Medicaid cohort study compares GLP-1 use with new depression diagnoses

A study published in September 2026 compared baseline GLP-1 receptor agonist use with newly documented depression diagnoses among adults with type 2 diabetes enrolled in Medicaid across multiple states ^[1]. The cohort study appears in Healthcare, an MDPI journal ^[1].

The comparison covered five drugs in the class: dulaglutide (Trulicity), exenatide (Byetta and Bydureon), liraglutide (Victoza and Saxenda), semaglutide (Ozempic, Wegovy and Rybelsus) and tirzepatide (Mounjaro and Zepbound) ^[1]. Researchers looked at whether patients were using one of these medicines at baseline and whether a new depression diagnosis was later documented in their claims ^[1].

The work follows the FDA's 2026 drug safety communication, which asked manufacturers to remove the suicidal behavior and ideation warning from the GLP-1 class ^[1].

What the study measured

The outcome was a newly documented depression diagnosis in insurance claims, not a clinical assessment of mood ^[1]. The paper notes that claims-based mental health outcomes are sensitive to diagnostic coding and to patients' access to care ^[1]. In practice, that means a new code could reflect a first episode of depression, the first time a clinician wrote the diagnosis down, or a screening or follow-up visit where it was entered. It could also miss depression that was never coded because a patient did not see a provider or was not screened.

Several details of the analysis have not been disclosed, including how many patients were included, how long they were followed, and how large any difference in new depression diagnoses was between GLP-1 users and non-users ^[1]. Whether the study separated patients by individual drug, by dose, or by whether they had a prior history of depression has not been reported ^[1].

Why it matters for patients

People taking or considering semaglutide, tirzepatide or other GLP-1 medicines often ask whether the drugs affect mood. This study compares patterns of recorded diagnoses across a Medicaid population; it does not measure what happened to any one person, and a cohort design of this kind cannot show that a drug caused or prevented depression ^[1].

Because the data come from Medicaid enrollees, the results may not describe people with commercial insurance, Medicare or no coverage at all ^[1]. And because the measure depends on whether a diagnosis was written down and coded, groups with less access to mental health care may appear to have less depression documented — a difference in paperwork and access rather than in how patients felt ^[1].

Nothing in the paper changes the approved labeling for any of the five drugs on its own, and no new prescribing guidance tied to this study has been announced ^[1].

What happens next

Under the FDA's 2026 communication, manufacturers were asked to remove the suicidal behavior and ideation warning from GLP-1 labeling ^[1]. The timing of those labeling changes has not been disclosed ^[1].

Whether additional studies will follow, or whether this cohort will be extended or reanalyzed with the full set of results, has not been reported ^[1]. For now, the paper adds population-level data to a question that regulators, researchers and patients have been weighing since the warning was first added and then targeted for removal ^[1].