Adults on methadone for opioid use disorder who also have type 2 diabetes were nearly twice as likely to enter OUD remission, and almost three quarters less likely to have suicidal thoughts or behaviors, when their diabetes was managed with a GLP-1 receptor agonist instead of without one.
That is the finding of a 2,314-patient propensity-matched cohort study published online May 18 in the Journal of General Internal Medicine ↗. The authors pulled adults aged 18 and older from the TriNetX research database between January 2015 and October 2024, kept only those on methadone with both an OUD diagnosis and a hemoglobin A1c at or above 6.5 percent, and split them into two groups of 1,157 each by whether they had ever been prescribed a GLP-1 RA. After propensity score matching for baseline characteristics, Cox proportional hazard models compared one-year incidence of cardiometabolic and mental-health endpoints.
GLP-1 RA patients had a 42 percent lower risk of myocardial infarction (HR 0.58, 95% CI 0.41 to 0.80). They had a 50 percent lower risk of hypoglycemia (HR 0.50, 0.33 to 0.77). Mean age was 57 years, half were women, and the follow-up was a single calendar year, so these are short-horizon effects in a high-event-rate group rather than long-tail prevention numbers.
The mental-health column is the surprise. Major depression dropped 29 percent (HR 0.71, 0.61 to 0.81). Anxiety disorders dropped 23 percent (HR 0.77, 0.67 to 0.88). Suicidal thoughts or behaviors dropped 73 percent (HR 0.27, 0.16 to 0.47). And OUD itself moved: the GLP-1 RA group was 75 percent more likely to enter remission of their opioid use disorder over the same year (HR 1.75, 1.24 to 2.47).
Why this is unusual
The trials that defined the GLP-1 class for cardiometabolic outcomes (LEADER, SUSTAIN-6, REWIND, SELECT) excluded active substance-use disorder, severe mental illness, and methadone maintenance therapy as a matter of routine. The class-effect literature was built on a population that was the literal opposite of the one here. So a retrospective cohort that takes the excluded patients and tracks them in a real-world prescribing context is one of the only ways to get any signal at all, and the signal points the same direction the trial population's signal pointed, only larger.
The OUD-remission and suicidality numbers also speak to a separate question the field has been circling for two years. Animal data and small clinical reads have suggested that GLP-1 receptors in the ventral tegmental area and nucleus accumbens modulate dopaminergic reward signaling, which would predict an effect on opioid craving distinct from the cardiometabolic axis. Whether the suicidality result here is mediated by less craving, less withdrawal-driven dysphoria, weight loss, glycemic stability, or some confounded combination is not something a propensity-matched cohort can answer. But the direction and magnitude are consistent with the dopaminergic-modulation hypothesis rather than against it.
Note what is not here. The study is not randomized, prescribers chose who got a GLP-1 RA and who did not, and propensity matching captures only observed covariates. The TriNetX database under-captures social determinants and severity of OUD at index. Methadone dose, retention in treatment, and concomitant naloxone-buprenorphine use are not in the published tables. The authors call their findings "hypothesis-generating" in the conclusion, which is the right register; do not read these as effect sizes a randomized trial would reproduce.
Where this lands
Semaglutide ↗ is the canonical GLP-1R agonist the rest of the class is judged against, and it is on formulary for type 2 diabetes in roughly the same prescribing rhythm that produced this cohort. The mechanism page for GLP-1R ↗ lists the receptor's expression in pancreas, gut, and central reward circuitry; the methadone-OUD finding is a reminder that the third tissue compartment is doing work the cardiometabolic trials were not designed to measure.
The class has now been associated with reduced incidence of stroke (LAMP post-hoc, HOMA-IR-stratified, May 2026), reduced cardiorenal events (meta-analysis, May 2026), reduced alcohol consumption (semaglutide AUD RCT, May 2024), and now reduced OUD recurrence and suicidality (this paper). The shape of the cumulative literature is no longer "GLP-1 RA lowers HbA1c and weight, and incidentally protects the heart." It is closer to "GLP-1 RA is a partial reset on the metabolic-and-reward circuit that most addiction and chronic-disease pathology eventually routes through, and the cardiometabolic numbers were the first ones we measured because they were the easiest to count."
A randomized trial of liraglutide or semaglutide in methadone-treated OUD patients with metabolic syndrome would be small (a few hundred patients), short (12 months for OUD-relapse endpoints), and feasible inside an existing harm-reduction clinic network. The hard part is the IRB. After this paper, the IRB conversation gets easier.