GLP-1 Drugs and Dementia Risk: What Observational Studies Cannot Prove
Several large observational studies link GLP-1 drugs to lower dementia risk in people with type 2 diabetes, but observational data cannot show the drugs caused it. Two large randomized trials, EVOKE and EVOKE+, found oral semaglutide did not slow progression in people who already had early Alzheimer’s symptoms.
Key points
- Observational studies compare people who happened to take different drugs. They can reveal a pattern but cannot rule out other explanations.
- Reported dementia reductions differ between studies because the comparison drugs, populations and drug classes differ.
- The best direct test so far, two phase 3 trials of semaglutide in early Alzheimer’s disease, did not show slower decline.
- Those trials tested treatment after symptoms began. They did not test prevention in people without dementia.
What the observational studies found
| Study | Design | Comparison | Reported finding | Main limit |
|---|---|---|---|---|
| Tang et al., JAMA Neurology, 2025 | Target trial emulation using real-world health data on 396,963 eligible adults with type 2 diabetes | GLP-1 receptor agonists or SGLT2 inhibitors vs other second-line glucose-lowering drugs | Lower risk of Alzheimer’s disease and related dementias for both classes (hazard ratios 0.67 and 0.57). No significant difference between the two classes. | Residual confounding and reliance on clinical coding to identify dementia |
| Wang, Renoux et al., Drug Safety, 2026 | Cohort of more than 450,000 adults aged 50 or older with type 2 diabetes, followed for about three years | DPP-4 inhibitors or GLP-1 receptor agonists vs sulfonylureas | DPP-4 inhibitors were linked to 23% lower dementia risk. GLP-1 drugs showed a similar pattern with less certainty because fewer people used them. | The comparison drug may itself raise risk, and the study is observational |
The Tang study used a method called target trial emulation, which tries to mimic a randomized trial using existing records. It is a more careful design than a simple comparison, but it still cannot randomize who receives a drug.
What the randomized evidence found
| Evidence | Participants | Question | Result |
|---|---|---|---|
| EVOKE and EVOKE+ (Novo Nordisk), topline November 2025 | 3,808 people aged 55 to 85 with mild cognitive impairment or mild dementia due to Alzheimer’s disease and confirmed amyloid | Does oral semaglutide slow disease progression over two years? | No statistically significant slowing on the main clinical scale (CDR-SB). Some biomarkers improved, but that did not translate into clinical benefit. The one-year extension was stopped. |
| AAN 2026 conference abstract, network meta-analysis | 26 randomized trials of glucose-lowering drugs reporting dementia outcomes, 164,531 participants | Does any individual drug lower all-cause dementia risk? | Liraglutide at 1.8 mg daily was associated with lower risk. Other GLP-1 drugs showed non-significant trends. |
The network meta-analysis is a conference abstract and has not been confirmed here as a peer-reviewed publication. Treat it as a signal worth following, not a settled result.
Why observational studies cannot prove cause
- Prescribing is not random. People who receive newer drugs may differ in health, access to care and monitoring in ways that also affect dementia risk.
- The comparison drug matters. Sulfonylureas carry a higher risk of low blood sugar than incretin drugs. Researchers on the Drug Safety study noted that less hypoglycemia may itself contribute to lower cognitive decline, so part of the gap may reflect the comparator.
- Follow-up is short for a slow disease. Dementia develops over many years, and an editorial accompanying the 2025 papers pointed to short follow-up and low dementia rates.
- Dementia is identified through records. Coding and diagnosis timing can differ between groups and distort comparisons.
- Drug classes are lumped together. Newer drugs such as semaglutide may behave differently from older ones, and combining them can hide those differences.
- The people studied have type 2 diabetes. The findings do not show what happens in people who take these drugs for weight loss, a group researchers have called out as worth studying.
What the trial result does and does not close
EVOKE and EVOKE+ asked a specific question: can semaglutide slow decline in people who already have early Alzheimer’s symptoms and confirmed amyloid? The answer was no over the two-year treatment period. That does not answer whether GLP-1 drugs lower the chance of developing dementia in the first place, whether longer treatment or an earlier start would change the picture, or whether a specific drug differs. Those remain open.
How this fits the bigger risk picture
Diabetes is one of 14 modifiable dementia risk factors identified by the 2024 Lancet Commission. That helps explain why diabetes drugs are being studied, and why any benefit seen in people with diabetes may reflect better metabolic control and fewer vascular problems rather than a direct effect on brain disease. Our explainer on relative and absolute risk shows why a percentage reduction needs the starting risk beside it.
Frequently asked questions
Do GLP-1 drugs prevent dementia?
That has not been shown. Observational studies in people with type 2 diabetes link them to lower dementia risk, but those studies cannot establish cause. A large Alzheimer’s treatment trial of semaglutide found no slowing of decline in people already affected.
Why do studies report different percentages?
They compare different drugs in different groups over different time periods. A reduction measured against sulfonylureas is not the same as one measured against several other glucose-lowering drugs, so the numbers should not be lined up side by side.
Should I take a GLP-1 drug to protect my brain?
The current evidence does not support that as a reason on its own. These drugs are prescribed for specific conditions, have side effects, and are chosen with a clinician. This article cannot advise on any individual decision.
Sources
- Tang H, Donahoo WT, DeKosky ST, et al. GLP-1RA and SGLT2i medications for type 2 diabetes and Alzheimer disease and related dementias. JAMA Neurology, 2025. DOI 10.1001/jamaneurol.2025.0353.
- Wang Y-H, Renoux C, et al. Incretin-based drugs and the risk of dementia among patients with type 2 diabetes. Drug Safety, 2026. McGill University summary: McGill news release.
- Novo Nordisk. Topline results of the evoke and evoke+ trials, with conference presentation at the American Academy of Neurology: AAN abstract. Summary: Alzheimer Europe.
- Drug-specific effects of GLP-1 receptor agonists on dementia risk: a network meta-analysis. AAN 2026 conference abstract: abstract.
- UCI MIND. Novo Nordisk announces disappointing GLP-1 results: university commentary.
- Livingston G, Huntley J, Liu KY, et al. Dementia prevention, intervention, and care: 2024 report of the Lancet standing Commission. The Lancet, 2024. DOI 10.1016/S0140-6736(24)01296-0.
