Published on 16 September 2026
GLP-1 receptor agonists and psychiatric symptoms: what a systematic review can and cannot show
In brief
GLP-1 receptor agonists are prescribed on a very large scale for type 2 diabetes and obesity, and the question of their psychiatric effects is now reaching the consulting room. This systematic review brings together 26 published studies, with a reported total of 3020 participants that counts some of them twice, and lists 10 registered trials. It calculates no pooled estimate: the authors cite the heterogeneity of designs, which range from randomised trials to case reports, and of populations. Where psychiatric symptoms were the primary outcome, in substance use disorders, the results are mixed and rather disappointing for exenatide. Where they were only a secondary outcome, most often in metabolic trials, mood and quality of life sometimes improve, with no way of separating a specific effect from the effect of weight loss or glycaemic control. On safety, the review did not analyse adverse events systematically. It reports isolated cases of depression, irritability and delusions on semaglutide, and in its discussion cites an analysis of the US FAERS database, outside the included studies, which reportedly found no disproportionate reporting of suicidal behaviour. The mean duration of the included studies is 26.4 weeks. The authors themselves describe the effects as mixed and inconclusive.
The context
A patient on semaglutide for diabetes says he is drinking less. Another asks whether the weight-loss treatment could make him depressed. Both questions are legitimate, and neither can be settled by a single trial. The class has already been the subject of two analyses on this site, one on GLP-1 receptor agonists and the risk of depression in older people with diabetes, the other on GLP-1 receptor agonists and alcohol consumption. This review has a broader ambition: to cover psychiatric symptoms as a whole (addictions, mood, anxiety, binge eating disorder, schizophrenia symptoms, quality of life), with neurodevelopmental and neurocognitive disorders excluded, as well as psychiatric safety.
Its main methodological idea is simple, and useful: to separate the studies in which the psychiatric symptom was what the treatment aimed at from those in which it was only an ancillary measure in a metabolic trial. In the first case, the study is in principle built to answer the question. In the second, it is not. The classification is not always applied rigorously: the secondary analysis by Probst et al. on alcohol consumption, drawn from a smoking cessation trial, appears among the studies with a psychiatric primary outcome.
In France, the marketed GLP-1 receptor agonists, semaglutide, liraglutide and dulaglutide, are authorised for type 2 diabetes or for weight management depending on the product, and none for a psychiatric disorder. Exenatide no longer holds an active authorisation there (French public medicines database and European Medicines Agency product information, consulted on 10 September 2026). Their use in a substance use disorder or a mood disorder therefore falls outside the authorised indications.
The study at a glance
| Question (PICO) | |
|---|---|
| Population | |
| People exposed to a GLP-1 receptor agonist: patients with a substance use disorder (cocaine, alcohol, nicotine), type 2 diabetes, obesity (including binge eating disorder and polycystic ovary syndrome), major depressive disorder or bipolar disorder, or schizophrenia with obesity or diabetes, with psychiatric symptoms measured. One study involves adolescents. 26 published studies, 3020 participants | |
| Interventions | |
| Exenatide, liraglutide, semaglutide (including an oral form), dulaglutide in the published studies; exenatide, liraglutide and semaglutide in the registered trials | |
| Comparators | |
| Variable across studies: placebo, another diabetes treatment (dulaglutide, insulin in particular), behavioural management alone, unexposed individuals, or a simple before and after comparison; no comparator in the case series and case reports | |
| Outcomes | |
| Psychiatric symptoms as primary outcome (substance use disorders) or secondary outcome (mood, anxiety, binge eating disorder, schizophrenia symptoms and cognition, quality of life). Adverse events are not among the data extracted from the published studies | |
| Design | |
| Systematic review following PRISMA, without meta-analysis. MEDLINE, PsycINFO and Embase searched through OVID from inception to November 2024, supplemented by the first ten pages of Google Scholar, the references of relevant articles and ClinicalTrials.gov (searched on 17 May 2024). Quality assessed with the JBI critical appraisal checklists. A body of evidence mixing randomised trials, observational studies, case series and case reports |
The quality check
| Criterion | Status |
|---|---|
| Literature search | Broad |
| FindingThree databases, the first ten pages of Google Scholar, the references of articles and a trial registry; selection and extraction carried out independently in duplicate | |
| Protocol registration | Not found |
| FindingNo registration (PROSPERO or other) and no protocol is mentioned; the attached PRISMA checklist marks both items “NA” | |
| Quantitative synthesis | Absent |
| FindingHeterogeneity of designs, populations and measures: descriptive synthesis only, no common effect size | |
| Publication bias | Not assessed |
| FindingA limitation acknowledged by the authors | |
| Funding and competing interests | Reservation |
| FindingNo specific funding declared. Competing interests declared by five of the fifteen authors, including fees from Novo Nordisk, Lundbeck and Janssen, among other companies, received by J. Swainson | |
The findings
| Domain | What the review reports |
|---|---|
| Cocaine use disorder | Exenatide 5 µg as a single dose (randomised trial, 13 participants): no significant difference from placebo on cocaine-induced euphoria (p = 0.21) or craving (p = 0.46). Case series of three patients on exenatide 2 mg weekly for 6 weeks: craving reduced in two patients, fluctuating in the third |
| PEB readingNot significant, small samples an absence of difference does not establish an absence of effect, and three uncontrolled cases do not establish an effect | |
| Alcohol use disorder | Exenatide 2 mg weekly for 26 weeks (randomised trial, 127 patients with alcohol use disorder): heavy drinking days clearly reduced in both groups, with no difference from placebo; DUDIT (+0.96 points; 95% CI 0.7 to 1.3) and AUDIT (+5.1 points at 6 months; 95% CI 0.9 to 9.3) scores worse on exenatide. Dulaglutide 1.5 mg weekly: alcohol consumption 29% lower than in the placebo group at 12 weeks (relative effect 0.71; 95% CI 0.52 to 0.97; p = 0.04), in a secondary analysis of a smoking cessation trial conducted in smokers, whether or not they drank. Case series of six patients on semaglutide: AUDIT down by 9.5 points on average |
| PEB readingConflicting results, non-comparable populations the only trial conducted in alcohol use disorder is negative; the dulaglutide signal comes from a secondary analysis in smokers, the semaglutide signal from six uncontrolled cases; no direct comparison between drugs | |
| Tobacco | Dulaglutide 1.5 mg weekly for 12 weeks, added to varenicline and behavioural support (randomised trial): abstinence 63% versus 65% on placebo at 12 weeks, 32% in both groups at 52 weeks; the abstract reports RR = 0.87 (p = 0.25) without specifying the time point. A registered trial of liraglutide reports 10.5% versus 9.5% abstinence at 12 weeks |
| PEB readingNo signal demonstrated | |
| Mood, anxiety and quality of life | Liraglutide 1.8 mg/day for 4 weeks in patients with major depressive disorder or bipolar disorder (open-label study without a control group, 19 participants according to the text, 17 according to table 1): improvement on the HAMD (Cohen’s d 0.68; p = 0.022) and the SHAPS (p = 0.010). Oral semaglutide compared with dulaglutide in 458 patients with diabetes over 52 weeks: better diabetes therapy-related quality of life (DTR-QoL); the abstract gives d = 0.48 for anxiety. Several studies found no difference: liraglutide (BDI, health-related quality of life, CES-D, depressive symptoms in binge eating disorder), exenatide versus active comparators (psychological well-being). A cross-sectional study associates exenatide with higher PHQ-9 and perceived stress scores |
| PEB readingDirect effect cannot be isolated secondary outcomes, often self-rated, most often in trials designed for weight or glycaemia; the only study conducted in depressed or bipolar patients is open-label and has no control group | |
| Binge eating disorder | Liraglutide 3.0 mg/day (randomised trial, 27 patients): binge episodes down by 4.0 per week versus 2.5 on placebo, a non-significant difference (p = 0.37), remission 44% versus 36%. Dulaglutide (pilot study, 60 patients with diabetes): reduction in BES score, linked to weight loss and to HbA1c. Semaglutide (retrospective cohort, 48 patients): greater reduction in BES than in the other treatment groups. Liraglutide combined with behavioural therapy (randomised trial, 150 patients): improvement in eating psychopathology at 24 weeks compared with therapy alone |
| PEB readingInconsistent signal, targeted trial not significant the only randomised trial conducted in binge eating disorder, with 27 patients, shows no difference; in the pilot study, the fall in score tracks weight loss | |
| Schizophrenia with obesity | Exenatide 2 mg weekly for 3 months (randomised trial, 45 patients): cognition and symptoms improved over time, with no group effect (p = 0.64) and no time by group interaction (p = 0.77) |
| PEB readingNo group effect demonstrated the improvement affects both arms and cannot be attributed to the treatment | |
| Psychiatric safety | Adverse events not systematically extracted from the published studies. Common effects cited in the discussion: early satiety, abdominal distension, nausea, fatigue, with references partly outside the included studies. Isolated cases on semaglutide: two depressive episodes (one in a patient with a history of depression), one case of irritability and anxiety, persecutory delusions in a patient with schizophrenia, which recurred when the dose was increased. Analysis of the FAERS database cited in the discussion, outside the body of included studies: no disproportionate reporting of suicidal behaviour |
| PEB readingNot systematically assessed, short follow-up an absence of analysis is not an absence of signal; isolated cases with no established causality, spontaneous reports, mean study duration 26.4 weeks; the reference cited for the FAERS analysis concerns, judging by its title, the neuroprotective effects of GLP-1 | |
The point that matters most for interpretation is the boundary between primary and secondary outcomes. The improvements in mood come mostly from studies in which diabetes or obesity was being treated; the only study conducted in depressed or bipolar patients was open-label and had no control group. Yet weight loss and better glycaemic control can improve mood in their own right, and that is one of the explanations the authors put forward. The review has no analysis that separates these two pathways, and it does not claim to.
Critical appraisal
| Domain | Judgement |
|---|---|
| Question and inclusion criteria | Clear, unevenly applied |
| FindingThe distinction between primary and secondary outcomes structures the whole synthesis, and it is the right one. It is unevenly applied: a secondary analysis of a smoking cessation trial is classed among the studies with a psychiatric primary outcome | |
| Level of evidence of included studies | Heterogeneous |
| FindingNine randomised trials according to the text (eleven studies rated with the JBI checklist for randomised trials in supplementary table S2) sit alongside observational studies, case series and case reports, which swell the body of evidence without adding comparative evidence. Three of these trials show a significant effect on the primary psychiatric outcome, six do not. The total of 3020 participants adds together the dulaglutide smoking cessation trial and its reanalysis on alcohol: some participants are counted twice | |
| Measurement of symptoms | Reservation |
| FindingVaried scales (BDI, HAMD, DUDIT, AUDIT, DTR-QoL), often self-rated, rarely the primary outcome | |
| Aggregate risk of bias | Not synthesised |
| FindingJBI checklist applied study by study (table S2), with no synthesis by criterion and no assessment of the certainty of evidence; the authors describe the randomised trials as of “moderate-to-high methodological quality” | |
| Quality of reporting | Internal inconsistencies |
| FindingThe abstract attributes to a case series (n = 12) a p value of 0.46 that the text reports for craving in the randomised exenatide trial; sample sizes differ between the text and table 1 (19 versus 17 participants for the liraglutide study in mood disorders; 225 in the table against 127 and 128 participants in the results of the dulaglutide trial); several statements in the discussion rely on references outside the included studies | |
| Fit between conclusion and evidence | Cautious, then optimistic |
| FindingThe authors conclude that the effects are mixed and inconclusive, but their discussion goes further than the data: they describe “a positive safety profile”, report “more consistent improvements” in the metabolic setting although the detailed results there are described as inconsistent, and present the class as “a promising integrative therapeutic target” | |
Level of evidence
PEB appraisal: reasonable confidence in the map, low confidence in each of the signals. What is established is the state of knowledge: few trials designed for a psychiatric question, results that conflict from one study and one drug to the next, and a psychiatric safety profile that has not been systematically assessed, with a few isolated cases and short follow-up. What is suggested, inconsistently, is a benefit on mood in patients treated for a metabolic reason, probably partly indirect. What still remains a hypothesis is a specific effect on reward circuits or on inflammation, which these studies did not measure.
The colleague test
What an experienced colleague would say if you put this study to them in two minutes, between two consultations.
“ I am not going to prescribe a GLP-1 agonist for an addiction or for depression, the data are mixed. My patients who are already taking one for their diabetes, on the other hand, I keep an eye on their mood, without turning it into an alarm. And when one of them is feeling better, the first thing I ask myself is whether they have lost weight. ”
What this means in practice: the class does not change psychiatric prescribing, it changes how you ask questions and how you interpret the answers in patients who are already receiving it.
What you can do with this
- Do not offer a GLP-1 receptor agonist to treat a substance use disorder or a mood disorder: the results are mixed and such use would fall outside the authorised indications.
- In a patient already treated for diabetes or obesity, ask about mood at each renewal, as for any recently started treatment, without presenting the class as causing depression: isolated cases of depression, irritability or delusions have been reported, without the link being established.
- When mood improves on treatment, work out what is due to weight, sleep, glycaemic control and a return to activity.
- Be ready to answer a patient who has read that a diabetes drug makes people stop drinking: signals exist (a secondary analysis of dulaglutide in smokers, a series of six cases on semaglutide), the only randomised trial conducted in alcohol use disorder, with exenatide, showed no difference from placebo, and no trial has yet settled the question.
Frequently asked questions
Do GLP-1 receptor agonists increase the risk of suicide?
This review cannot answer that. It did not analyse this risk in the included studies. In its discussion it cites an analysis of the US FAERS pharmacovigilance database that reportedly found no disproportionate reporting of suicidal behaviour, but the reference given does not correspond, judging by its title, to that work, and the same paragraph mentions studies associating semaglutide and liraglutide with suicidal ideation and self-injury. A spontaneous reporting database does not measure a risk, and the mean duration of the included studies is short (26.4 weeks): the question remains open, in either direction.
Can GLP-1 receptor agonists be used to treat alcohol use disorder?
Not on current evidence. The only randomised trial conducted in alcohol use disorder, with exenatide, shows no difference from placebo on heavy drinking days. The dulaglutide signal comes from a secondary analysis in smokers, the semaglutide signal from a series of six cases. Such use would fall outside the authorised indications.
Why might mood improve on a GLP-1 receptor agonist?
Two explanations coexist. One is indirect: weight loss and glycaemic control could improve mood, which the authors themselves put forward. The other, a direct effect on the brain, remains a hypothesis resting mainly on preclinical work, which these studies did not test.
Why did this systematic review not include a meta-analysis?
Because the studies differ in their designs, from randomised trials to case reports, in their populations and in what they measure. Pooling these results would have produced a figure that was precise and misleading.
Annotated bibliography
Source study. Meshkat S, Di Luciano C, Swiderski A, Li G, Janssen Aguilar R, Dunkley BT, Reichelt AC, Zhang Y, Greenshaw A, Vermetten E, Jetly R, Dash S, Agarwal SM, Swainson J, Bhat V. Efficacy and Safety of Glucagon-Like Peptide-1 Agonists for Psychiatric Symptoms: A Systematic Review. Brain and Behavior. 2025; 15(7): e70661. DOI: 10.1002/brb3.70661 · PMID 40635383, identifier confirmed in the PubMed record on 10 September 2026. S. Meshkat and C. Di Luciano are co-first authors. Received 14 January 2025, revised 16 May 2025, accepted 5 June 2025. Funding: no specific funding declared. Declared competing interests: J. Swainson, speaking or advisory fees from AbbVie, Bausch Health, Biron, Eisai, Idorsia, Janssen, Lundbeck, Novo Nordisk and Otsuka; R. Jetly, chief medical officer of Mydecine Innovation Group; B. T. Dunkley, chief scientific officer of MYndspan Ltd, funding from the Department of National Defence (Government of Canada), the Canadian Institutes of Health Research, the National Institutes of Health and MITACS; S. M. Agarwal, fees from HLS Therapeutics and Boehringer-Ingelheim Canada, supported by an Academic Scholar Award from the University of Toronto Department of Psychiatry; V. Bhat, Academic Scholar Award from the same department, research support from the Canadian Institutes of Health Research, the Brain & Behavior Foundation, the Ontario Ministry of Health Innovation Funds, the Royal College of Physicians and Surgeons of Canada, the Department of National Defence (Government of Canada), the New Frontiers in Research Fund, Associated Medical Services Inc. Healthcare, the American Foundation for Suicide Prevention, Roche Canada, Novartis and Eisai. Registration: no registration or protocol mentioned. Supplementary material consulted: search strategy (table S1), JBI quality assessment (table S2) and PRISMA checklist.