FDA drops its suicide warning from GLP-1 obesity drugs

The warning had been on the label since reports filed from mid-2023 prompted a review, first previewed by the FDA in January 2024. Two years and two large analyses later, the agency concluded there was nothing there, and asked drugmakers to take the paragraph out. The review itself is substantial: a meta-analysis spanning 107,910 randomized-trial participants, plus a Sentinel System cohort of more than two million patients tracked against an active comparator rather than a placebo. Both point the same way, and European and British regulators reached the same conclusion earlier, in 2024. For an agency that is often accused of moving cautiously, this is a clear reversal, not a hedge.

Who the trials never enrolled

The reversal rests on data that was never built to answer the question a psychiatrist now has to ask. The 91 trials in the meta-analysis are obesity and diabetes trials, and obesity trials of this kind have a standard exclusion list: a current major depressive episode, a history of schizophrenia or bipolar disorder, a suicide attempt within the past two years. An independent analysis of the SURMOUNT program, cited alongside the FDA's own review, makes this exclusion explicit. The Sentinel cohort is administrative claims data, drawn from insurance records rather than a study designed around psychiatric status, so it cannot fill the gap either way. Neither dataset can tell a clinician what happens when a GLP-1 agonist meets an active psychotic disorder or a recent mood episode, because neither dataset contains that patient.

None of this makes the FDA's finding wrong. A meta-analysis of over 100,000 people and a national claims cohort of over two million are not small evidence. But the population now receiving GLP-1 agonists is no longer only the population the trials enrolled. As their metabolic benefits reach patients with schizophrenia, bipolar disorder and treatment-resistant depression – exactly the patients carrying the metabolic burden this issue covers – the question the label warning used to flag has moved to a group the underlying trials excluded, at the same time the warning is coming off.

One more thread sits outside the FDA's own analysis. A separate study using the WHO's VigiBase pharmacovigilance database reported a disproportionate suicidality signal for semaglutide, more pronounced among patients also taking an antidepressant. VigiBase reports are not causal evidence – they cannot rule out confounding by indication – but they mark a combination, GLP-1 plus antidepressant, that neither the meta-analysis nor the Sentinel cohort was built to isolate.

The FDA's own instruction to patients has not changed: anyone on a GLP-1 agonist should still tell a clinician about new or worsening suicidal thoughts. A removed label warning changes what the paperwork says. It does not change what belongs in the intake conversation with a patient whose psychiatric history the pivotal trials would have screened out.