EVOKE: Can Semaglutide Slow Cognitive Decline in Alzheimer's? The Phase 3 Result.

Novo Nordisk's EVOKE and EVOKE+ trials — 3,808 participants across 40 countries, the largest GLP-1 neurodegeneration program ever conducted — returned negative on the primary cognitive endpoint. Here's what the trials measured, what changed, what didn't, and what it means.

Why this trial existed in the first place

The hypothesis that GLP-1 medications might slow cognitive decline rested on three converging lines of evidence. First, type 2 diabetes is an established Alzheimer's risk factor — patients with poorly controlled diabetes have meaningfully higher rates of dementia. Second, observational data from diabetes registries showed that patients on GLP-1 medications had lower rates of incident Alzheimer's disease compared to patients on other diabetes treatments. Third, preclinical work in animal models showed GLP-1 agonists reduced amyloid plaque accumulation, neuroinflammation markers, and cognitive impairment in transgenic mouse models of AD.

That convergence was strong enough for Novo Nordisk to commit to two parallel phase 3 trials — a $1+ billion investment in proving the hypothesis in humans.

Trial design — what was actually tested

Both trials enrolled large, biomarker-confirmed populations — the design quality was as rigorous as any AD trial in the field. Of 3,808 randomized participants, 2,746 (72.1%) had MCI and 1,034 (27.2%) had mild AD dementia at baseline.

The headline result: no cognitive benefit

The November 24, 2025 topline announcement and March 2026 AD/PD presentation showed:

• Primary endpoint (CDR-SB at week 104): No significant difference between oral semaglutide and placebo in either trial
• Secondary endpoint (ADCS-ADL-MCI): No significant difference
• MMSE score: No significant difference
• ADAS-Cog score: No significant difference
• Pooled analysis of MCI-to-mild-AD progression: No delay with semaglutide vs placebo

Across every cognitive and functional measure used in the trials, oral semaglutide did not differ meaningfully from placebo. Based on the lack of efficacy in the main 2-year phase, Novo Nordisk discontinued the planned 1-year extension period.

What did move — biomarkers

The trial was not entirely null. Inflammation markers improved meaningfully:

High-sensitivity C-reactive protein — a systemic inflammation marker that has been independently associated with AD risk and progression — decreased significantly in semaglutide arms in both trials. Other AD-related biomarkers were also reported to improve, though the company has not yet released the full biomarker dataset publicly.

Safety — consistent with prior semaglutide data

Safety and tolerability across both trials were consistent with prior semaglutide experience in obesity and diabetes populations. Gastrointestinal adverse events (nausea, vomiting, diarrhea) were the most common, typically transient and dose-dependent. No new safety signals emerged. Dropout rates were similar between treatment and placebo arms.

Why this matters beyond Alzheimer's

Three implications:

1. The metabolic-pathway hypothesis isn't dead — it's narrower

The biomarker improvements (especially hsCRP) suggest GLP-1-mediated effects on systemic inflammation are real. The failure to translate to cognitive benefit suggests the inflammatory pathway alone is insufficient to alter AD trajectory at the MCI/mild dementia stage. This may reframe the hypothesis toward earlier intervention windows or combination therapies that pair metabolic modulation with anti-amyloid approaches.

2. GLP-1s for other neurodegenerative indications still warrant investigation

Parkinson's disease trials (notably exenatide in PD) have shown more encouraging early signals. The EVOKE failure does not generalize to all neurodegenerative conditions. Different mechanisms, different populations, different stages.

3. The risk-benefit framing for cognitive use is now clear

Patients should not start semaglutide off-label for "brain health" or to prevent dementia. The largest, best-designed trial of this hypothesis has now reported negative — and the result was negative across multiple cognitive endpoints, not a borderline miss.

What semaglutide is still appropriate for

This result has no implications for semaglutide's established uses. Patients on semaglutide for weight management or type 2 diabetes should continue therapy as prescribed. The cardiovascular benefits demonstrated in SELECT, the kidney benefits in FLOW, the obstructive sleep apnea benefits in SURMOUNT-OSA, and the MASH benefits in ESSENCE are all unaffected by EVOKE's negative result.

For patients pursuing approved indications

What's next in the GLP-1 cognitive research program

Two open questions remain:

• Earlier intervention. EVOKE enrolled symptomatic populations (MCI and mild AD). Whether GLP-1 intervention in preclinical AD (amyloid-positive but cognitively normal) might show different results requires its own trial — and is unlikely to be funded by Novo Nordisk given the EVOKE result.
• Combination with anti-amyloid therapy. Lecanemab and donanemab (now FDA-approved anti-amyloid antibodies) plus GLP-1 anti-inflammatory effects is a theoretically interesting combination. No such trial is currently registered.

For other neurodegenerative conditions — particularly Parkinson's disease — the GLP-1 hypothesis remains active and several mid-stage trials continue. The EVOKE result is specific to Alzheimer's at the MCI and mild dementia stage with monotherapy; it should not be over-extrapolated.

The honest summary

EVOKE and EVOKE+ were the largest, best-designed test of the GLP-1-for-Alzheimer's hypothesis ever conducted. They returned negative on the primary cognitive endpoint and on every secondary cognitive measure. Biomarkers moved; clinical outcomes did not. For now, this closes the case for using semaglutide to prevent or slow Alzheimer's disease in symptomatic populations. The broader GLP-1 cardiometabolic story — and the open questions in Parkinson's and earlier-stage AD — continues.