The semaglutide Alzheimer's trial result is a clear no. In two large studies called EVOKE and EVOKE+, 3,808 people with early Alzheimer's disease took a daily semaglutide tablet or a placebo, a look-alike pill with no drug in it. After two years, memory, thinking and everyday function had declined at the same pace in both groups. Semaglutide did not slow the disease, and no trial has shown that it prevents dementia.
Novo Nordisk, which makes semaglutide and paid for both trials, announced the headline result on November 24, 2025. As of September 22, 2026, the day this article was written, the full results are published in The Lancet (online March 19, 2026). The numbers below come from that paper.
What the semaglutide Alzheimer's trial set out to test
The question was simple: could a medicine already used for diabetes and weight slow Alzheimer's disease once symptoms have started?
It looked like a reasonable bet. Health-record studies had linked GLP-1 drugs with fewer dementia diagnoses in people with type 2 diabetes, and animal studies suggested they calm brain inflammation. Why those signals and this trial disagree comes further down.
Who took part in EVOKE and EVOKE+
The trials enrolled people aged 55 to 85 at 566 sites in 40 countries (registry records NCT04777396 and NCT04777409).
- Everyone had amyloid, the sticky protein that builds up in an Alzheimer's brain, confirmed by a scan or spinal-fluid test.
- About 72% had mild cognitive impairment, thinking problems that show up on testing but don't yet disable someone. The rest had mild dementia.
- Average age was 72. Only 13.6% had type 2 diabetes, and about 1 in 6 had obesity.
Half took semaglutide, stepping up from 3 mg a day to 14 mg, the top strength of the tablet sold as Rybelsus. Doses could be lowered or paused for side effects; by week 104, about 6 in 10 semaglutide participants still being followed were on the full 14 mg. This was the pill, not the weekly shot, and how the semaglutide pill works explains the difference.
What EVOKE measured
The main yardstick was the CDR-SB, short for Clinical Dementia Rating–Sum of Boxes. A trained rater interviews the person and a care partner about six areas of thinking and daily life, scoring each from 0 to 3. The total runs from 0 to 18; higher means more impairment. Participants started at an average of 3.7.
Researchers compared how much that score rose over 104 weeks. They also tracked everyday activities and how quickly people with mild cognitive impairment moved on to dementia. The trials were sized to detect a slowing of 0.44 points, about a fifth of the expected decline.
EVOKE and EVOKE+ results: no difference from placebo
Over two years, the CDR-SB rose 2.3 points on semaglutide and 2.3 on placebo in EVOKE, and 2.2 versus 2.1 in EVOKE+. The differences, −0.08 and +0.10 points, were nowhere near statistical significance, and the range the data allow stops short of the benefit the trials were built to find.
Nothing else separated the groups either. Daily-activity scores matched, and people with mild cognitive impairment reached dementia at the same rate in both groups.
Some lab markers did move. In a substudy of 199 people, semaglutide lowered several tau and inflammation markers in cerebrospinal fluid, the fluid around the brain and spinal cord, by up to 10%, and it lowered CRP, an inflammation marker in the blood. Very little drug reached the brain, though: 0.08 nmol/L in spinal fluid, against a modelled average of about 17 nmol/L in the blood. The markers shifted; people's lives didn't.
Weight fell an average 5.8% on semaglutide, against a 0.6% gain on placebo.
Side effects in the EVOKE trials
The side effects looked like semaglutide's anywhere else, and the authors found no new safety concerns. Stomach and appetite problems, and weight loss, drove most of the stopping. These rows come from the paper's safety table:
| Reported event | EVOKE semaglutide | EVOKE placebo | EVOKE+ semaglutide | EVOKE+ placebo |
|---|---|---|---|---|
| Serious adverse event | 19.0% | 25.0% | 21.6% | 22.7% |
| Stopped the drug because of an adverse event | 17.2% | 8.1% | 16.7% | 8.6% |
| Decreased appetite | 31.7% | 5.7% | 34.4% | 6.3% |
| Nausea | 25.2% | 6.9% | 23.4% | 6.8% |
| Diarrhea | 14.7% | 11.2% | 14.3% | 8.3% |
| Vomiting | 12.3% | 4.1% | 12.3% | 3.2% |
Falls were reported less often on semaglutide in both trials, though that wasn't a tested outcome, and deaths were equal at 2.7% in each group. Our semaglutide side effects guide covers the wider picture.
Does semaglutide prevent dementia? Why earlier studies disagreed
Most of the hopeful evidence came from observational studies, which look back at people who happened to take a drug instead of assigning it by chance.
- Health records. Among 1,094,761 people with type 2 diabetes, semaglutide users were 40% to 70% less likely to get a first Alzheimer's diagnosis within three years than users of other diabetes drugs (Wang et al., Alzheimer's & Dementia 2024).
- Pooled heart trials. Three diabetes trials of about 15,800 people recorded fewer dementia diagnoses on GLP-1 drugs than placebo, with wide uncertainty; dementia wasn't what they were built to measure (Nørgaard et al., 2022).
- A smaller trial. Liraglutide, an older daily GLP-1 shot, missed its main brain-scan endpoint in 204 people with Alzheimer's (Edison et al., Nature Medicine, December 2025).
Record studies can't fully separate a drug from the kind of person who gets it, and their diagnoses come from billing codes, not brain scans. A randomized trial lets chance decide instead.
The EVOKE authors offer reasons the two could differ: the record studies looked at preventing dementia in people with diabetes, often over longer periods, while EVOKE tried to slow a disease already causing symptoms, mostly in people without diabetes. They also note that weight loss and nausea are hard to hide, so some participants may have guessed their group.
As of September 22, 2026, the day this article was written, that debate is still live. In an August 2026 letter to The Lancet, Ivan Koychev and Rury Holman argued that EVOKE tested treatment, not prevention, which remains an open question (Koychev and Holman, 2026). A record study published September 20, 2026, of 1,320 people with type 2 diabetes and mild cognitive impairment, found dementia in 15.5% of those starting a GLP-1 drug versus 20.4% of those starting a DPP-4 inhibitor, another type of diabetes pill, over an average 3.9 years (Schechter et al., 2026). Its authors call it hypothesis-generating; it shares every limit above.
What EVOKE means if you take semaglutide now
It doesn't change why semaglutide is prescribed. The current Rybelsus label (revised January 2026) covers blood-sugar control and lowering the risk of major heart and stroke events in type 2 diabetes, with nothing about memory. Separate, mostly animal research on semaglutide and longevity is covered in semaglutide and aging.
EVOKE tested a branded tablet, not a compounded preparation. Through Promise, semaglutide is dispensed as a compounded medication, which is different from an FDA-approved product: the formulation offered here is not FDA-approved. The FDA explains that it does not review compounded drugs for safety, effectiveness or quality before they are marketed. A licensed provider may still prescribe a compounded formulation; that decision is between the patient and the doctor. A licensed provider reviews every request, and not everyone qualifies.
Tirzepatide, the other common weekly option, has no published randomized trial in Alzheimer's disease, so EVOKE can't be read across to it either way. Tirzepatide vs semaglutide covers the practical differences.
What to watch next
The open question is prevention. A UK trial called ISAP was designed to test daily oral semaglutide against placebo for one year in up to 88 people aged 55 or older who have amyloid in the brain but no or only mild memory problems, measuring tau buildup on brain scans (ISAP protocol, BMJ Open 2024). A scan change would be a clue, not proof of fewer dementia cases.
For now: in the only large randomized trials run on the question, semaglutide did not slow early Alzheimer's disease. Whether it can help prevent dementia is still unanswered.