Inside evoke and evoke+: How the Largest GLP-1 Dementia Trial Was Run

3,808 volunteers, 40 countries, two years. A close look at the trial design behind a landmark null result — and why the way it was built matters as much as its answer.

Behind every headline result sits a study design, and the design is where trust is earned or lost. The Evoke trial and its twin, Evoke+, tested whether oral semaglutide could slow the progression of early Alzheimer’s disease. The answer, reported in late 2025 and published in 2026, was that it could not. The reason the answer carries weight is the rigor of how the question was asked.

Who was enrolled — and why it was strict

Novo Nordisk built evoke and evoke+ to a demanding standard. The trials enrolled 3,808 adults aged 55 to 85 with mild cognitive impairment or mild dementia due to Alzheimer’s. Critically, eligibility required confirmed amyloid pathology — biomarker evidence that participants actually had Alzheimer’s, not another cause of memory loss. That single criterion has sunk many older dementia trials, which quietly enrolled people whose decline had nothing to do with the mechanism under test.

Recruitment spanned roughly 40 countries, a scale that guards against quirks of any one health system or population. Participants received oral semaglutide or placebo on top of their standard care in a double-blind, double-masked, double-anonymized study, so neither patients nor investigators knew who received what.

What CDR-SB actually captures

The primary endpoint was the Clinical Dementia Rating–Sum of Boxes, or CDR-SB. It is worth understanding because it is why the result is meaningful. CDR-SB does not test memory in isolation; it grades function across everyday domains — memory, judgment, community affairs, home life, personal care. A clinician builds the score from interviews with the patient and a care partner.

That design makes CDR-SB a measure of lived decline, not a lab abstraction. A drug that moves it is helping people navigate their days. A drug that does not move it is not — however encouraging its biomarkers look.

The result, and the discipline around it

Over two years, semaglutide and placebo groups declined at the same rate on CDR-SB. The drug was safe and well tolerated, and it shifted some Alzheimer’s biomarkers favorably — but the function that CDR-SB tracks did not diverge.

Novo then did the disciplined thing: it discontinued the planned one-year extensions. Once a treatment shows no benefit, continuing to ask volunteers to take it — and to keep some on placebo — is hard to justify. Stopping is the ethical signal of a trial run in good faith.

Why a well-built null result is valuable

It is tempting to read a negative trial as a failure. In clinical science, a large, clean, well-controlled null is a contribution. evoke’s strict amyloid confirmation, its functional endpoint, its scale, and its blinding mean the result can be trusted: at this stage of Alzheimer’s, in this form, semaglutide does not slow decline.

That precision lets the field redraw its map. Researchers can now redirect resources rather than chase a signal that observational data had made tantalizing. The next questions — whether GLP-1 biology matters earlier, in prevention, or only in combination — are sharper because they close the gap left by the broad one.

What it means

For the families who enrolled hoping semaglutide might slow a parent’s fade, the outcome is a quiet loss, and their contribution was not wasted. They helped convert a decade of hope and correlation into a firm, usable fact.

Evoke is a reminder of what large trials are for. Real-world signals generate hypotheses. Only a study built like this one can tell you whether the hypothesis was true.

Frequently asked questions

Did semaglutide work for Alzheimer’s? No. In the evoke and evoke+ trials, oral semaglutide did not slow cognitive or functional decline over two years compared with placebo, despite favorable movement in some biomarkers.

What is CDR-SB? The Clinical Dementia Rating–Sum of Boxes is a scale that grades function across everyday domains — memory, judgment, home life, personal care — built from interviews with the patient and a care partner. It measures lived decline, not just a lab value.

Is Novo Nordisk still testing GLP-1 drugs for dementia? Novo discontinued the evoke extensions after the null result. Researchers now suspect that if GLP-1 biology matters for the brain, it may matter earlier — in prevention — or only in combination, questions that the Evoke trials could not address.

Sources

1. “Efficacy and safety of oral semaglutide in early-stage symptomatic Alzheimer’s disease (evoke and evoke+).” The Lancet, 2026. ClinicalTrials.gov NCT04777396 and NCT04777409.
2. Novo Nordisk A/S — evoke/evoke+ topline results, November 24, 2025 (CTAD conference, San Diego).
3. Cummings, J., et al. — background on CDR-SB as an Alzheimer’s trial endpoint, Alzheimer’s Research & Therapy.
4. Alzheimer’s Association — trial design and biomarker-confirmed enrollment standards, 2025–2026.