Do GLP-1 Drugs Raise Suicide Risk? What Trials Show

GLP-1 receptor agonists such as semaglutide and liraglutide do not show an increased risk of suicide or self-harm in the best available placebo-controlled evidence. A 2025 meta-analysis found a non-significant result that pointed in the opposite direction: a pooled rate ratio of 0.76 (95% CI, 0.48 to 1.21). That is not proof that these drugs protect mental health. It does mean the trials did not confirm the harm signal that prompted regulatory scrutiny.

The concern was reasonable to investigate. Spontaneous reports to Iceland's medicines agency described suicidal thoughts or behavior among people taking liraglutide and semaglutide. Regulators at the FDA and European Medicines Agency opened reviews. The alarm also drew on the observation that suicide risk can rise after bariatric surgery. But a safety signal is a question, not an answer. Spontaneous-report databases cannot tell whether a drug caused an event, especially when the treated population already has high rates of depression, anxiety, and other psychiatric illness.

What the randomized trials found

The strongest evidence comes from a 2025 JAMA Psychiatry systematic review and meta-analysis. The investigators screened 144 placebo-controlled GLP-1 receptor agonist trials lasting at least six months, but only 27 trials had prospectively recorded suicide or self-harm outcomes. That is a major weakness in the development record: metabolic trials collected cardiovascular and metabolic outcomes more consistently than psychiatric safety outcomes.

The 27 eligible trials included 32,354 participants assigned to a GLP-1 receptor agonist and 27,042 assigned to placebo. They contributed 69,653 and 63,853 person-years of follow-up, respectively. Suicide or self-harm events were rare: 0.047 per 100 person-years in the GLP-1 groups and 0.042 per 100 person-years in the placebo groups.

After pooling the trials, the investigators reported a rate ratio of 0.76 (95% CI, 0.48 to 1.21; P = 0.25). The confidence interval includes 1.0, so the result was not statistically significant. It remains compatible with lower risk, no meaningful difference, or some increase in risk. There was no significant heterogeneity across trials, and subgroup analyses did not find differences by diabetes status or specific GLP-1 drug.

That last point deserves restraint. Rare outcomes demand very large studies. With event rates this low, the analysis likely lacked power to rule out a clinically meaningful increase, including a possible drug-specific effect. A non-significant result does not certify psychiatric safety; it narrows the case for a broad class-wide suicide hazard.

Why the bariatric-surgery comparison fits poorly

Bariatric surgery does have a documented association with elevated postoperative suicide risk, but surgery and GLP-1 treatment do not create the same biological circumstances. Surgery can cause micronutrient malabsorption involving nutrients such as vitamin B12, folate, and zinc; alter alcohol pharmacokinetics after bypass; and impose an acute surgical and psychological burden. GLP-1 receptor agonists do not produce those surgical exposures.

The comparison did provide a prudent reason to look for psychiatric harms. It did not provide a mechanism showing that GLP-1 drugs should cause them. In fact, GLP-1 receptors occur in reward and limbic circuits, including the mesolimbic dopamine system, hippocampus, and amygdala. A 2025 review of GLP-1 drugs in psychiatric illness describes hypotheses in which receptor activation affects neuroinflammation, dopamine signaling, and hypothalamic-pituitary activity in directions that could benefit mood. Those are plausible mechanisms, not evidence that a prescription prevents suicide.

Observational and psychiatric evidence point the same way

Randomized trials are not the only relevant evidence. A 2024 Annals of Internal Medicine target-trial emulation used U.S. electronic health records from older adults with type 2 diabetes. It compared GLP-1 receptor agonist use with active comparators and found no increased risk of suicidal ideation or behaviors. Observational studies cannot remove every source of confounding, but active comparators and a target-trial framework make this design more informative than a raw count of adverse-event reports.

Psychiatric research is beginning to address a population that metabolic trials usually left out. A 2026 randomized clinical trial of semaglutide in people with active major depressive disorder reported pro-cognitive effects. The supplied result concerns cognition, not suicide or self-harm, so it cannot settle the safety question. Still, it offers direct human evidence that semaglutide did not obviously precipitate psychiatric deterioration in that trial population.

A 2025 network meta-analysis also reported concentration-dependent protective associations for delirium, depression, and dementia. These outcomes differ from suicidality, and associations across a network meta-analysis do not establish a suicide-prevention effect. Together with preclinical work on neuroinflammation and neurotrophic signaling, including a 2022 review of GLP-1 analogues in neurological disease models, the evidence makes a simple suicidogenic mechanism less convincing. It does not eliminate the need to watch individual patients.

The signals that started the alarm still matter

The regulatory concern did not arise from nowhere. A 2023 pharmacovigilance analysis described the Icelandic reports and noted a slight numerical imbalance in suicidal ideation with liraglutide in pooled weight-management trial adverse-event reporting. This is the most substantial evidence supporting a possible liraglutide-specific concern.

Yet the same trials did not show that imbalance when they used prospective structured questionnaires. That disagreement matters because unstructured adverse-event reporting can vary with patient reporting, clinician attention, and how events are coded. It weakens the argument for a direct pharmacological effect, but it does not fully explain the discrepancy.

A 2025 VigiBase disproportionality study also found psychiatric adverse-event reports disproportionately associated with dulaglutide, semaglutide, and liraglutide. Such analyses generate hypotheses; they cannot establish incidence or causation. People with obesity and type 2 diabetes have substantial baseline psychiatric burden, and reporting can rise sharply once a possible drug risk receives public attention.

Individual reports remain clinically useful as warnings. A 2024 case report described depressive symptoms in one person with type 2 diabetes after liraglutide dose escalation, with symptoms resolving after discontinuation. One case cannot quantify risk or prove causation, but it supports a practical point: population averages do not prevent idiosyncratic reactions.

The evidence gap is concentrated in high-risk patients

Most metabolic GLP-1 trials excluded people with active psychiatric disorders, a history of suicidal behavior, or concurrent psychotropic medication. The trial meta-analysis therefore speaks most directly to a psychiatrically lower-risk population, not to every person now receiving these drugs for obesity or diabetes.

Follow-up also limits reassurance. Most trials were too short to detect psychiatric effects that might emerge after prolonged treatment or major weight change, and five of the 27 included trials had more than 5% loss to follow-up. If psychiatric symptoms led some participants to stop treatment and disappear from outcome collection, trial estimates could undercount them. There are also no long-term psychiatric safety data beyond two years for high-dose obesity formulations such as semaglutide 2.4 mg weekly or for tirzepatide in people with psychiatric risk factors.

For people represented in the trials, current evidence does not support a GLP-1-related increase in suicide or self-harm. For people excluded from those trials, the honest answer is that the evidence remains thin.

What to do with this information

For a person without active psychiatric illness, prior suicidal behavior, or psychotropic medication use, the controlled evidence does not justify avoiding a GLP-1 receptor agonist because of a presumed suicide risk. The point estimate in the randomized evidence leans below neutral, but its uncertainty is too wide to call that protection.

For someone with active depression, suicidal thoughts or history, or changing psychiatric medication, treatment deserves individualized discussion and active symptom monitoring. That approach does not assume GLP-1 drugs are dangerous. It recognizes that the people with the most to lose from a psychiatric adverse effect were largely absent from the trials designed to detect one. The next useful study is not another speculative count of reports; it is prospective psychiatric safety research in the patients clinicians actually worry about.