No rise in suicidal thinking among 9,222 teenagers on GLP‑1 drugs, and the apparent benefit did not survive a second test

The largest study so far of whether obesity medicines harm teenagers' mental health found no increase in suicidal thinking, depression, anxiety, insomnia or eating disorders over three years. It also found something its own authors flag as a warning about the method: the apparent benefit disappeared when the comparison group was changed.

The cohort study was published online by Diabetes, Obesity and Metabolism on September 14, 2026. It matched 9,222 adolescents starting semaglutide, liraglutide or tirzepatide against 9,222 receiving lifestyle care, using deidentified electronic health records from the TriNetX Global Collaborative Network.

Hundreds of times larger than the 44-patient chart review published the same day

OTN covered a 44-patient chart review from a single Wisconsin clinic on September 14, which found depression scores fell and no suicide screen worsened. This study asks the same question at a scale that can actually detect an uncommon harm, across many health systems rather than one clinic, and over up to three years rather than a few visits. Each arm here holds more than 200 times the patients in that entire review.

Both point the same way. The difference is that a 15-patient and 23-patient analysis could not have ruled out a meaningful risk, and a 9,222-per-group analysis has a real chance of doing so.

Suicide-related events were less common in the treated group, not more

Outcomes were counted from day 30 to day 1,095 after treatment started. The primary outcome combined suicidal ideation and suicide attempt.

Treated adolescents had a hazard ratio of 0.74 for suicide-related events, with a 95% confidence interval from 0.57 to 0.95, and 0.73 for suicidal ideation alone, from 0.56 to 0.96. Anxiety came in at 0.92, from 0.84 to 0.99, a difference small enough to sit at the edge of the range.

For suicide attempt, insomnia, depression and eating disorders, the study found no significant difference either way. The reductions were statistically clear among girls, hazard ratio 0.61 from 0.46 to 0.82, and among adolescents without type 2 diabetes, 0.63 from 0.47 to 0.85. They were not clear among boys or among those with type 2 diabetes.

Against metformin instead of lifestyle care, the benefit vanished

This is the part that deserves more attention than the headline hazard ratios. The authors also ran the comparison against adolescents taking metformin, another prescribed medicine, and found no significant difference.

That matters because it suggests the apparent protection is not coming from the drug. Teenagers who get any prescription have already seen a clinician who chose to treat them. Comparing them to a lifestyle-care group compares two different kinds of patient and two different levels of medical contact, not just two treatments. When both groups are on a prescription, the gap closes.

The authors say so plainly, writing that the lower risk should be interpreted cautiously given potential selection bias and the absence of a significant difference in the metformin analysis. The safety reading holds. The benefit reading does not.

Records show what was coded, not what teenagers felt

An electronic health record study counts diagnoses that a clinician entered. Suicidal thoughts that a teenager never disclosed, or that were discussed but never coded, do not appear. If treated adolescents are seen more often, that can cut either way: more chances to record a problem, or more attention that heads one off.

The 30-day lead-in before counting begins is a standard guard against picking up events that were already underway when treatment started. It also means anything happening in the first month is not in these numbers.

Where this sits against the regulatory record

US regulators reviewed this question directly. On January 13, 2026, FDA said a comprehensive review had found no increased risk of suicidal thoughts or behavior and asked the manufacturers of Saxenda, Wegovy and Zepbound to remove the suicidal ideation and behavior warning from their prescribing information. This study is consistent with that position and adds adolescent-specific evidence to it.

None of this replaces monitoring. New or worsening low mood, anxiety or thoughts of self-harm in a young person on any of these medicines is a reason to contact the prescribing clinician promptly rather than wait for a scheduled visit.