaviors Reports in 2026. Semaglutide was the most studied drug, in 15 trials, followed by exenatide in 8 and tirzepatide in 6.

For opioids, the human signal so far is one small liraglutide study that, according to the Grigson review, found a 40% reduction in opioid craving. Larger trials are testing semaglutide and tirzepatide as add-ons to buprenorphine or methadone. The National Drug Abuse Treatment Clinical Trials Network is running one of them, called TAB. Described by Theresa Winhusen and colleagues in Contemporary Clinical Trials in 2026, it plans to randomize about 310 adults with moderate to severe opioid use disorder who have recently started buprenorphine to tirzepatide or placebo across 9 sites. The main outcome is whether people stay on buprenorphine for 6 months, which is the point where many treatment courses fail.

Why is the evidence still not enough to recommend GLP-1s for addiction?

What observational data can and cannot show

Apart from the small liraglutide craving study, every positive human result in this article is observational, and observational studies cannot rule out that GLP-1 users differ in ways the records miss. People who get a newer, often expensive medicine may have better insurance, more regular doctor visits and more stable lives, all of which lower addiction risk on their own. Comparing against another diabetes drug helps, as the BMJ study did, but the Kennedy results show the answer can change with the comparison.

The outcomes are also blunt. A diagnosis code or an emergency visit captures the most severe end of drug use, not day-to-day consumption or craving. And almost all the large studies involve people with type 2 diabetes or obesity, which leaves open whether the effect, if real, applies to anyone else. These are the questions the randomized trials are designed to answer.

Does a microdose GLP-1 protocol reduce drug cravings?

Untested, and dose may cut both ways

No study has tested a microdose GLP-1 protocol for opioid, cannabis, cocaine or any other substance use disorder, so any effect at those doses is unknown. The large cohorts drew on standard diabetes and obesity prescriptions and did not report results by dose.

The dose question is genuinely open in both directions. The cocaine experiment that found nothing used a low single dose of a short-acting drug, which could mean low doses are too weak. The rodent work found tolerance with chronic higher doses, which could mean more is not always better. Neither finding tells us what a low weekly dose does in a person. Anyone hoping a GLP-1 protocol will help with drug use should treat that as a question for an addiction specialist, not something to test alone.

What should someone with a substance use disorder know before trying a GLP-1?

Practical points from the evidence

Frequently Asked Questions