As of April 2026, more than 40 million GLP-1 receptor agonist prescriptions are dispensed annually in the United States alone, driven by blockbuster drugs like Ozempic (semaglutide) and Mounjaro (tirzepatide). The medications have transformed treatment for type 2 diabetes and obesity. But a string of peer-reviewed pharmacovigilance studies mining the FDA’s own adverse-event database has surfaced a troubling pattern: disproportionate reports of dementia and other cognitive problems among people taking GLP-1 drugs.
The signal is not settled science. Large observational studies point in the opposite direction, linking GLP-1 use to lower rates of Alzheimer’s disease and other dementias. That contradiction sits at the heart of an active scientific debate with real consequences for the tens of millions of patients and the clinicians advising them.
What the adverse-event data show
Several independent research teams have combed the FDA Adverse Event Reporting System (FAERS) for neurological signals tied to GLP-1 receptor agonists. A pharmacovigilance study by Duan et al. (2025) published in Therapeutic Advances in Neurological Disorders examined drug-induced dementia signals across a broad range of medications using FAERS data, establishing how dementia-related adverse events are detected and classified through the MedDRA coding system used by regulators worldwide.
A separate analysis by Wang et al. (2025), covering FAERS reports from 2005 through the third quarter of 2024, specifically assessed neurological adverse-event patterns for individual GLP-1 drugs, including exenatide, liraglutide, lixisenatide, dulaglutide, semaglutide, and tirzepatide. That study, indexed in PubMed, represents one of the most targeted efforts to quantify neurological signals for the drug class as a whole. A third FAERS-based study by Chen et al. (2024), published in European Psychiatry, analyzed neuropsychiatric adverse events associated with GLP-1 receptor agonists from the first quarter of 2010 through the first quarter of 2024, providing concrete event counts and trend data over that 14-year window.
None of these studies prove that GLP-1 drugs cause cognitive decline. FAERS data reflect voluntary reports, not controlled experiments, and they can be skewed by media attention, prescribing volume, and reporting habits. But the consistency of the signal across multiple independent teams and overlapping time periods gives it more weight than any single analysis would carry on its own.
The case for neuroprotection
Running against those safety flags is a large propensity-matched cohort study published in International Immunopharmacology. Researchers matched obese adults who used GLP-1 receptor agonists with similar non-users across multiple healthcare organizations and countries. The GLP-1 group showed reduced risks for Alzheimer’s disease, Lewy body dementia, and vascular dementia, along with lower overall mortality.
Propensity matching attempts to control for confounders like age, sex, comorbidities, and other medications, but it cannot eliminate them entirely. People who stay on GLP-1 therapy long enough to be counted as exposed may differ from those who stop early or never start in ways that independently affect dementia risk: better access to specialty care, more aggressive cardiovascular management, or higher health literacy. Still, the study offers a different and arguably stronger lens than FAERS signal detection, and it suggests the drugs could be neuroprotective in metabolically unhealthy populations.
What clinical trials have and have not shown
No prospective randomized controlled trial has yet measured long-term cognitive outcomes in cognitively healthy adults taking GLP-1 drugs for weight loss. The trial evidence that does exist is limited and mixed.
A proof-of-concept trial evaluating long-acting exenatide in patients with mild cognitive impairment found the drug did not prevent cognitive decline. The phase 2b ELAD trial (Edison et al., 2024, published in Alzheimer’s & Dementia) tested liraglutide in 204 participants with mild to moderate Alzheimer’s disease who did not have diabetes, using brain metabolism as the primary endpoint and cognitive measures as secondary outcomes. Neither trial demonstrated clear cognitive benefit from GLP-1 treatment in patients who already had significant neurodegeneration.
That does not close the book. A failure to reverse or slow decline in people with established Alzheimer’s pathology says little about whether these drugs might protect the brain if given earlier, before damage accumulates. Notably, Novo Nordisk’s ongoing EVOKE and EVOKE Plus trials are testing oral semaglutide in more than 3,700 people with early-stage Alzheimer’s disease, with results expected in the coming years. Those trials could reshape the conversation significantly.
Meanwhile, a published case report described severe neuropsychiatric symptoms after semaglutide was started in an older adult with both dementia and type 2 diabetes. A single case cannot establish causation, but it illustrates how patients with existing cognitive vulnerability may experience central nervous system effects that healthier users do not.
A retraction and a regulatory gap
The evidence base has its own quality problems. A 2024 systematic review and meta-analysis examining GLP-1 receptor agonists and cognitive function in type 2 diabetes patients (Yao et al.) was retracted by the journal Diabetes, Obesity and Metabolism after failing to meet peer-review standards. The retraction, recorded in PubMed, is a reminder that synthesis-level research in this area has already been withdrawn at least once. Clinicians and patients relying on aggregated findings should verify whether the underlying studies hold up individually.
The FDA has not issued a safety communication specifically addressing cognitive impairment signals from FAERS pharmacovigilance studies of GLP-1 drugs. The agency’s most directly relevant recent action involved a different neuropsychiatric concern: it requested the removal of suicidal behavior and ideation warnings from the labeling of GLP-1 medications including Saxenda, Wegovy, and Zepbound. That decision followed a multi-pronged evaluation using clinical-trial meta-analysis, a Sentinel claims study, and a review of observational studies and case reports.
The process shows how the FDA evaluates postmarketing safety signals, but the cognitive impairment question has not yet received comparable formal scrutiny. Regulatory silence should not be mistaken for a clean bill of health; it may simply reflect limited data, competing priorities, or a judgment that current evidence does not meet the threshold for label changes.
How to read the conflicting evidence
Understanding why these findings seem to contradict each other requires recognizing what each type of study can and cannot do.
FAERS-based pharmacovigilance studies are built to detect signals, not confirm causation. A spike in adverse-event reports for semaglutide partly reflects the fact that millions more people are taking it compared with five years ago. Reporting biases, media coverage, and the “stimulated reporting” effect that follows high-profile news stories can all inflate numbers.
Observational cohort studies, even well-matched ones, cannot fully account for the healthy-user bias that may make GLP-1 patients look better on paper. And randomized trials in this space have been small, short, and focused on people who already have dementia, a population where reversing damage is far harder than preventing it.
“Cognitive risk” itself is not a single outcome. FAERS reports encompass everything from transient confusion and memory lapses to formally diagnosed dementia. Some of those events may stem from hypoglycemia, dehydration, medication interactions, or progression of an underlying disease rather than a direct effect of the GLP-1 drug. On the other side, the dementia diagnoses captured in large cohort studies may miss subtler deficits in attention, executive function, or mood that still matter to patients day to day.
Monitoring guidance for patients starting GLP-1 therapy
The most defensible position as of May 2026 is that GLP-1 receptor agonists remain highly effective for blood sugar control and weight reduction, with possible brain benefits in high-risk metabolic populations and unresolved questions about cognitive safety in vulnerable groups.
Patients with pre-existing dementia, significant psychiatric illness, or a history of unexplained cognitive changes may warrant closer monitoring when a GLP-1 drug is started or the dose is increased. Family members and caregivers can help by watching for new confusion, personality shifts, or functional decline and reporting these promptly to a clinician.
Clinicians should document baseline cognitive status when feasible, especially in older adults, and revisit it periodically during treatment. When neuropsychiatric symptoms arise, a structured assessment is preferable to assuming they are caused by the drug or dismissing them as unrelated. Temporarily reducing the dose or pausing the medication may help clarify whether there is a temporal relationship.
Given the conflicting data, shared decision-making matters more than usual. Patients deserve to understand both the substantial cardiometabolic benefits and the current uncertainties about long-term brain effects. Until large, long-duration randomized trials with standardized cognitive endpoints, stratified by baseline brain health and age, deliver results, the story of GLP-1 drugs and cognition will remain unfinished. Careful, individualized conversations between patients and their doctors are the safest path through the uncertainty.
More from Morning Overview
*This article was researched with the help of AI, with human editors creating the final content.
