The same gut-brain receptor that tells you when you have eaten enough sits all over the brain’s reward circuit. The drug everyone calls an appetite shot binds it. So when 606,000 U.S. veterans with type 2 diabetes were tracked for three years and the ones on GLP-1 receptor agonists turned in lower rates of new addiction, fewer overdoses, and fewer drug-related deaths across alcohol, opioids, cocaine, cannabis, and nicotine, the result is less surprising than it looks. It is the receptor map finally getting a population-sized test.

The study landed in The BMJ on March 4, 2026, led by Ziyad Al-Aly at Washington University in St. Louis and the VA Saint Louis Health Care System, and ScienceDaily picked it up again this week. Al-Aly’s team pulled VA electronic health records and compared veterans on GLP-1 receptor agonists (semaglutide, liraglutide, dulaglutide, and others in the class) with veterans on SGLT2 inhibitors, a different modern diabetes class that lowers blood sugar through an entirely separate pathway. That comparison choice matters, and I will come back to it.

Among the 524,817 veterans with no substance use disorder at baseline, the GLP-1 group had a 14% lower risk of developing one over follow-up. Alcohol fell 18%, cannabis 14%, cocaine 20%, nicotine 20%, opioids 25%. Roughly seven fewer new diagnoses per 1,000 people on a GLP-1. Modest at the individual level, large at the population level.

The numbers in the other arm of the study are what made me sit up. Among the 81,617 veterans who already had a substance use disorder when they started a GLP-1, overdoses dropped 40%. Drug-related deaths dropped 50%. Addiction-related ER visits fell 30%, hospitalizations 25%. The authors translate that to roughly 12 fewer serious addiction-related events per 1,000 users. That is not a craving score shifting on a self-report. That is people who would have died, not dying.

So why would a drug designed to slow gastric emptying and goose pancreatic insulin show up in the overdose statistics? Because the receptors it activates are not parked exclusively in the gut and pancreas. They are dense on neurons in the ventral tegmental area, the nucleus accumbens, the nucleus of the solitary tract, and the lateral hypothalamus. Roughly: the dopamine-source neurons that fire when something feels rewarding, the dopamine-target neurons that say “do that again,” and the hunger-and-drive nucleus tying both to homeostasis. Animal work links GLP-1 receptor activation in those accumbens and VTA pathways to blunted dopamine signaling after alcohol and other addictive substances, as the Frontiers in Pharmacology systematic review out this year laid out in detail. The pleasure spike gets smaller. The signal to repeat gets weaker. The same hormone the gut uses to tell the brain “I am full” appears to dampen the brain’s response to “I want more,” across alcohol, opioids, stimulants, and nicotine at once.

A small randomized trial in alcohol use disorder published in JAMA Psychiatry in 2025 found that semaglutide reduced drinks per drinking day and craving, although it did not move every weekly drinking measure the team tracked. Smaller scale and a narrower question than the VA paper, but the direction matches. The mechanism story has been stacking up quietly. This cohort is what gave it population-scale numbers.

There is a lens issue, though, and it shapes how much the cohort can carry. This is a VA population: heavily older, heavily male, heavily white, and selected for being sick enough with diabetes to be on either of these drug classes. The team emulated a target trial through statistical adjustment, but they could not randomize who got what. They cannot rule out that the kind of veteran prescribed a GLP-1 and kept on it is, in some unmeasured way, the kind of veteran with more engaged primary care, more stable housing, more of every confound that lowers addiction risk on its own. The authors say so themselves. The accompanying BMJ editorial from Fares Qeadan at Loyola spells it out: established medications for substance use disorder remain the preferred treatment, and any GLP-1 benefit on addiction outcomes is, for now, a bonus in patients for whom the drug is already clinically warranted.

And the comparison arm is SGLT2 inhibitors, not placebo. The question this study can answer is whether GLP-1s do better on addiction outcomes than another modern diabetes drug. The answer there is yes. What the addiction-outcome curves look like against no drug at all, or against a randomized non-pharmacological intervention, is a different study that has not been run.

Sit with that while reading the next round of headlines. Novo Nordisk and Eli Lilly are running one of the most aggressive label-expansion campaigns the industry has put together in decades, because every new indication unlocks an enormous prescribing surface for an already-blockbuster class. Numbers this favorable, on outcomes this hard, in a population this big, will be moving through industry-funded continuing-medical-education decks long before any randomized addiction-outcome trial reads out. The data are the data. The interpretation that arrives wrapped in those decks will not be neutral, and the gap between an observational signal and a labeled indication is where pharma marketing lives.

On the safety side, the FDA in January asked manufacturers to remove the suicidal-behavior-and-ideation warning language from the labels of Saxenda, Wegovy, and Zepbound, citing a meta-analysis of 91 placebo-controlled trials in roughly 108,000 patients that found no signal. Worth knowing, with the standard asterisk: regulator track records on revising warnings for blockbuster drugs are not unblemished, and a chronic weekly injectable is a different shape of intervention than buprenorphine, methadone, naltrexone, acamprosate, or nicotine replacement, all of which already exist, all of which are evidence-based for substance use disorder, and all of which remain underused.

What I make of all this. Addiction medicine has, institutionally, been treated as a problem to be solved with substance-specific medications inside abstinence-heavy and punishment-heavy systems. A gut-brain peptide quietly turning down reward gain across multiple substances at once points at a unified mechanism that decades of single-substance research kept missing. It makes “is this also a metabolic disease” a serious question rather than a rhetorical one.

What I want to see next are randomized trials with the same hard endpoints this paper reported: overdose, hospitalization, death. The authors and the editorialist both flag those trials as the next move, and they are right to. If those trials replicate even a fraction of these effect sizes, the addiction-medicine pipeline gets rebuilt around a new class. If they don’t, we will have learned something important about how much an observational VA cohort can really tell us about a brand-new question. Either way, the gut hormone keeps showing up in the reward circuit. At this point, you have to take that seriously.

Sources

  1. The BMJ – GLP-1 receptor agonists and risk of substance use disorders among US veterans with type 2 diabetes: cohort study (Al-Aly et al., 2026)
  2. Washington University School of Medicine – GLP-1 medications get at the heart of addiction: study (2026)
  3. BMJ Group press release – GLP-1 diabetes drugs linked to reduced risk of addiction and substance-related death (2026)
  4. Frontiers in Pharmacology – The potential role of GLP-1 receptor agonists in substance use disorders: systematic review (2025)
  5. ScienceDaily – Popular GLP-1 weight-loss drugs linked to lower risks of addiction and overdose (2026)
  6. FDA – Removal of suicidal behavior and ideation warning from GLP-1 RA medications (2026)
  7. JAMA Psychiatry – Hendershot et al., Once-weekly semaglutide in adults with alcohol use disorder: a randomized clinical trial (2025)