GLP-1 and opioid overdose risk in veterans

A target-trial emulation published in the Journal of Clinical Psychiatry on 15 July 2026 found that veterans with type 2 diabetes and opioid use disorder who started semaglutide or tirzepatide had a lower 12-month risk of all-cause overdose compared with those who started insulin or SGLT2 inhibitors^[1]. The signal did not reach statistical significance against three other comparator drug classes, and the study's authors call for randomised controlled trials before drawing clinical conclusions.

What the study did

Kevin P. Kennedy and colleagues at the University of Pennsylvania used Veterans Health Administration records spanning January 2020 to December 2024. They identified veterans who had both type 2 diabetes and opioid use disorder and had newly started a GLP-1 receptor agonist (semaglutide or tirzepatide). Using propensity score matching and Cox proportional hazards regression, they built five separate comparisons^[1]:

GLP-1 RA vs insulin (n=630); vs SGLT2 inhibitors (n=432); vs metformin (n=1,016); vs sulfonylureas (n=710); vs DPP4 inhibitors (n=858). The primary outcome was any overdose event at 12 months.

The findings in full

Against insulin, the GLP-1 RA group had a hazard ratio of 0.32 (95% CI 0.14 to 0.71; P=.006), meaning roughly a 68% lower observed rate of overdose events^[1]. Against SGLT2 inhibitors, the hazard ratio was 0.25 (95% CI 0.09 to 0.68; P=.006). Neither comparator result against metformin, sulfonylureas, nor DPP4 inhibitors reached significance. The authors note that insulin and SGLT2 inhibitors share no known neurobiological pathway relevant to opioid reward, which makes them cleaner comparators for detecting a pharmacological signal specific to GLP-1 RA mechanisms.

The result against metformin being null is worth pausing on. Metformin has its own emerging evidence base for metabolic-neurological effects, and a null finding against it does not rule out a genuine GLP-1 effect, it just means the separation did not appear in this sample at this follow-up duration.

Why the biology makes sense

GLP-1 receptors are expressed in brain regions involved in reward processing, including the nucleus accumbens and the ventral tegmental area. Animal research has shown that GLP-1 receptor agonists reduce opioid self-administration and relapse-like behaviour in rodent models^[3]. The proposed mechanism involves cross-talk between GLP-1 signalling and the mesolimbic dopamine system, the same pathway that mediates the rewarding effects of opioids.

A 2025 real-world analysis published in Addiction by Qeadan and colleagues, which used records from 136 US health systems, found that GLP-1 or GIP receptor agonist prescriptions were associated with a 40% lower adjusted incidence rate of opioid overdose in patients with opioid use disorder (adjusted incidence rate ratio 0.60; 95% CI 0.43 to 0.83)^[2]. The Kennedy 2026 VHA study adds a more tightly matched, propensity-scored cohort to that signal, narrowing the comparison to a specific veteran population with both T2D and OUD.

What this is not

This is an observational emulation, not a randomised trial. The design cannot rule out residual confounding: veterans who receive GLP-1 receptor agonists may differ from those who receive insulin in ways that affect overdose risk, even after propensity score matching. There is no established causal pathway confirmed in humans. The VHA population is predominantly male and may not generalise to women or to civilian populations.

The study also does not address whether GLP-1 receptor agonists reduce opioid craving, relapse rates, or the subjective reinforcing effects of opioids directly. The outcome is overdose events, a composite of pharmacological, social, and behavioural factors.

What we do not yet know

No randomised controlled trial of semaglutide or tirzepatide specifically for opioid use disorder has yet reported results in humans. The most persuasive evidence for a mechanistic link remains preclinical. The dose required to produce any neurobiological effect relevant to addiction may differ from standard metabolic dosing, and this study cannot address that question. Whether the finding holds in people without type 2 diabetes (the conventional GLP-1 RA indication) is also untested.

The broader substance use picture

A 2026 systematic review in Frontiers in Pharmacology by Völker and colleagues identified 41 studies (35 preclinical, 6 clinical) across alcohol, nicotine, cocaine, and opioid use contexts^[3]. The preclinical signal was consistent: GLP-1 agonists reduced substance intake and relapse-like behaviour across compound classes. The six human studies offered preliminary support for alcohol use disorder but were small and variable. The Kennedy VHA result sits in that broader landscape: one of the first human studies with a comparator-controlled design to show a statistically significant overdose signal for opioids specifically.

Clinical and editorial note

Neither semaglutide nor tirzepatide carries an indication for opioid use disorder from the FDA, the EMA, or any national agency this site tracks. Both are approved for type 2 diabetes and, at higher doses, for chronic weight management. Anyone with opioid use disorder should discuss treatment options with a clinician experienced in addiction medicine. Evidence-based first-line pharmacotherapy for OUD remains buprenorphine and methadone. For more on GLP-1 receptor agonist pharmacology, see the semaglutide overview and the tirzepatide overview pages.

This article is for educational purposes only and does not constitute medical advice. Consult a qualified healthcare provider before making any treatment decisions.