Swedish Register Study Links Semaglutide and Liraglutide to Lower Risk of Worsening Depression and Anxiety

A within-individual cohort analysis of nearly 100,000 patients with co-occurring psychiatric and metabolic conditions suggests GLP-1 receptor agonists may offer psychiatric benefit, but causal proof requires randomised trials and findings should not be generalised beyond the studied population.

Why This Matters

Depression and anxiety disorders are among the most common comorbidities in people living with type 2 diabetes, affecting an estimated one in four patients and substantially worsening glycaemic control, treatment adherence, functional outcomes, and mortality risk. The bidirectional relationship between metabolic disease and psychiatric illness creates a therapeutic challenge: medications that address one domain may inadvertently worsen or improve the other, and clinicians managing these overlapping conditions need evidence about the neuropsychiatric profiles of the drugs they prescribe. GLP-1 receptor agonists, originally developed for glucose regulation, have demonstrated effects on central nervous system pathways including neuroinflammation, reward circuitry, and hypothalamic-pituitary-adrenal axis signaling, raising legitimate scientific questions about whether their benefits extend beyond metabolic control. With semaglutide and liraglutide prescribing expanding at historic rates, driven by both diabetes management and obesity treatment, determining whether these drugs carry psychiatric risk or benefit has become one of the most urgent pharmacovigilance and clinical questions in contemporary medicine.

Clinical Summary

The intersection of type 2 diabetes and mood or anxiety disorders represents one of the most clinically challenging comorbidity clusters in primary and specialty care. Patients with both conditions experience worse outcomes across nearly every relevant domain: poorer glycaemic control, higher cardiovascular event rates, greater disability, and elevated suicide risk compared with either condition alone. Existing treatments for depression and anxiety in this population are complicated by metabolic side effects of many psychotropic medications, limited efficacy of standard antidepressants in the context of chronic systemic inflammation, and the cumulative burden of polypharmacy. A therapeutic agent that could simultaneously address metabolic dysfunction and reduce psychiatric morbidity would represent a significant advance, but the evidentiary bar for such a claim is necessarily high.

Published in Lancet Psychiatry in 2026, this study by Fazel and colleagues drew on comprehensive Swedish national health registers to examine whether GLP-1 receptor agonist use was associated with reduced psychiatric deterioration in people with pre-existing depression or anxiety who were also using antidiabetic medications. The mechanistic rationale for a potential neuropsychiatric benefit is grounded in several converging lines of preclinical evidence. GLP-1 receptors are expressed in brain regions critical to mood regulation, including the hippocampus, amygdala, and prefrontal cortex. Preclinical studies have demonstrated that GLP-1 receptor activation can reduce neuroinflammation, enhance neuroplasticity, and modulate dopaminergic reward pathways. Beyond direct central nervous system effects, improvements in glycaemic variability, insulin resistance, and body weight could indirectly improve mood through reduced inflammation, improved sleep, and enhanced self-efficacy. However, teasing apart direct neuropsychiatric effects from indirect metabolic benefits is precisely the challenge that observational data, even well-designed observational data, cannot fully resolve.

The study examined 95,490 individuals identified through Swedish registers between 2009 and 2022, of whom 22,480 used a GLP-1 receptor agonist during the study period. The primary outcome was a composite measure of worsening mental illness defined as psychiatric hospitalisation, prolonged psychiatric sick leave exceeding 14 days, hospitalisation for self-harm, or death by suicide. Using a within-individual stratified Cox regression design, each patient’s periods of GLP-1 receptor agonist use were compared with their own periods of non-use, substantially reducing confounding from stable between-person characteristics. Semaglutide showed the strongest association with reduced risk, with an adjusted hazard ratio of 0.58 for the composite outcome, and specific reductions for worsening depression (aHR 0.56), worsening anxiety (aHR 0.62), and worsening substance use disorder (aHR 0.53). Liraglutide showed a smaller but statistically significant association (aHR 0.82). Notably, exenatide and dulaglutide showed no significant benefit whatsoever, with adjusted hazard ratios of 1.01 for both agents, suggesting this is not a uniform GLP-1 receptor agonist class effect and raising important questions about receptor binding kinetics, brain penetrance, and dosing differences across these molecules.

The within-individual design is a genuine methodological strength, eliminating confounding from genetics, stable personality traits, socioeconomic background, and other fixed characteristics that plague conventional cohort comparisons. However, the design cannot address time-varying confounders that coincide with GLP-1 receptor agonist initiation: concurrent changes in psychotropic medications, lifestyle modifications prompted by a new prescription, intensified clinical monitoring, weight loss itself, or natural fluctuations in illness trajectory. The composite outcome, while pragmatically useful for capturing severe psychiatric deterioration, combines events of very different clinical significance and may obscure heterogeneous effects across outcome types. The population is restricted to individuals with both a psychiatric diagnosis and antidiabetic medication use in Sweden, and findings should not be extended to GLP-1 receptor agonist use for obesity in psychiatrically healthy populations or to healthcare systems with different prescribing patterns. The authors appropriately conclude that randomised controlled trials are the necessary next step to determine whether these associations reflect a causal neuropsychiatric benefit.

Dr. Caplan’s Take

What this study gets right is the analytical architecture. The within-individual design is not just a statistical convenience; it directly addresses the most damaging criticism of pharmacoepidemiological studies, which is that the people who receive a given drug differ systematically from those who do not. By comparing each patient to themselves across time periods of GLP-1 receptor agonist use and non-use, the investigators substantially reduce the confounding that makes so much observational research unreliable. The differential findings across specific agents also add biological plausibility. If this were simply a healthy-user effect or a marker of better overall medical care, we would expect similar associations across all GLP-1 receptor agonists, not a strong signal for semaglutide, a moderate signal for liraglutide, and no signal at all for exenatide and dulaglutide.

In my clinical practice, I encounter this question with increasing frequency. Patients with depression or anxiety who are also managing diabetes or obesity ask whether semaglutide might help their mood, often citing media coverage or personal anecdotes. What they are really asking is whether there is a single intervention that can address both the metabolic burden and the emotional weight they carry. That is a deeply understandable hope. But an honest clinical response requires acknowledging that we do not yet know whether semaglutide directly improves psychiatric outcomes, whether the improvement seen in this study comes from weight loss, better glycaemic control, reduced inflammation, enhanced self-image, or some combination of all of these. Telling a patient “the data look promising but we cannot yet recommend this for your depression” is unsatisfying, but it is the only defensible position.

The specific methodological gaps that prevent confident recommendation are important to name. Time-varying confounders remain the central concern: when a patient starts semaglutide, that event may coincide with a period of heightened clinical engagement, medication review, dietary change, or a decision point in their psychiatric illness trajectory that independently predicts improvement. The study could not capture changes in psychotherapy, exercise, or social circumstances that might cluster around the initiation of a new metabolic medication. Additionally, the composite outcome, while practical, mixes psychiatric hospitalisation with prolonged sick leave, events that may have quite different clinical and mechanistic drivers. We need randomised trial data that isolates the psychiatric signal from the metabolic noise before we can tell patients that semaglutide treats depression.

What I actually do, given the current evidence, is straightforward. For patients already indicated for a GLP-1 receptor agonist based on their metabolic profile, I discuss the emerging observational data as an additional consideration that may inform agent selection, particularly if choosing between semaglutide and a molecule with no psychiatric signal. I document the psychiatric comorbidity, establish baseline symptom severity using validated scales, and monitor mood prospectively after initiation. I do not prescribe GLP-1 receptor agonists for a psychiatric indication, and I am explicit with patients about why. This population-level register signal, however promising, does not replace the need for rigorous trials. If patients are interested in contributing to the evidence base, I direct them toward clinical trial registries where randomised studies of GLP-1 receptor agonists for depression are beginning to recruit.

Clinical Perspective

This study occupies a pivotal position in the rapidly evolving landscape of GLP-1 receptor agonist research. It arrives at a moment when pharmacovigilance agencies, clinicians, and patients are simultaneously asking whether the neuropsychiatric effects of these drugs are beneficial, neutral, or harmful. Early post-marketing signals raised concerns about suicidal ideation with semaglutide, prompting regulatory review in multiple jurisdictions. The Fazel et al. findings push back against those concerns by suggesting the opposite direction of effect in people with pre-existing mood and anxiety disorders, though it is critical to note that these are different questions applied to different populations. The differential signal across specific agents is the study’s most scientifically provocative finding, because it implies that receptor affinity, brain penetrance, or pharmacokinetic profile may determine neuropsychiatric impact, not merely GLP-1 pathway activation itself.

For patient-facing communication, clinicians should be prepared to acknowledge the study without overstating its implications. The appropriate framing is that a large, well-designed observational study suggests semaglutide and liraglutide may be associated with psychiatric benefit in people who have both diabetes and depression or anxiety, but that observational evidence, no matter how carefully analysed, cannot prove causation. Patients who have seen media coverage of this study may arrive with the impression that semaglutide is a treatment for depression. It is not, based on current evidence. Clinicians can validate the patient’s interest while being clear that the mechanism driving the association remains unknown, and that prescribing a GLP-1 receptor agonist solely for psychiatric benefit would be premature and unsupported by any clinical guideline.

Specific pharmacological considerations deserve attention. The differential effect across GLP-1 receptor agonists raises the question of whether semaglutide’s greater CNS penetrance relative to exenatide, for example, accounts for the psychiatric signal. Clinicians managing patients on both GLP-1 receptor agonists and psychotropic medications should be aware that significant weight loss can alter the pharmacokinetics of lipophilic drugs, including many antidepressants and anxiolytics, potentially requiring dose adjustment. Nausea, the most common GLP-1 receptor agonist side effect, can mimic or exacerbate anxiety symptoms and reduce adherence to both metabolic and psychiatric treatment regimens. Additionally, rapid weight loss in individuals with eating disorder histories or body image disturbance requires careful psychiatric monitoring regardless of the metabolic indication.

The most actionable recommendation for clinicians right now is to integrate structured psychiatric monitoring into the care of patients who carry both metabolic and mood disorder diagnoses and are initiating or continuing GLP-1 receptor agonist therapy. This means documenting baseline depression and anxiety severity using validated instruments such as the PHQ-9 and GAD-7, reassessing at regular intervals after GLP-1 receptor agonist initiation, and recording both metabolic and psychiatric outcomes in a way that contributes to real-world evidence generation. If a patient’s psychiatric status appears to improve on semaglutide or liraglutide, document the timeline carefully but do not discontinue evidence-based psychiatric treatment on the basis of observational data alone. Where available, consider referring patients to ongoing randomised trials examining GLP-1 receptor agonists for major depressive disorder, as these trials represent the only pathway to the causal evidence that clinical practice requires.

Study at a Glance

Study Type

National register-based cohort study with within-individual (self-controlled) stratified Cox regression

Country

Sweden

Study Period

2009 to 2022