Why semaglutide failed in Alzheimer’s: the trial leaders’ own post-mortem
Two large trials found no benefit. Semaglutide lowered inflammation markers, but only slightly in the brain and with no effect on decline. The trial leaders now think lowering inflammation in the body may not be enough.
In March 2026 the two phase 3 EVOKE trials of oral semaglutide in early Alzheimer’s disease were published in The Lancet, and they were negative. They enrolled 3,808 people aged 55 to 85 with amyloid-confirmed mild cognitive impairment or mild dementia, at 566 sites in 40 countries. After 104 weeks, scores on the main measure of decline, the Clinical Dementia Rating Sum of Boxes, had worsened by almost exactly the same amount with semaglutide as with placebo: 2.3 versus 2.3 points in EVOKE and 2.2 versus 2.1 in EVOKE+. Neither difference was statistically significant.
On September 24, 2026, the trials’ steering-committee co-chairs, Jeffrey Cummings and Philip Scheltens, published a review in The Journal of Prevention of Alzheimer’s Disease on what the program taught. Both disclose consulting ties to Novo Nordisk, the trials’ sponsor.
What the drug did, and did not, change
Semaglutide produced a robust, statistically significant fall in high-sensitivity C-reactive protein, a blood marker of inflammation in the body. It had no effect on most other blood measures. In spinal fluid, two markers of activated support cells in the brain (YKL-40 and GFAP) fell, along with nominal shifts in core Alzheimer’s markers, but the changes were small, about 7% to 10%, and they did not track with how fast anyone declined.
Lowering inflammation in the body may not be enough to change the course of Alzheimer’s once symptoms have begun. They do not rule out benefit in patients selected for high inflammation, or from GLP-1 drugs designed to reach the brain more easily.
Why the early signal misled
The trials were built on observations that people with type 2 diabetes taking semaglutide developed dementia less often. The review argues that treating biomarker-confirmed, symptomatic Alzheimer’s is a different problem from lowering dementia risk in cognitively healthy people with diabetes. Most EVOKE participants, about 86%, did not have diabetes.
That prevention question is still open. A separate study published September 20, 2026 emulated a trial in health records from 1,320 matched adults with both type 2 diabetes and mild cognitive impairment: those who started a GLP-1 drug developed dementia less often than those who started a DPP-4 inhibitor (hazard ratio 0.74). Its authors call the result hypothesis-generating; it is not a randomized trial.
What it means for you
Semaglutide is not a treatment for Alzheimer’s disease. Its proven benefits for weight, blood sugar and heart and kidney outcomes are unchanged. Our GLP-1 guide covers those trials, and our Alzheimer’s prevention article covers the risk factors with stronger evidence.