Largest Subject-Level Trial Analysis Finds No Raised Suicide Risk With GLP-1 Drugs

A sweeping reanalysis of subject-level data from 91 placebo-controlled clinical trials has found no evidence that glucagon-like peptide-1 receptor agonists, the drug class behind some of the world’s most widely prescribed diabetes and weight-loss medicines, increase the risk of suicidal ideation or behavior. The analysis, published in Drugs & Therapy Perspectives, pooled individual participant data across GLP-1 RA development programs approved in the United States from 2005 to 2023, covering 107,910 randomized participants and roughly 271,000 person-years of follow-up. It is described by its authors as the largest known subject-level analysis of clinical trial participants across these development programs.

The work was carried out after postmarketing reports of suicidal ideation and behavior in patients taking GLP-1 RAs prompted the US Food and Drug Administration to open an investigation in July 2023. A preliminary FDA review of clinical trial and postmarketing data, including observational studies and case reports, did not support an association, but regulators noted considerable uncertainty in the risk estimates because individual trials contained only small numbers of these rare events. That uncertainty motivated the present analysis: by combining subject-level records across every GLP-1 RA product, investigators could sharpen the precision of the risk estimate far beyond what any single trial, or a meta-analysis of trial-level summaries, could achieve.

The trials included were prospective, randomized, placebo-controlled phase 3 or phase 4 studies evaluating glycemic control in people with type 2 diabetes, weight reduction in people with obesity or overweight, or cardiovascular risk reduction. Of the 160 controlled trials that met the prespecified criteria, 91 had a placebo control: 62 enrolled participants with type 2 diabetes as the primary population and 29 enrolled participants with obesity or overweight, a group that could include people with diabetes. Eight trials were cardiovascular outcomes trials, each enrolling at least 3,000 participants, and together these eight contributed 63 percent of the total analysis population. Twenty-two trials collected the Columbia Suicide Severity Rating Scale, a prospective instrument for assessing suicidal ideation and behavior.

The primary outcome was the first occurrence of suicidal ideation or behavior, defined as a binary composite endpoint. Events were identified retrospectively using the Standard Medical Dictionary for Regulatory Activities Query for suicide and self-injury, broad scope, and all potential events were adjudicated and classified by applicants using the Columbia Classification Algorithm of Suicide Assessment, which distinguishes suicidal ideation, both passive and active, from suicidal behavior such as attempted or completed suicide. Secondary endpoints covered anxiety, psychosis, irritability, insomnia, and depression, identified through FDA-prespecified groupings of MedDRA preferred terms. Statistical comparisons used Mantel-Haenszel methods stratified by trial to estimate both a rate ratio and a rate difference comparing GLP-1 RA with placebo, with a rate ratio above 1 indicating potential harm.

The headline numbers were strikingly balanced. The GLP-1 RA arm recorded 55 events of suicidal ideation or behavior during 14.3 × 10⁴ person-years, corresponding to an incidence rate of 3.6 events per 10⁴ person-years. The placebo arm recorded 44 events during 12.8 × 10⁴ person-years, also 3.6 events per 10⁴ person-years. The pooled rate ratio was 1.0 with a 95 percent confidence interval of 0.7 to 1.5, and the rate difference was 0.0 with a 95 percent confidence interval of −1.4 to 1.5. Twenty-seven of the 91 trials contributed events to the rate ratio estimate, and a forest plot showed every individual trial estimate with a confidence interval that included the null value of 1.0. Heterogeneity statistics did not indicate substantial inconsistency across trials.

Sensitivity and subgroup analyses reinforced the null result. Alternative follow-up windows of on-treatment plus 90 days and on-treatment plus 365 days yielded rate ratios of 1.2 (95 percent CI 0.8 to 1.8) and 1.0 (95 percent CI 0.7 to 1.6), respectively, consistent with the primary on-study analysis. In trials grouped by population, the rate ratio was 1.0 (95 percent CI 0.5 to 1.7) in type 2 diabetes trials and 1.0 (95 percent CI 0.6 to 1.8) in obesity or overweight trials. Subject-level subgroups by sex, age, race, ethnicity, body mass index, and psychiatric history produced estimates ranging from 0.8 to 4.0, but every subgroup with a rate ratio above 1.5 had a very wide confidence interval that included the null, reflecting statistical variability from tiny event counts rather than a credible signal of harm.

A complementary analysis of the 22 trials that prospectively collected the Columbia Suicide Severity Rating Scale also supported the null finding. In that subset there were more suicidal ideation than suicidal behavior events, 109 versus 6, and more passive than active ideation events, 58 versus 51, a pattern the authors attribute to the instrument’s high sensitivity for detecting ideation. The rate ratio was 0.7 (95 percent CI 0.5 to 1.0), and the rate difference of −25.8 (95 percent CI −51.6 to 0.0) did not indicate a harmful effect. For the other psychiatric endpoints, rate ratios ranged from 0.9 to 1.3 with confidence intervals that included 1.0 for anxiety, depression, irritability, and psychosis. The only nominally elevated estimate was insomnia in the on-treatment plus 365 days analysis, with a rate difference of 5.6 events per 10⁴ person-years (95 percent CI 0.1 to 11.1). Notably, 31 of the 44 participants reporting a psychosis-related adverse event also had a history of a psychotic disorder, such as schizophrenia, at baseline.

The findings align with a broader body of evidence. The authors note consistency with null results from published cohort studies and with meta-analyses of trial-level summaries, including one of 27 studies reporting a rate ratio of 0.76 (95 percent CI 0.48 to 1.21) for suicidality and self-harm and another of 80 studies reporting 0.98 (95 percent CI 0.82 to 1.19) for serious psychiatric adverse events. FDA analyses of claims data in the Sentinel distributed database likewise found no increased risk of intentional self-harm comparing GLP-1 RAs with sodium-glucose cotransporter-2 inhibitors (hazard ratio 0.93; 95 percent CI 0.81 to 1.08) or dipeptidyl peptidase-4 inhibitors (hazard ratio 0.94; 95 percent CI 0.82 to 1.07) among people with type 2 diabetes. A review by the European Medicines Agency’s Pharmacovigilance Risk Assessment Committee also concluded there was no link between GLP-1 RA use and suicidal ideation or behavior.

The study has limitations the authors acknowledge candidly. Suicidal ideation and behavior were identified retrospectively through the standardized query in most trials, although adjudication confirmed that accidental overdose and injury cases were not misclassified, and the prospective C-SSRS subset corroborated the null result. Targeted collection of baseline psychiatric diagnoses was lacking in many trials, so a large share of participants had missing psychiatric history data, and estimates for the no-diagnosis category may not accurately reflect the population. Some weight-reduction trials excluded people with severe psychiatric disease, limiting generalizability to high-risk groups, and future work comparing GLP-1 RAs against active controls rather than placebo would be informative for clinical decisions.

The practical consequences are already visible. Based on the combined evaluation, including this subject-level analysis, the Sentinel System analysis, and reviews of published observational and pooled studies, the FDA issued an updated Drug Safety Communication stating that its evaluation did not identify an increased risk of suicidal ideation or behavior with GLP-1 RA medications, and asked drug application holders to remove the relevant risk language from prescribing labels where it appeared.

Subject of Research: Subject-level analysis of placebo-controlled GLP-1 receptor agonist trials evaluating suicidal ideation and behavior and other psychiatric endpoints

Article Title: GLP-1 receptor agonists and suicidal ideation/behavior: subject-level analysis of clinical trials

Article References: Tran, V., Golden, J., Higginbotham, L., Kalaria, S., Kim, C. Y., Nguyen, E., Penzenstadler, J., Sharretts, J., & Houstoun, M. (2026). GLP-1 receptor agonists and suicidal ideation/behavior: subject-level analysis of clinical trials. Drugs & Therapy Perspectives. https://doi.org/10.1007/s40267-026-01266-5

Image Credits: AI Generated

DOI: 10.1007/s40267-026-01266-5

Keywords: GLP-1 receptor agonists, suicidal ideation, clinical trials, FDA, placebo-controlled trials, psychiatric safety, type 2 diabetes, obesity, drug safety, rate ratio, pharmacovigilance, semaglutide