A 600,000-veteran study links GLP-1 drugs to lower substance use disorder risk
Compared with a diabetes drug from another class, GLP-1 agonists were followed by fewer new diagnoses of alcohol, opioid, nicotine and other substance use disorders.
A cohort study published this week in The BMJ adds to a growing body of observational evidence that GLP-1 receptor agonists — the same drug class as semaglutide and tirzepatide — are associated with reduced risk of substance use disorders, this time in a population of more than 600,000 US veterans with type 2 diabetes [s1].
The design
Researchers used US Department of Veterans Affairs electronic health records to emulate eight target trials, comparing new users of GLP-1 receptor agonists against new users of SGLT-2 inhibitors, a different diabetes drug class used as an active comparator [s1]. From a base population of 606,434 veterans with type 2 diabetes, two protocols were run. The first, in people with no prior history of substance use disorder, followed 524,817 veterans — 124,001 who started a GLP-1 drug and 400,816 who started an SGLT-2 inhibitor — for up to three years, tracking new diagnoses of alcohol, cannabis, cocaine, nicotine, and opioid use disorders, plus a composite outcome [s1]. The second protocol followed 81,617 veterans who already had a substance use disorder — 16,768 GLP-1 starters and 64,849 SGLT-2 starters — for substance-use-related emergency visits, hospital admissions, and deaths [s1]. Because this is an emulation of a trial using observational data rather than a randomized trial itself, it can show association but cannot prove that GLP-1 drugs caused the differences.
What it found
Against SGLT-2 inhibitors, starting a GLP-1 receptor agonist was associated with lower three-year risk across every substance category measured [s1]:
- Alcohol use disorder: hazard ratio 0.82 (95% CI 0.78–0.85); 5.57 fewer cases per 1,000 people
- Cannabis use disorder: 0.86 (0.81–0.90); 2.25 fewer per 1,000
- Cocaine use disorder: 0.80 (0.72–0.88); 0.97 fewer per 1,000
- Nicotine use disorder: 0.80 (0.74–0.87); 1.64 fewer per 1,000
- Opioid use disorder: 0.75 (0.67–0.85); 0.86 fewer per 1,000
- Other substance use disorders: 0.87 (0.81–0.94); 1.12 fewer per 1,000
- Composite of all incident disorders: 0.86 (0.83–0.88); 6.61 fewer per 1,000
Among veterans who already had a substance use disorder, starting a GLP-1 drug was associated with fewer substance-related emergency department visits (hazard ratio 0.69, 8.92 fewer per 1,000), fewer substance-related hospital admissions (0.74, 6.23 fewer per 1,000), and lower substance-related mortality (0.50, 1.52 fewer per 1,000) [s1].
Reading the numbers carefully
These are hazard ratios in the range of 0.75–0.87 for new diagnoses — a 13–25% relative reduction — which is a real but moderate effect size, not a dramatic one. The absolute risk differences are small in per-person terms (roughly one to six fewer cases per 1,000 people over three years, depending on the substance) because substance use disorder diagnoses are relatively uncommon events even in a population this large. The mortality reduction among veterans with pre-existing substance use disorder is the more striking figure: a hazard ratio of 0.50 implies roughly half the risk of substance-related death, though the absolute reduction — 1.52 fewer deaths per 1,000 — reflects how the underlying event rate and confidence interval width.
The population is nearly entirely US military veterans with type 2 diabetes, a group that differs demographically and in health-system access from the general population; how this generalizes to people without diabetes, who now make up a large share of GLP-1 prescriptions, is not addressed by this study. The comparator, an SGLT-2 inhibitor, is itself an active diabetes drug rather than a placebo, so the results describe a difference between two treatments rather than GLP-1 drugs against no treatment.
Why researchers think this might be biologically plausible
GLP-1 receptors are expressed in brain regions involved in reward processing, and animal studies have previously suggested GLP-1 agonism can blunt reward-seeking behavior — a mechanism distinct from the drugs' metabolic effects on blood sugar and appetite. This cohort study does not test that mechanism directly; it is an epidemiological association drawn from prescribing and diagnosis records, not a trial designed around addiction as a primary outcome.
What to watch
Randomized trials of GLP-1 drugs specifically for substance use disorders, several of which are underway, will be needed to establish whether the association reflects a causal drug effect rather than differences between people whose clinicians chose a GLP-1 drug over an SGLT-2 inhibitor. This article describes an observational association involving a drug class with known safety considerations of its own; it is not medical advice and does not suggest GLP-1 drugs be used or prescribed for addiction treatment, which remains outside their approved indications.
Sources
Sources
More on
Semaglutide and tirzepatide cut alcohol-related hospitalizations by up to 65%
Four separate trial emulations, each with a different comparator, all pointed the same direction. The strongest effect appeared among people being treated for alcohol use disorder.
GLP-1 drugs, death rates and serious mental illness: a very large, very soft signal
A 1.5-million-person records study reports lower mortality with GLP-1 drugs than with SGLT2 inhibitors, and the largest absolute gains in people with serious mental illness. The design cannot prove it.
Ecuador logged 93,680 mental health hospitalisations in a decade
A national registry analysis found the highest burden among 20-to-29-year-olds and a notable rise among girls aged 10 to 19, with rates concentrated in the Sierra region.
Among GLP-1 users, oxycodone carried the highest short-term gut risk
A 411,000-patient claims study found adults on GLP-1 drugs who started oxycodone had more severe constipation and bowel obstruction at 30 days than those starting hydrocodone or tramadol.