Clinical Insight | CED Clinic

De Giorgi and colleagues studied 13,007 adults with psychiatric diagnoses in a multicenter US electronic health record dataset. Among 1,261 semaglutide initiators, clinician-recorded cognitive signs over 12 months were lower than with no antidiabetic treatment, glipizide, or empagliflozin, but not significantly lower than with sitagliptin after correction. This is a clinically relevant association, not evidence that semaglutide should be prescribed to improve cognition. Residual confounding, treatment selection, missing data, and routine documentation can all shape the estimate. Randomized trials using standardized cognitive and functional outcomes are needed before efficacy or clinical utility can be claimed.

What This Study Teaches Us

The adjusted 12-month cognitive score was lower among semaglutide initiators than among people receiving no antidiabetic treatment, glipizide, or empagliflozin. The comparison with sitagliptin was not statistically significant after correction. Associations were most consistent in major depression, while the study did not demonstrate a causal treatment effect.

Why This Matters

Cognitive symptoms are common across major depression, psychosis, and bipolar disorder, yet treatment evidence is limited. A large routine-care dataset can identify a clinically relevant association, but prescribing decisions require randomized evidence, validated cognitive testing, and a complete assessment of metabolic and psychiatric risks.

Study Snapshot

Clinical Bottom Line

Semaglutide initiation was associated with fewer recorded cognitive signs than three of four comparison strategies. This is hypothesis-supporting real-world evidence, not proof that semaglutide improves cognition or treats psychiatric illness.

Three Comparisons Favored Semaglutide

Adjusted cognitive-sign scores were lower with semaglutide than with no treatment, glipizide, or empagliflozin.

The estimates describe associations in routine clinical records and should not be read as randomized treatment effects.

The Sitagliptin Comparison Was Null

The mean ratio versus sitagliptin was 0.89 with a 95% confidence interval from 0.77 to 1.02.

That comparison was not statistically significant after Bonferroni correction, so the paper did not show a consistent advantage over every active comparator.

Major Depression Had the Most Consistent Signal

Effect sizes were similar across diagnostic groups, but findings were only consistently significant in major depression.

Subgroup consistency does not remove confounding or establish that semaglutide treats depression-related cognitive dysfunction.

Routine Ratings Are Not a Neuropsychological Battery

The outcome came from semistructured clinician-rated assessments documented during usual care.

This can capture real-world signs, but it differs from standardized, blinded cognitive testing and may be influenced by visit patterns and documentation practices.

Disclosures Matter

Two authors reported employment and equity ownership in the company group connected with the NeuroBlu data platform, and one author reported consultancy fees from Holmusk.

A disclosure does not invalidate the analysis, but independent replication is important when data access and commercial relationships intersect.

What This Paper Does Not Show

The study does not show that semaglutide prevents dementia, reverses cognitive impairment, treats depression, psychosis, or bipolar disorder, improves daily functioning, or is superior to sitagliptin. It also does not establish causality, a cognitive prescribing indication, or a complete safety profile.

How This Fits With the Broader Clinical Conversation

Observational GLP-1 research has produced signals across neurologic and psychiatric outcomes, but estimates vary with the population, comparator, indication, and outcome definition.

For clinical care, cognition should be assessed alongside psychiatric symptoms, metabolic status, medication burden, sleep, substance exposure, nutrition, and functional change rather than inferred from one EHR association.

Dr. Caplan’s Take

I find the size and active-comparator design useful, especially because the signal was not identical across all comparisons. The null sitagliptin result is part of the finding, not a detail to hide.

This paper supports a better randomized question. It does not support prescribing semaglutide as a cognitive enhancer or changing a psychiatric treatment plan on the strength of routine-care associations.

What a Careful Reader Should Take Away

A careful reader should retain both halves of the result: fewer recorded cognitive signs appeared with semaglutide versus three comparison strategies, while one active comparison was null and the nonrandomized design cannot establish treatment efficacy.

Evidence Interpretation Guide

A Four-Step Reading Frame