Doctors assume weight-loss drugs help mental health by helping people lose weight. A study of 63,000 patients found the dose they reached mattered more than the pounds they lost.
If weight-loss drugs help mental health, the obvious explanation is the weight loss itself. A study of 63,215 patients on semaglutide, the drug in Ozempic and Wegovy, tested that idea by comparing two things: how much weight patients lost and the highest dose they reached. Patients who reached higher doses went on to have fewer new mood, anxiety, and substance-use diagnoses. For most of those conditions, patients who lost more weight did not. The authors then explain why that result may not mean what it seems.
Researchers at the health data company nference analyzed electronic health records from a network of US academic medical centers that included 489,785 patients prescribed semaglutide. The main analysis used the 63,215 of them who already had a neurological or psychiatric condition before starting the drug, so the findings do not describe people with no such history. Follow-up ran through the end of 2025.
In a first comparison, the team matched semaglutide patients one to one with people starting metformin, a common diabetes drug. Over two years, semaglutide patients had lower rates of mood disorders (10.5% versus 13.2%), substance-related disorders (3.8% versus 6.1%), psychotic disorders (0.31% versus 0.93%), and dementia or degenerative brain disease (0.95% versus 1.96%). Metformin did better on neuromuscular disease. Comparisons with two other diabetes drug classes showed mixed but broadly similar patterns.
Dose tracked with diagnoses, weight loss mostly did not
The second analysis stayed inside the semaglutide group. The researchers recorded the highest dose each patient reached and the most weight each patient lost during the first two years, then counted new diagnoses over the next two. They compared 39,795 patients who never went above 1.0 mg with 15,515 who reached 1.7 mg or more. The higher-dose group had lower rates of substance-related disorders (relative risk 0.71), mood disorders (0.82), and anxiety and stress-related disorders (0.88).
When they sorted the same patients by weight lost instead, from under 5% to 20% or more, the rates of mood disorders, substance-related disorders, and degenerative brain conditions did not separate significantly. Two outcomes went the other way. Cognitive symptoms and speech or language symptoms tracked with weight loss, and cognitive symptoms were more common in the bigger weight-loss groups: 5.5% among patients who lost 15% to 20% of their weight versus 2.1% among those who lost under 5%. The authors say this should not be read as semaglutide harming cognition. They point to frailty, illness-related weight loss, and reverse causation. Dementia rates did not differ between the high-dose and low-dose groups.
Why the authors urge caution
The paper is blunt about what the dose result does not show. Maximum dose is not a clean pharmacological variable, the authors write, because it also reflects how well patients tolerated the drug, whether they kept taking it, whether they could access care, and their insurance. The low-dose group probably includes people who never really started semaglutide or quit during dose escalation, while the high-dose group is likely healthier and more engaged with their doctors. This is called healthy-adherer bias, and the authors say the dose associations “should not be interpreted as a dose-response signal in the pharmacological sense.” They also note that the null weight-loss result may simply mean dose captures more of that bias.
Other limits apply. The data are retrospective, and adherence is not well captured in the records. Diagnoses are defined by the first diagnostic code in the record, which reflects when something was documented rather than when it began. With cohorts this large, tiny differences can reach statistical significance. The study did not assess side effects, and the authors say it does not rule out acute psychiatric events or support off-label prescribing. Other evidence is mixed: a separate analysis of more than 300,000 patients with obesity and no psychiatric history found GLP-1 drug use linked to more depression, anxiety, and suicidal behavior.
The authors call their findings hypothesis-generating and say prospective trials are needed. They are employees of nference, which runs research collaborations with drug companies whose products appear in the study. The paper states that none of those companies funded the work or had a role in it.
Source
Murugadoss K, Venkatakrishnan AJ, Soundararajan V. “Higher semaglutide dose is associated with lower neuropsychiatric event incidence independent of weight loss.” npj Metabolic Health and Disease, 2026;4:36.
DOI: 10.1038/s44324-026-00131-3
