Semaglutide users had 51% lower suicidal ideation risk than older GLP-1 users

Concerns about depression and suicidal ideation as potential neuropsychiatric side effects of GLP-1 receptor agonists have circulated since the drugs' rise in obesity and diabetes treatment, but whether risk differs meaningfully across the different drugs in this class has been unclear. A large real-world cohort study published this month in Diabetes, Obesity and Metabolism tests that question directly, comparing tirzepatide against semaglutide, and separately, semaglutide against older GLP-1 drugs [s1].

The design

Researchers used the TriNetX Global Federated Network, covering more than 192 million patients, to conduct a retrospective cohort study [s1]. Adults with type 2 diabetes, obesity, or both, who started tirzepatide, semaglutide, or another GLP-1 receptor agonist between July 2022 and June 2025, were identified using a new-user design with a 12-month washout period (excluding people who'd recently used a comparable drug, to better isolate new-start risk) [s1]. Propensity score matching balanced the comparison groups on demographic, clinical, and metabolic characteristics [s1]. Two separate comparisons were run: tirzepatide versus semaglutide, and semaglutide versus other (earlier-generation) GLP-1 receptor agonists [s1]. Outcomes — incident depression, anxiety, and suicidal ideation — were tracked using Cox models across two follow-up windows: Year 1 (days 31–365) and a Year 2 landmark analysis (days 366–730) [s1].

What it found

After matching, 85,546 pairs were analyzed for tirzepatide versus semaglutide, and 80,115 pairs for semaglutide versus other GLP-1 drugs [s1]. Tirzepatide and semaglutide showed similar risk for a composite psychiatric outcome across both years (Year 1 hazard ratio 0.984, 95% CI 0.950–1.019; Year 2 hazard ratio 1.002, 95% CI 0.960–1.046) [s1]. A nominally higher anxiety hazard appeared with tirzepatide specifically in Year 2 (hazard ratio 1.052, 95% CI 1.001–1.106) — a result the study's own authors say should be interpreted cautiously given the number of comparisons run across the analysis [s1].

The more striking finding came from the second comparison: against other, earlier-generation GLP-1 receptor agonists, semaglutide was associated with meaningfully lower risk across every psychiatric outcome measured during Year 1 — depression (hazard ratio 0.811, 95% CI 0.770–0.855), anxiety (hazard ratio 0.915, 95% CI 0.871–0.961), suicidal ideation (hazard ratio 0.488, 95% CI 0.339–0.702), and the composite outcome (hazard ratio 0.866, 95% CI 0.832–0.901) [s1].

Why the suicidal ideation number needs careful handling

A hazard ratio of 0.488 for suicidal ideation is a large effect — implying roughly half the risk with semaglutide compared with older GLP-1 drugs — but its confidence interval (0.339 to 0.702) is noticeably wider than the other outcomes in this comparison, reflecting the fact that suicidal ideation, while a serious outcome worth taking seriously, occurs less frequently than depression or anxiety diagnoses in a matched cohort like this, producing fewer events for the statistical model to work with. The direction and general magnitude of the finding — semaglutide performing better than older comparators — is consistent across all four outcomes measured, which lends some coherence to the pattern, even as the suicidal ideation estimate specifically carries more statistical uncertainty than the others.

What "caution given multiple comparisons" means, and why the authors flagged it themselves

The study's own explicit caveat about the tirzepatide-versus-semaglutide anxiety finding (Year 2 hazard ratio 1.052) is worth taking at face value: when a study runs many statistical comparisons — as this one does, across two drug pairings, three outcome measures, a composite outcome, and two time windows — some results will cross the threshold for statistical significance by chance alone, even if there's no true underlying difference. The authors flagging their own single nominally-significant result this way is a sign of appropriately careful interpretation, not evidence the finding should be dismissed outright — simply that it shouldn't be treated as a confirmed anxiety risk specific to tirzepatide without further, more targeted study.

What this doesn't establish

This is a retrospective, real-world cohort study — even with propensity score matching, unmeasured confounding by indication (why a clinician or patient chose one drug over another) remains possible, particularly relevant here since which GLP-1 drug someone starts may correlate with disease severity, cost or insurance access, or other factors that could independently relate to psychiatric outcomes. The study measures diagnosed depression, anxiety, and suicidal ideation via healthcare records, which captures clinically recognized and coded events rather than the full spectrum of psychiatric symptoms patients may experience.

What to watch

Whether the semaglutide-versus-older-drugs psychiatric advantage, and the tirzepatide anxiety signal specifically, are confirmed in independent cohorts or in the psychiatric safety data reported from ongoing and future GLP-1 clinical trials. This article is not medical advice.

Sources

1. Neuropsychiatric Outcomes With Tirzepatide, Semaglutide, and Other GLP-1 Receptor Agonists — Diabetes, Obesity and Metabolism, 8 July 2026