Semaglutide and Alzheimer's: Reading the EVOKE Trial Results

TL;DR: Oral semaglutide did not work for cognition. In two large phase 3 trials, EVOKE and EVOKE+, once-daily oral semaglutide 14 mg failed to slow decline on the primary endpoint (CDR-SB) versus placebo over two years, and failed on the secondary functional endpoint too. A pooled analysis found no delay in progression from mild cognitive impairment to dementia. The one genuinely interesting finding is that semaglutide did lower a marker of inflammation by roughly 30%, and that biological change did not translate into any measurable cognitive or functional benefit at all.

Why this trial got run in the first place

Semaglutide’s path into an Alzheimer’s trial followed a real, published line of reasoning: chronic low-grade inflammation is one of the mechanisms thought to drive Alzheimer’s progression, GLP-1 receptors sit on microglia and other immune cells in the brain, and semaglutide reliably lowers systemic inflammatory markers in its metabolic trials. If peripheral and central inflammation talk to each other the way some preclinical literature suggests, a drug that turns down inflammation might, in principle, turn down one of the engines pushing Alzheimer’s forward. That is the neuroinflammation hypothesis in a sentence, and it was plausible enough, alongside earlier observational signals of lower dementia rates in people with diabetes on GLP-1 therapy, for Novo Nordisk to fund two phase 3 trials built to test it directly.

Those trials are EVOKE and EVOKE+, and they are now read out, published, and stopped early. This is the honest version of what they found.

What EVOKE and EVOKE+ actually tested

Both trials enrolled adults aged 55 to 85 with amyloid-confirmed, early-stage symptomatic Alzheimer’s disease, meaning mild cognitive impairment (MCI) or mild dementia due to AD, not advanced disease. Combined, they randomized 3,808 people: 1,855 in EVOKE (928 semaglutide, 927 placebo) and 1,953 in EVOKE+ (976 semaglutide, 977 placebo). The trials differed on one inclusion detail: EVOKE excluded people with cerebrovascular disease on brain imaging, while EVOKE+ allowed it.

Participants took oral semaglutide, titrated from 3 mg to 7 mg at four weeks and up to a flexible 14 mg dose at eight weeks, the same escalation logic used in oral semaglutide’s Wegovy pill form, just aimed at a different endpoint. The primary analysis was scheduled at two years (week 104), with a planned one-year extension discontinued once those results came in.

The primary endpoint was change in CDR-SB, the Clinical Dementia Rating - Sum of Boxes. It is worth defining plainly because the whole result hinges on it: CDR-SB is a clinician-rated scale that scores memory, judgment, home and community function, and personal care across six domains, summed into a single number where higher scores mean worse impairment. It is one of the standard yardsticks regulators and researchers use to ask “did this person’s dementia get measurably worse,” and it is what both EVOKE and EVOKE+ were built around.

The primary result: no separation from placebo

The result is not ambiguous. In EVOKE, mean CDR-SB change from baseline to week 104 was 2.3 points with semaglutide versus 2.3 with placebo, an estimated difference of -0.08 (95% CI, -0.35 to 0.20; p=0.57). In EVOKE+, the numbers were 2.2 versus 2.1, a difference of 0.10 (95% CI, -0.17 to 0.38; p=0.46). Both confidence intervals straddle zero comfortably. Reporting from the CTAD conference described the two curves as declining “neck-and-neck,” essentially on top of each other for the full two years.

The secondary endpoint, ADCS-ADL-MCI (a functional scale measuring everyday activities like managing finances or medications), showed the same pattern, and so did every other cognitive measure collected, including MoCA, ADAS-Cog13, MMSE, and ADCOMS.

The most concrete way to see the null result is the progression analysis. Researchers pooled data across both trials to ask whether semaglutide delayed the transition from MCI to full dementia, arguably the outcome patients and families care about most. It did not: participants progressed from MCI to dementia at effectively the same rate whether they were taking semaglutide or placebo.

The part that is actually interesting: biomarkers moved, outcomes didn’t

This is where EVOKE stops being just another negative trial and becomes a useful piece of science. Semaglutide did do something measurable in the body. Plasma high-sensitivity C-reactive protein, hs-CRP, a standard systemic inflammation marker, dropped by roughly 30% in the semaglutide group in both trials. A subset who underwent additional biomarker testing also showed small but statistically detectable reductions, generally under 10%, in several cerebrospinal fluid markers tied to Alzheimer’s pathology, including phosphorylated tau species, total tau, neurogranin, and YKL-40. None of that moved the needle on cognition or function.

That dissociation is the scientifically useful takeaway. Semaglutide reaches the brain only in small amounts: EVOKE’s own pharmacokinetic sub-analysis found a cerebrospinal-fluid-to-plasma ratio of about 0.4%, meaning the drug barely crosses into the central nervous system at the doses tested. The peripheral inflammation-lowering effect may be real and mechanistically sound, yet still too indirect, too small, or arriving too late to slow neurodegeneration once amyloid and tau pathology are already established. A hypothesis being wrong for the clinical endpoint is not the same as a mechanism being fake; it can also mean the mechanism was real but not sufficient, or not delivered to the right tissue in the right amount.

Timeline: from topline to full data

That gap between topline and full publication is normal for a trial this size, and it is why the CRP and CSF biomarker detail only became public at the March conference and Lancet paper, months after the headline “semaglutide failed in Alzheimer’s” had already circulated.

What the safety data showed

Semaglutide’s safety profile in EVOKE and EVOKE+ tracked its established pattern from diabetes and weight-management trials. Gastrointestinal effects were the dominant complaint: nausea in roughly a quarter of the semaglutide group, diarrhea in about 14%, vomiting in about 12%, and more participants on semaglutide discontinued for these reasons than on placebo. Weight loss followed the expected pattern too, averaging about 5.8% over two years versus a roughly 0.6% gain on placebo. Serious, severe, and fatal adverse events did not differ meaningfully between arms. None of that changes the efficacy story: a drug can be safe and simply not do the job it was tested for.

What this does and doesn’t mean for GLP-1 drugs generally

It is worth being precise about the boundaries of this result. EVOKE and EVOKE+ tested one drug, one route, one dose range, in people who already had confirmed amyloid pathology and measurable symptoms. They do not test whether a GLP-1 drug given years earlier, before amyloid accumulates, could behave differently; they do not test tirzepatide or other dual agonists, which act on an additional receptor and are covered in our what is tirzepatide guide; and they do not touch the separate, better-supported story of semaglutide’s effect on alcohol cravings, which runs through a different reward-circuit mechanism and has its own randomized trial evidence. Failing on Alzheimer’s cognition says nothing about that separate finding.

What EVOKE does responsibly close off is the specific idea it was built to test: that lowering peripheral inflammation with a GLP-1 drug, in people who already have symptomatic Alzheimer’s, meaningfully slows the disease. Two large, well-conducted trials looked for that effect directly and did not find it, even while confirming the drug did lower inflammation as predicted. That is a clean, informative failure rather than a muddy one, and it narrows the next round of hypotheses rather than leaving the question wide open. It is simply not a win for the patients and families who were hoping this drug, at this dose, in this population, would slow things down. For background on what the drug does outside this trial, see what semaglutide is, and for the tolerability side independent of this result, our GLP-1 side effects management guide covers the same gastrointestinal pattern EVOKE reported.

Frequently asked questions

Did semaglutide work for Alzheimer’s disease?

No. In both EVOKE and EVOKE+, oral semaglutide 14 mg showed no significant difference from placebo on the primary endpoint, change in CDR-SB score at two years, and no benefit on the secondary functional endpoint (ADCS-ADL-MCI) either. A pooled analysis also found no delay in progression from mild cognitive impairment to dementia.

What were the EVOKE and EVOKE+ trials?

They were two phase 3, randomized, placebo-controlled trials run by Novo Nordisk testing oral semaglutide in 3,808 adults aged 55 to 85 with amyloid-confirmed mild cognitive impairment or mild dementia due to Alzheimer’s disease. EVOKE excluded participants with cerebrovascular disease on imaging; EVOKE+ allowed it. Both ran a two-year primary analysis, with a planned one-year extension discontinued after the null result.

If semaglutide reduced inflammation, why didn’t it help cognition?

Semaglutide lowered plasma CRP by about 30% and nudged several cerebrospinal fluid biomarkers in both trials, confirming it affects inflammation-related biology systemically. But the drug barely crosses into the brain, with a cerebrospinal-fluid-to-plasma ratio of roughly 0.4%, and by the time someone has symptomatic Alzheimer’s, amyloid and tau pathology may already be too advanced for a peripheral anti-inflammatory effect to change the disease trajectory.

What is CDR-SB, the main measure used in these trials?

CDR-SB, the Clinical Dementia Rating - Sum of Boxes, is a clinician-administered scale scoring six domains, including memory, judgment, and daily function, where higher numbers mean worse impairment. It is a standard endpoint in Alzheimer’s trials because it captures both cognitive and functional decline in one number that researchers can compare across studies.

Does this mean GLP-1 drugs don’t do anything in the brain?

Not necessarily. EVOKE tested one specific hypothesis, peripheral inflammation reduction slowing established symptomatic Alzheimer’s, and found it did not hold up. It says nothing about GLP-1 effects on other brain-related outcomes, such as the separate, trial-supported reduction in alcohol cravings, or about whether starting a GLP-1 drug much earlier, before amyloid accumulates, could behave differently.

Was the EVOKE trial safe, even though it didn’t work?

Yes. Its safety profile matched semaglutide’s known pattern from diabetes and weight-management trials: mostly gastrointestinal complaints like nausea, diarrhea, and vomiting, with no meaningful increase in serious, severe, or fatal adverse events versus placebo. The trial failed on efficacy, not on safety.

References

1. Efficacy and safety of oral semaglutide 14 mg (flexible dose) in early-stage symptomatic Alzheimer’s disease (evoke and evoke+): two phase 3, randomised, placebo-controlled trials. The Lancet. 2026; PIIS0140-6736(26)00459-9.
2. Baseline characteristics from evoke and evoke+: Two phase 3 randomized placebo-controlled trials of semaglutide in participants with early-stage symptomatic Alzheimer’s disease. PMC.
3. Novo Nordisk A/S. Evoke phase 3 trials did not demonstrate a statistically significant reduction in Alzheimer’s disease progression. Company announcement, November 24, 2025.
4. GLP-1 Semaglutide Fails to Outperform Placebo in Phase 3 EVOKE Trial of Alzheimer Disease. NeurologyLive, CTAD 2025 conference coverage.
5. Semaglutide Does Not Treat Alzheimer’s. Could It Prevent Dementia? ALZFORUM, AD/PD 2026 conference coverage.

This article is an educational research reference and is not medical advice; consult a qualified clinician about any personal treatment decision.