Semaglutide and Suicidality in Adults with Overweight or Obesity: A Bayesian Meta-Analysis of Randomized Trials

There is no conclusive evidence linking semaglutide to increased or decreased suicidality, and findings are highly imprecise.

Summary and findings

This study evaluated the association between semaglutide and suicidality in adults with overweight or obesity through a Bayesian meta-analysis of randomized trials. The primary outcome was completed suicide, with secondary outcomes including suicide attempts and suicidal ideation. The findings indicated no conclusive evidence of increased or decreased suicidality associated with semaglutide.

Abstract

<title>Abstract</title> <p>Background Concerns have been raised regarding a potential association between glucagon-like peptide-1 receptor agonists and suicidality. Because suicidal outcomes are exceptionally uncommon in randomized trials, conventional meta-analytic estimates may be unstable and may exclude trials with no events. We evaluated semaglutide specifically across the clinical spectrum of suicidality. Methods We conducted a systematic review and meta-analysis of randomized trials comparing semaglutide with placebo in adults with overweight or obesity, with or without type 2 diabetes. The protocol was prospectively registered in PROSPERO (CRD420251059430) and substantively amended before data extraction to include completed suicide as the primary outcome and suicide attempt, suicidal ideation, and overall suicidal behavior/events as secondary outcomes. Arm-level counts were analyzed with a Bayesian random-effects binomial-logit model that retained double-zero studies without continuity correction. The primary prior for the pooled log-odds ratio (OR) was Normal (0,1), and an empirical Turner prior was used for between-study heterogeneity. The prior distribution represents the range and relative plausibility of treatment effects assumed before incorporating the observed trial data. We report posterior medians, 95% credible intervals (CrIs), 95% prediction intervals, and P(OR > 1) and P(OR > 1.25). Prior and frequentist sensitivity analyses were performed. Certainty of evidence was assessed separately for each outcome using GRADE across risk of bias, inconsistency, indirectness, imprecision, and publication bias. Results Fifteen unique placebo-controlled trials contributed to at least one outcome. Completed suicide was reported in eight trials (8/17,086 semaglutide vs 4/15,304 placebo; 12 events): pooled OR 1.56 (95% CrI 0.55–4.67), P(OR > 1) = 80.6%, and P(OR > 1.25) = 66.7%. For suicide attempt (12 trials; 5/19,046 vs 10/17,190), the OR was 0.51 (0.19–1.32), with P(OR > 1) = 9.0%. For suicidal ideation (eight trials; 12/17,366 vs 7/14,833), the OR was 0.93 (0.38–2.43), with P(OR > 1) = 44.8%. For overall suicidal behaviour/events (four trials; 12/11,061 vs 10/9,670), the OR was 1.06 (0.42–2.65), with P(OR > 1) = 54.6%. All credible and prediction intervals crossed unity, conclusions were qualitatively stable across all nine prior combinations for every outcome, and frequentist estimates were directionally concordant. GRADE certainty was very low for every outcome because of very serious imprecision and serious indirectness. Conclusions Randomized evidence did not provide conclusive evidence of either increased or decreased suicidality with semaglutide, and GRADE certainty was very low for all outcomes. The posterior distribution for completed suicide leaned toward increased odds, whereas suicide attempt leaned toward decreased odds; both estimates were highly imprecise and compatible with clinically important effects in either direction. Continued active surveillance and adequately powered studies with systematic psychiatric ascertainment are warranted.</p>

Limitations

Not reported in abstract.

Elsewhere in the Semaglutide corpus