Most GLP-1 coverage is about the scale. This study is about something else entirely: whether these medications also touch the brain’s reward circuitry in a way that affects drinking, smoking, and drug use. It’s a genuinely interesting finding, and it’s also a good case study in how a striking headline and a careful reading of the same data can pull in different directions.
The study
Published March 4, 2026 in The BMJ, researchers led by Ziyad Al-Aly followed more than 600,000 US veterans with type 2 diabetes over three years, comparing people who started a GLP-1 receptor agonist against people who started a different class of diabetes medication, SGLT-2 inhibitors. Because both groups were starting a new diabetes drug around the same time, the comparison is a reasonable way to isolate something about GLP-1s specifically, rather than just comparing “on a diabetes drug” to “not on one.”
Among veterans with no prior history of a substance use disorder, starting a GLP-1 was associated with:
- 14% lower risk of developing any substance use disorder
- 25% lower risk specifically for opioid use disorder
- 20% lower risk for cocaine and nicotine use disorders
- 18% lower risk for alcohol use disorder
- 14% lower risk for cannabis use disorder
Among veterans who already had a substance use disorder, starting a GLP-1 was associated with fewer downstream consequences: 31% fewer related ER visits, 26% fewer hospital admissions, 39% fewer overdoses, 25% fewer instances of suicidal ideation or attempts, and a 50% reduction in substance-related deaths.
Those last numbers are the ones that got this study picked up widely, and they’re also the ones worth reading most carefully.
The proposed mechanism
The leading hypothesis isn’t mysterious. GLP-1 receptors exist in the mesolimbic system, the brain circuitry that governs reward, motivation, and impulse control — the same machinery involved in food cravings, which is where the “food noise” description came from originally. Researchers describe a similar quieting effect across other reward-driven behaviors: less “drug noise,” not just less appetite. It’s a plausible biological story. Plausible is not the same as proven.
What independent reviewers flagged
When outside experts weighed in on the study, the caveats were specific, not generic hand-waving:
The effect sizes are more modest than the percentages suggest. A 14% relative risk reduction sounds large. In absolute terms, several of the individual substance use disorder outcomes moved by single digits per 1,000 people over three years. Both numbers are true at once — relative risk reductions and small absolute effects routinely coexist in this kind of research, and headlines tend to report only the first one.
This is an observational cohort study, not a randomized trial. Veterans weren’t randomly assigned to GLP-1s or SGLT-2 inhibitors — clinicians and patients chose, for reasons a database can’t fully capture. People started on a GLP-1 may have differed systematically from the comparison group in ways that also affect substance use risk: engagement with care, other health conditions, motivation to manage their diabetes more broadly. The researchers adjusted for known factors, but as reviewers noted, residual confounding from things like severity of prior substance use or social determinants of health can’t be ruled out.
The population is narrow. This is a VA cohort — predominantly older male veterans with type 2 diabetes. Whether the association holds in younger people, women, or people without diabetes taking a GLP-1 for weight management is genuinely unknown from this data.
One expert’s summary is the right way to hold this finding: hypothesis-generating, not confirmatory. Randomized trials designed specifically to test substance use outcomes would be needed before anyone could call GLP-1s a substance use disorder treatment. Established, evidence-based SUD treatments remain the standard of care — this data is not a reason to set them aside.
The trial that’s actually testing it
That’s not a hypothetical suggestion — it’s already happening. The Department of Veterans Affairs began recruiting in late July 2026 for a randomized, placebo-controlled trial testing semaglutide against placebo in more than 600 veterans with moderate-to-severe alcohol use disorder, across 18 VA medical centers. Participants get weekly injections for 24 weeks with a safety follow-up period after.
This is exactly the right next step, and its existence says something honest about where the science currently stands: promising enough to justify a real trial, not established enough to skip one.
Why this matters beyond the headline
If you’re taking a GLP-1 for weight management or metabolic health, this study is not a reason to expect it will change your relationship with alcohol, nicotine, or anything else — and it’s not something we’d ever suggest a compounded medication is prescribed to treat. The trials behind this hypothesis, including the one now enrolling, involve a specific manufactured product studied in a specific population for a specific condition. None of that data has been generated on a compounded preparation, and a real answer here is still years away.
What it does illustrate is worth sitting with regardless: a large, well-designed observational study can be both a genuinely useful signal and something that needs a randomized trial before anyone treats it as an answer. Both of those things are true of this one at the same time.
This post is general information, not medical advice. It does not describe a benefit of any NoBsRx program, and it is not a claim that any medication treats substance use disorders. Compounded medications are not FDA-approved. Treatment is never guaranteed — eligibility, diagnosis, and every treatment decision are made by independent licensed providers based on an individual clinical assessment, and availability varies by state. If you or someone you know is struggling with substance use, please talk to a licensed clinician or contact SAMHSA’s National Helpline at 1-800-662-4357.