Scientific deep-dive

GLP-1s and Worsening Mental Illness

Semaglutide was associated with 42% lower risk of worsening mental illness in people with depression and anxiety. Exenatide and dulaglutide showed nothing. Four drugs, one class, two very different results.

By Nora Bissett · Pricing Editor
Editorially reviewed. Not medical advice · How we verify contentLast reviewed
5 min read·1 citation

A Swedish cohort followed people with depression and anxiety who took GLP-1 drugs, and found something more interesting than a simple answer. Semaglutide was associated with a 42% lower risk of worsening mental illness. Liraglutide was associated with an 18% lower risk, and exenatide and dulaglutide with none at all.[1]

What was found

Adjusted hazard ratios for worsening mental illness against non-use, in people with depression and anxiety.
DrugWorsening mental illness
Semaglutide0.58 (95% CI 0.51–0.65)
Liraglutide0.82 (95% CI 0.76–0.89)
Exenatide1.01 (95% CI 0.69–1.46)
Dulaglutide1.01 (95% CI 0.85–1.20)

For semaglutide the effect appeared across all three specific outcomes: worsening depression at 0.56, worsening anxiety at 0.62, and worsening substance use disorder at 0.53. Liraglutide moved only depression, at 0.74. Across the class as a whole, self-harm was lower, at 0.56 (confidence interval 0.34 to 0.92).[1]

Why the split between drugs is the finding

We have now seen this shape twice. In a meta-analysis of atrial fibrillation, the class was neutral and the signal appeared specific to semaglutide, which we covered in reading a 42% headline. Here again, two drugs move and two do not. The wider record on suicidal thoughts and psychiatric safety is gathered in ten studies on GLP-1s and mental health.

An effect that lands on some members of a drug class and not others is either real pharmacology or a description of who receives which drug.

Both explanations are live. The design removes one large source of bias: it compares each person’s periods on a drug with their own periods off it, so stable differences between people cancel out.[1] It cannot remove everything. Each drug’s estimate comes only from the people who took it, and semaglutide, the newest and most effective of the four, reached different people than exenatide. Who reaches any of them is skewed as well: in diabetes, people with a mental health diagnosis have a quarter of the odds of being prescribed a GLP-1. Whatever changes around the time someone starts a drug, such as closer contact with their care, can also move a mental-health trajectory.

There is also a straightforward mediating route that has nothing to do with the brain: semaglutide produces far more weight loss than exenatide, and losing a great deal of weight changes mobility, sleep, pain and how people are treated by others. A drug that improves depression through those channels is still helping, but that is not evidence of a direct psychiatric mechanism.

The self-harm result, and where it sits

The class-wide self-harm finding matters because it speaks to the harm this drug class has been most publicly accused of. A hazard ratio of 0.56 points in the opposite direction from harm, in a population selected for existing depression and anxiety, the group in which any risk would show up first.

Set that against the spontaneous-reporting signal we covered in the FAERS analysis, which pointed the other way. These are not equally strong designs: a national cohort that compares each person with themselves beats voluntary reports with no denominator, and the regulatory picture is covered in what happened with the suicidality signal.

None of this makes a GLP-1 a psychiatric treatment, or a replacement for one; how these drugs sit alongside an existing antidepressant is covered in our antidepressant interaction guide. The authors’ own conclusion is that for anxiety and depression co-occurring with diabetes and obesity these drugs might be dually effective options, and that randomized trials are warranted. A randomized trial of semaglutide for cognitive dysfunction in depression has since missed its primary endpoint, which we covered separately.

Frequently Asked Questions

This cohort found the opposite. Among people with depression and anxiety, semaglutide was associated with a 42% lower risk of worsening mental illness and liraglutide 18% lower, while exenatide and dulaglutide showed no association either way.
It is either a real pharmacological difference or a description of who gets which drug. The study compared each person with themselves, which cancels out stable differences between people. But each drug’s result comes from the people who took it, semaglutide reaches different patients than exenatide, and changes around the time a drug is started can also affect mental health.
Plausibly. Semaglutide produces far more weight loss than exenatide, and losing a great deal of weight changes mobility, sleep, pain and how people are treated. A drug helping depression through those channels is still helping, but that is not evidence of a direct psychiatric effect.
Across the class, self-harm was lower, with a hazard ratio of 0.56. That is notable because the population was selected for existing depression and anxiety, the group in which any risk would surface first.
No, and the authors do not suggest it. Their conclusion is that these might be dually effective options where depression or anxiety co-occurs with diabetes and obesity, and that randomized trials are needed. A randomized trial in depression has since missed its primary endpoint.

References

  1. 1.Taipale H, Taylor M, Lähteenvuo M, et al. Association between GLP-1 receptor agonist use and worsening mental illness in people with depression and anxiety in Sweden: a nationwide cohort study The Lancet Psychiatry. 2026. PMID: 41862258.

GLP-1s and Antipsychotic Weight Gain: A Trial Cut Short

Semaglutide produced 13.88% weight loss against 0.42% on placebo in people taking clozapine, without altering clozapine levels or psychotic symptoms. The trial ended at 31 of an intended 80 because the drug ran out.

6 min read

Antidepressants and a GLP-1: Ozempic, Wellbutrin, Lexapro and What the Labels Say

No GLP-1 label names an antidepressant as an interaction, and people on antidepressants lost as much weight on semaglutide as everyone else. The real overlaps: stacked nausea, slowed absorption of tablets, and two bupropion rules.

11 min read

Tirzepatide Versus Semaglutide on Mood

Across more than 85,000 matched pairs over two years, psychiatric outcomes were the same for both drugs. Against older GLP-1s, semaglutide was associated with about half the hazard of suicidal ideation.

5 min read

Semaglutide and Motivation

Semaglutide reduced how heavily people with depression discounted effort, while leaving their sensitivity to probability unchanged. That dissociation is what makes the finding hard to dismiss.

5 min read

When the Primary Endpoint Fails and a Secondary Doesn't

A randomized trial in 72 people with major depressive disorder found no effect on executive function, its primary outcome. A secondary cognition measure cleared p=0.03, and that is the one that will be reported.

5 min read

Metformin and a GLP-1 Together: The Risk Is Not the One People Expect

The diabetes trials behind these drugs were largely run on people already taking metformin. Stomach side effects did not stack in the largest analysis, and low blood sugar comes from other drugs. The real overlap is dehydration.

8 min read

Where to get semaglutide online, safely: sellers our editors have checked

These are telehealth sellers our editors have checked. For each one we hold a price, the form the drug comes in, and the states it reaches.

No insurance needed · vetted by our editors

Some of the links on this page earn us money. If you sign up with a provider after following one, that provider may pay GLP Watchdog a commission. Learn more

9.1

Collective

From $59/mo + membership

Get started →
Read full review →
6.4

MyDrHank

An oral route if you will not self-inject

About this analysis

GLP Watchdog publishes independently. Each seller in our dataset is checked against that seller's own website, and the dataset is updated continuously. Read our full methodology →