A review of GLP-1 and dual GIP/GLP-1 drugs found no consistent pattern of psychiatric harm in the human evidence, but it did not rule out every possible risk. Its central judgment is that a link is biologically plausible, yet the evidence remains too mixed to determine whether most reported neuropsychiatric outcomes are treatment-related.
A broad review found a mixed picture
The review examined mood, suicidal ideation, self-harm, anxiety, cognitive effects, eating disorders and substance misuse. It brought together molecular and systems-level brain studies, animal models, randomized trials, observational cohorts, case-control analyses and post-marketing pharmacovigilance.
This was a narrative appraisal, not a single pooled estimate. The authors searched six bibliographic and scholarly sources using drug and neuropsychiatric terms, then synthesized studies with different designs and outcome definitions. Those differences meant they did not conduct a formal meta-analysis.
Under the Bradford Hill framework, the overall evidence was rated weak to moderate and the criteria were only partly met. Animal models showed strong consistency, but that pattern did not translate consistently to human outcomes. That split helps explain why the review can describe biological plausibility without concluding that treatment caused psychiatric problems.
The largest human studies found no rise in self-harm
One of the clearest checks came from an FDA clinical-trial meta-analysis covering 91 trials and 107,910 patients. Of those patients, 60,338 received a GLP-1 receptor agonist and 47,572 received placebo. The analysis found no increased risk of suicidal ideation or behavior versus placebo, with similar results for anxiety, depression, irritability and psychosis.
An FDA Sentinel retrospective cohort provided another large comparison. It included 1,161,983 people who started a GLP-1 receptor agonist and 1,081,155 who started an SGLT2 inhibitor. GLP-1 use was not associated with an increased risk of intentional self-harm compared with SGLT2 inhibitor use.
Still, post-marketing evidence raised questions. Pharmacovigilance sources reported possible associations between liraglutide, semaglutide and tirzepatide and suicidal thoughts or actions, with higher reporting probabilities described for semaglutide and liraglutide than for other GLP-1 receptor agonists. The review presented these as possible associations rather than a settled treatment effect and said the wider evidence was insufficient to determine whether most outcomes were treatment-related.
Other outcomes offered few clear answers
Anxiety results were largely unremarkable. Clinical data showed low, placebo-comparable rates, while genetic and meta-analytic analyses reported null associations. The review found no consistent signal for psychosis or mania.
Cognition was similarly reassuring, with an important caveat about the evidence. A systematic review of 22 studies involving 186,847 participants found no robust evidence of clinically meaningful cognitive toxicity. In some cases, it reported meaningful improvement in cognitive and affective functioning.
Evidence on eating disorders was mixed. One meta-analysis found a significant difference in eating-disorder outcomes between GLP-1 exposure and control, while a retrospective cohort of adults with obesity found no significant worsening of eating-disorder risk or psychological distress. The contrasting findings do not settle whether the medicines improve, worsen or have no effect on these outcomes.
Outside mental health, a separate pharmacovigilance signal involved eye disorders. In a FAERS analysis of disproportionate reporting, using metformin and orlistat as controls, the review examined reporting odds ratios above 4. Among people without type 2 diabetes, compared with metformin, the reported ratios were 9.424 for semaglutide and retinopathy, 10.253 for semaglutide and retinal hemorrhage, 6.475 for tirzepatide and eye swelling, and 9.628 and 9.557 for liraglutide and cataract and macular degeneration, respectively. A reporting odds ratio compares how often an event is reported between groups, so these figures do not provide incidence rates or establish direct retinal toxicity. The review also says some displayed confidence intervals were internally inconsistent and that absolute risks were not reported.
The review also found a potentially favorable association in substance use. In a cohort of 606,434 US veterans with type 2 diabetes followed for up to three years, starting a GLP-1 receptor agonist was associated with a lower risk of incident substance use disorders, or SUDs. Among participants who already had an SUD, GLP-1 use was associated with lower risks of several SUD-related harms. Because this was observational evidence, it cannot establish that treatment caused the lower risks.
The uncertainty is the main finding
Taken together, the review's message is cautious. The medicines do not show a consistent psychiatric harm signal across the largest analyses, but the evidence is not strong enough to determine whether most reported outcomes are treatment-related or whether rare reactions are absent. The authors call for prospective, mechanism-informed studies to resolve the remaining uncertainty.
Paper data and sources
Original title: GLP-1 and dual GIP/GLP-1 receptor agonists' psychopharmacology and putative neuropsychiatric associated effects: a Bradford Hill-informed, systematic, evaluation.
Authors: F Schifano, M A De Luca, S Bonaccorso et al.
Journal/Repository: Current psychiatry reports
Status: Peer-reviewed
First online: 2026-08-21
DOI: 10.1007/s11920-026-01709-w
Original paper · Full text