In a US observational cohort of 70,257 adults with schizophrenia and type 2 diabetes, starting a GLP-1 receptor agonist was associated with lower hazards of psychiatric hospitalization and psychiatric emergency-department visits. These findings are encouraging, but they do not establish that the medications caused the difference.
The population is large enough to matter. Roughly one in eight adults with schizophrenia also carries a diagnosis of type 2 diabetes, a 13.2% prevalence drawn from a meta-analysis of thirteen studies, and that overlap surprises nobody who treats this group. For context on the stakes, excess all-cause inpatient and emergency-department costs associated with schizophrenia have been estimated at $30.2 billion and $5.6 billion respectively in 2019 dollars. Those are all-cause figures rather than psychiatric care alone, and they are not an estimate of what GLP-1 treatment might save.
The cohort drew on electronic health records from January 2018 onward, with a mean age of 53.1 years. GLP-1 initiators showed lower psychiatric acute-care use not only against non-initiators but against patients starting DPP-4 inhibitors or SGLT2 inhibitors, both of which carry some evidence of mood-related benefit of their own. A propensity-score matched cohort from Hong Kong found SGLT2 inhibitors associated with lower new-onset depression risk compared with DPP-4 inhibitors, which is why a GLP-1 advantage over both classes is worth examining rather than dismissing as a class effect.
What the Randomized Trials Show
Randomized evidence in this population is metabolic, not psychiatric. In COaST (Siskind et al., 2025), 31 clozapine-treated participants given semaglutide had substantial weight loss over 36 weeks, with no detected difference in PANSS scores. In HISTORI (2025), 154 participants with schizophrenia, prediabetes and overweight or obesity saw improvements in HbA1c and weight over 30 weeks, without significant improvement in PANSS-6 scores or mental quality of life. Sass et al. (2026) followed 73 participants with schizophrenia-spectrum disorders taking clozapine or olanzapine and found semaglutide improved HbA1c and weight over 26 weeks, again with no detected difference in psychiatric symptoms. Earlier, Larsen et al. (2017) found liraglutide improved glucose tolerance and produced 5.3 kg greater weight loss than placebo over 16 weeks in 103 participants taking clozapine or olanzapine.
Taken together, randomized trials support metabolic benefits of GLP-1 therapy in selected schizophrenia populations. Whether those benefits translate into fewer psychiatric hospitalizations remains an open question.
Possible Mechanisms
Several lines of work point toward the target without settling it. A Mendelian randomization study reported that genetically proxied GLP1R expression was associated with lower schizophrenia risk, with an odds ratio of 0.84 (95% CI: 0.71 to 0.98). That concerns genetically proxied receptor expression rather than actual medication treatment, so it supports further investigation of the target rather than showing that GLP-1 medications prevent or treat schizophrenia.
GLP-1 receptors are expressed in limbic and prefrontal regions, which raises the possibility of central effects including modulation of neuroinflammation. The mechanistic differences between GLP-1 agonists and DPP-4 inhibitors at the CNS level remain an active area of investigation. Whether any central effect translates into clinically meaningful stabilization in people already taking antipsychotics is precisely what observational data cannot confirm.
A separate line concerns the metabolic injury itself. Work in isolated human islets suggests that antipsychotic-related dopamine-receptor blockade can alter insulin and glucagon secretion. Whether that progresses to diabetes in patients, and whether correcting it yields any direct psychiatric benefit, should be treated as hypotheses rather than established clinical pathways.
There is also a simpler possibility worth holding alongside the others. Metabolic destabilization in schizophrenia is not a quiet comorbidity sitting beside the psychiatric illness. Uncontrolled hyperglycemia is associated with worse cognitive function, disrupted sleep and higher inflammatory burden, each of which can erode whatever stability a patient has achieved. Some part of any apparent psychiatric benefit may reflect the removal of a chronic destabilizer rather than a central drug effect. Distinguishing the two is the work ahead.
Why Confounding Is the Central Problem
The study authors are appropriately cautious, and residual confounding and selection bias are real concerns in any EHR-based cohort. Patients who receive a GLP-1 prescription in a schizophrenia population may have more engaged care teams, better adherence support or more stable social circumstances, and every one of those predicts fewer hospitalizations independent of the drug. That caution is warranted, and it describes exactly the kind of confounding only a randomized trial can resolve.
What Future Trials Need to Test
The randomized trials conducted so far were designed and powered around metabolic endpoints, which is why they can report weight and HbA1c with confidence and say little about hospitalization. Future trials should prespecify psychiatric outcomes and include enough participants and follow-up to evaluate them reliably. That is a question of adequate design rather than a requirement that psychiatric measures become co-primary endpoints in every study.
The population is identifiable in EHR systems, as a cohort of 70,257 demonstrates. Retention is the harder problem, since this group has high dropout in real-world care, and any design that assumes standard psychiatric trial retention will be underpowered before the first interim look. The open question is whether a sponsor with a schizophrenia and diabetes population in its registry data moves to prespecify a prospective sub-study, and whether regulators treat psychiatric outcomes in metabolic trials as a signal worth powering for.
References
Lancet Psychiatry, “GLP-1 receptor agonist initiation and risk of psychiatric hospitalisation and emergency department presentation in adults with schizophrenia and type 2 diabetes in the USA”
Siskind et al. (COaST, 2025), semaglutide in clozapine-treated participants
HISTORI (2025), JAMA Psychiatry, semaglutide in schizophrenia with prediabetes and overweight or obesity
Sass et al. (2026), JAMA Psychiatry, semaglutide in schizophrenia-spectrum disorders on clozapine or olanzapine
Larsen et al. (2017), liraglutide in patients taking clozapine or olanzapine
Dove Press, “Diabetes Mellitus in Patients with Schizophrenia: A Systematic Review and Meta-Analysis” (Kake and Jalilian)
Journal of Clinical Psychiatry, “Economic Burden of Schizophrenia in the United States” (Kadakia et al.)
Mendelian randomization study of genetically proxied GLP1R expression and schizophrenia risk (OR 0.84, 95% CI: 0.71 to 0.98)
Study of dopamine-receptor blockade and insulin and glucagon secretion in isolated human islets
ResearchGate, “Comparing SGLT2 inhibitors and DPP-4 inhibitors on new-onset depression: a propensity score-matched study in Hong Kong”
Dr. Leonardo Vando, MD is Principal Investigator of Psychiatry at Vando Medical Services.
