September 1, 2026 · Science · Nutrition
What Was Published
A review published in the journal Diabetology on August 31, 2026 consolidated five years of mechanistic, pharmacovigilance, observational and regulatory evidence on GLP-1 receptor agonists and mental health. As reported by MedicalXpress, it found no link between these medications and increased suicidal thoughts, depression or other serious psychiatric harm, and referenced a pooled analysis of 91 clinical trials showing no increase in psychiatric risk.
That conclusion is consistent with where regulators landed two years earlier. In April 2024, the European Medicines Agency Pharmacovigilance Risk Assessment Committee concluded that the available evidence "does not support a causal association" between GLP-1 receptor agonists and suicidal and self-injurious thoughts and actions. The substances named were dulaglutide, exenatide, liraglutide, lixisenatide and semaglutide. The committee had reviewed non-clinical studies, clinical trial data, post-marketing surveillance, a real-world electronic health records study finding no causal link between semaglutide and suicidal thoughts, and an analysis EMA conducted itself. The review had opened in July 2023 after case reports involving liraglutide and semaglutide. No product information changes were required.
The short version
Why the Confidence Interval Is the Story
The 2025 systematic review and meta-analysis in Diabetes/Metabolism Research and Reviews identified 11 studies, of which four were eligible for pooling. Its reported risk ratio of 0.568 sits below 1, which points toward lower risk, and its interval of 0.077 to 4.205 crosses 1 by a wide margin in both directions. In practice that means the data are consistent with these drugs reducing suicidal outcomes by more than 90% and with them increasing such outcomes fourfold. A result that broad does not distinguish between those possibilities.
The authors' own limitations explain why the interval is so wide. Heterogeneity across studies was reported at 98%, meaning the included studies disagreed with each other far more than chance would explain. Most of the included studies drew on pharmacovigilance databases, which the authors note are prone to underreporting and reporting bias. The prediction interval, which describes the range a future study might land in, was reported as 0.001 to 218.938. And they identify the exclusion of people with a history of mental illness from most clinical trials as a source of selection bias.
A phrase like "no significant link" carries two very different meanings that get collapsed in coverage. It can mean a study was precise enough to rule out a meaningful effect. It can also mean a study lacked the precision to detect one. Here the confidence interval identifies which of those applies.
The Signal That Did Appear
The more actionable finding in the 2026 review is not the headline. As reported, patients already taking antidepressants or benzodiazepines showed a substantially amplified reporting signal for suicidal ideation, suggesting risk concentrated among people with pre-existing psychiatric vulnerability rather than spread evenly across everyone treated.
That finding needs its own caveat. A reporting signal derives from spontaneous adverse event databases, where clinicians and patients voluntarily submit suspected reactions. Such databases have no denominator, so they cannot produce a rate, and they are subject to stimulated reporting, in which publicity about a suspected harm increases reports of it regardless of whether the underlying rate changed. Someone taking an antidepressant is also, by definition, someone with a documented psychiatric history and a higher baseline risk of the outcome being counted. A raised signal in that group is expected on those grounds alone.
What the finding supports is the review's reported recommendation: individualized screening for psychiatric history and suicidality before starting treatment, and closer monitoring for people with a history of mood disorders or concurrent psychiatric medication use. That is a process recommendation, and it holds whether or not a causal effect exists.
Why This Matters on a Testing Platform
Several of the platforms benchmarked here prescribe GLP-1 medications inside a subscription built around lab panels and a dashboard. That model is efficient at ordering bloodwork and structurally weak at psychiatric assessment, which depends on history, continuity and a clinician who knows the person rather than on any marker a panel reports.
The screening the review recommends is a conversation, not a test. There is no blood marker for psychiatric history, current antidepressant use, or suicidality, and no panel a longevity platform sells substitutes for asking. For anyone considering one of these programs, whether the intake process includes a structured mental health history, and whether there is a named clinician to contact between appointments, is a reasonable question to ask before enrolling. Our guide to doctor-led options covers how the clinical models behind these subscriptions differ, and the metabolic reversal hub covers programs that address the same conditions without medication.
Anyone currently taking one of these medications who notices a change in mood should raise it with the prescribing clinician promptly. Anyone experiencing thoughts of suicide or self-harm should seek immediate help from a clinician or a local crisis service. Nothing on this page is a basis for starting, stopping or changing a medication.
What Would Change This Read
A randomized trial that deliberately enrolled people with psychiatric history would change it most. The central weakness in this evidence base is that the design capable of establishing causation excluded the population of interest. A trial that included them, with psychiatric outcomes as prespecified endpoints, would produce the estimate that currently does not exist.
Large linked health-record cohorts with adequate follow-up would narrow the interval. Electronic health records carry a denominator, which pharmacovigilance databases lack, and EMA has already used one such study. More of them, in different health systems, would tighten a pooled estimate that four studies currently leave very wide.
Related Coverage
A separate briefing covers the updated Annals of Internal Medicine review of 38 GLP-1 weight loss trials, which reports the efficacy figures alongside a gastrointestinal adverse event rate of 76.0% and the authors' note that safety outcomes were inconsistently reported. That inconsistency and the gap described here are the same underlying problem seen from two directions. A third briefing reports a Mendelian randomization study on GLP-1 receptor activity and male pattern hair loss, where the signal comes from the gene rather than from anyone taking the drug. The nutrition hub and the metabolic reversal hub collect the guides these decisions start from.
Sources
- European Medicines Agency, "Meeting highlights from the Pharmacovigilance Risk Assessment Committee (PRAC) 8-11 April 2024," ema.europa.eu (accessed September 1, 2026). Source of the PRAC conclusion and its quoted wording, the five named active substances, the July 2023 start of the review and the case reports that prompted it, the categories of evidence reviewed including the electronic health records study, and the statement that no product information changes were required.
- "Association of GLP-1 Receptor Agonists With Risk of Suicidal Ideation and Behaviour: A Systematic Review and Meta-Analysis," Diabetes/Metabolism Research and Reviews, February 13, 2025, PMC11823376 (accessed September 1, 2026). Source of the 11 studies identified and 4 pooled, the risk ratio of 0.568 with its 95% confidence interval of 0.077 to 4.205, the finding of no statistically significant difference, the 98% heterogeneity figure, the prediction interval of 0.001 to 218.938, the note on pharmacovigilance underreporting and reporting bias, and the selection bias from excluding people with a history of mental illness.
- MedicalXpress, "GLP-1s do not cause major psychiatric harm, according to new study," August 31, 2026, medicalxpress.com (accessed September 1, 2026). Source of the description of the 2026 Diabetology review, its DOI 10.3390/diabetology7080144, its integrative design across five years of evidence, its reported conclusion on suicidal ideation, depression and anxiety, the referenced pooled analysis of 91 clinical trials, the amplified reporting signal among patients on antidepressants or benzodiazepines, the screening and monitoring recommendation, and the identified research gaps in adolescents and people with serious mental illness.
- The Diabetology paper's publisher page at mdpi.com returned an access error at the time of writing, and a PubMed search for DOI 10.3390/diabetology7080144 returned no matching record. Every statement about that paper on this page is attributed to the MedicalXpress report above, and no figure is taken from the paper itself.
Frequently Asked Questions
What did European regulators actually conclude?
In April 2024, the European Medicines Agency Pharmacovigilance Risk Assessment Committee concluded that the available evidence does not support a causal association between GLP-1 receptor agonists and suicidal and self-injurious thoughts and actions. The five substances named were dulaglutide, exenatide, liraglutide, lixisenatide and semaglutide. The committee reviewed non-clinical studies, clinical trial data, post-marketing surveillance, a real-world electronic health records study and an analysis EMA conducted itself. No product information changes were required.
Why did the review start in the first place?
It began in July 2023 following case reports of suicidal thoughts and thoughts of self-injury in people using liraglutide and semaglutide. Regulators investigate signals of this kind routinely, and starting a review is not itself evidence that a drug causes harm. The purpose is to determine whether reported cases reflect a causal effect or the background rate in a treated population.
Does "no significant link" mean the drugs are proven safe for mental health?
No, and the distinction matters. A 2025 systematic review and meta-analysis in Diabetes/Metabolism Research and Reviews pooled four studies and reported a risk ratio of 0.568 with a 95% confidence interval of 0.077 to 4.205. That interval stretches from a large reduction in risk to a roughly fourfold increase, so it is compatible with a wide range of true effects. The finding is best read as insufficient precision to detect a difference, not as a demonstration that no difference exists.
Who is missing from this evidence base?
The people the question is most about. The same meta-analysis lists as a limitation that most clinical trials excluded individuals with a history of mental illness, which introduces selection bias. Randomized trials are the strongest design available, and if they systematically omitted people with psychiatric history, they cannot answer what happens to that group. The newer review is reported to identify remaining gaps in adolescents and in people with serious mental illness.
What is the reported concern about people already taking psychiatric medication?
The 2026 review is reported to find a substantially amplified reporting signal for suicidal ideation among patients already taking antidepressants or benzodiazepines. A reporting signal comes from pharmacovigilance databases, which collect voluntary reports of suspected adverse events and cannot establish causation or measure a rate. It is a flag for closer attention rather than a measured risk, and the review is reported to recommend individualized screening for psychiatric history and suicidality before starting treatment.
What should someone on a GLP-1 who notices mood changes do?
Raise it with the clinician who prescribed it, and do so promptly rather than waiting for a scheduled review. That applies whether or not any drug is suspected of causing the change. Nothing on this page is a basis for starting, stopping or altering a medication independently. Anyone experiencing thoughts of suicide or self-harm should seek immediate help from a clinician or a local crisis service.