GLP-1 drugs and mental health: a 2026 study

On 29 June 2026, Communications Medicine published an observational cohort study by Huang et al. examining whether adults starting tirzepatide or semaglutide experienced different rates of neuropsychiatric diagnoses compared with people starting other approved weight-loss medications^[1]. The findings show lower hazard ratios for anxiety and depression in both drug groups, with some nuance depending on whether participants had type 2 diabetes (T2D).

If you are currently taking either drug and have questions about your own mood or mental health, the right person to speak with is your prescribing clinician, not an article. Do not adjust or stop your medication based on population-level data.

What the study did

Huang and colleagues used U.S. electronic health records from 2020 to 2025, matching adults newly starting tirzepatide or semaglutide one-to-one with users of naltrexone-bupropion, phentermine, or phentermine-topiramate. Participants were then tracked for up to 24 months for a first recorded neuropsychiatric diagnosis^[1]. The analysis split results by T2D status, because the metabolic profile and baseline mental-health burden differ between the two groups.

The comparison group matters here. Naltrexone-bupropion is itself a drug with neuropsychiatric mechanisms (bupropion is an antidepressant). Choosing it as the active comparator is methodologically deliberate, but it also means the hazard ratios are relative to a medication that already has mood-related pharmacology. For background on how semaglutide and tirzepatide work, see their respective peptide pages.

The headline numbers

For semaglutide compared with naltrexone-bupropion, the study found lower anxiety rates in both the T2D group (hazard ratio 0.67, 95% CI 0.51-0.87) and the non-T2D group (HR 0.73, CI 0.64-0.83). Depression rates were lower in both groups as well (T2D: HR 0.69, CI 0.50-0.95; non-T2D: HR 0.53, CI 0.45-0.61).

For tirzepatide, the anxiety and depression associations were in the same direction and, on the point estimates, somewhat stronger. T2D patients on tirzepatide had a hazard ratio for depression of 0.49 (CI 0.33-0.72) versus naltrexone-bupropion. Non-T2D patients had HR 0.57 (CI 0.49-0.66).

One finding runs the other way: in the non-T2D group, tirzepatide was associated with a modestly higher rate of anxiety (HR 1.12, CI 1.06-1.18) and insomnia (HR 1.13, CI 1.05-1.21) compared directly with semaglutide. The absolute magnitude is small, and the authors note it requires replication with validated psychiatric assessments.

Why these findings need careful reading

Observational data do not establish causation. The authors themselves flag residual confounding and the risk of diagnosis misclassification as key limitations. Electronic health records capture diagnoses that were recorded, not the true incidence of symptoms; people on newer, more closely monitored drugs may simply be seen by clinicians more often, leading to more diagnoses being filed.

Context from the broader literature adds caution in both directions. A 2023 pharmacovigilance analysis in European Psychiatry (PMID 38031404) documented case reports of suicidal and self-injurious behaviours with liraglutide and semaglutide submitted to regulators, though disproportionality in a reporting database is not the same as an elevated incidence rate. Separately, a 2026 adverse-event analysis found semaglutide carried the strongest signal for psychiatric adverse events, notably anxiety, in its dataset. The Huang et al. study points in a more favourable direction, but the literature as a whole is not yet settled.

For context, the EMA and FDA have both reviewed the neuropsychiatric safety of GLP-1 receptor agonists and to date have not added new neuropsychiatric warnings to the product labels for tirzepatide or semaglutide at approved doses.

What this is not

This study is not a clinical trial. It does not prove that tirzepatide or semaglutide protect against depression or anxiety. It does not show that either drug causes the favourable difference, only that the difference existed in this dataset. It cannot tell individual patients what to expect from their own treatment course. And it does not override any prior concern you or your clinician have discussed.

It is also not a reason to start either drug specifically for a neuropsychiatric indication. Both are approved for type 2 diabetes and obesity management. Prescribing them outside those indications is a clinical decision that belongs with a qualified healthcare provider who has reviewed your full history.

What to watch next

The authors call for follow-up studies using standardised psychiatric assessments, not just diagnostic codes. Longer follow-up periods would help characterise whether any difference persists beyond 24 months. The slight divergence between tirzepatide and semaglutide on anxiety in the non-T2D group is worth watching in prospective data. Regulators continue to monitor post-market safety reports for both drugs, and any label changes would be the clearest signal that the evidence base has shifted^[1].