Semaglutide and Alzheimer's: EVOKE review

In November 2025 Novo Nordisk announced that oral semaglutide did not slow cognitive decline in two large Alzheimer's trials, EVOKE and EVOKE+^[1]. The primary endpoint, change in Clinical Dementia Rating (CDR-SB) at week 104, was negative. The company discontinued the planned extension periods. That looked like the end of the story.

A June 2026 commentary in the Journal of Alzheimer's Disease by Christian Holscher argues the book may have been closed too early^[2]. Holscher re-examines the trial data at timepoints beyond week 104 and identifies a set of limited but real drug-placebo separations that the initial announcement did not emphasize.

What the trials tested

EVOKE and EVOKE+ were parallel Phase 3 studies, each enrolling roughly 1,840 adults aged 55 to 85 with mild cognitive impairment or mild dementia confirmed by amyloid biomarkers^[3]. Participants received oral semaglutide titrated to 14 mg daily (the same dose marketed as Rybelsus for type-2 diabetes^[4]) or placebo, on top of their existing Alzheimer's medications. The main difference between the two trials was that EVOKE excluded participants with subcortical vascular disease while EVOKE+ did not^[3].

The sole primary outcome was change in CDR-SB at week 104. Secondary measures included the ADAS-cog-13 (a cognitive test battery), the ADCS-ADL-MCI (a functional activities scale), and cerebrospinal fluid biomarkers in a substudy of about 200 patients^[1].

Where the later timepoints showed signals

The week 104 CDR-SB result was unambiguously negative: semaglutide did not separate from placebo on the primary endpoint^[1]. Holscher's re-analysis focuses on what happened after that cutoff.

On the ADCS-ADL-MCI (Activities of Daily Living for MCI), a statistically significant difference favouring semaglutide appeared at week 130 (p = 0.0039)^[2]. The ADAS-cog-13 cognitive scores showed a trend toward improvement by week 156, though the paper describes these as trends rather than firm statistical separations^[2]. Some CSF Alzheimer's biomarkers also showed small but significant changes.

To put the biomarker findings in proportion: the CSF substudy showed reductions of 10% or less in p-tau181, p-tau217, YKL-40, total tau, and neurogranin. The blood inflammatory marker hsCRP dropped by about 30%^[1]. None of that rescued the primary endpoint.

The blood-brain barrier problem

Holscher points to a pharmacokinetic explanation for the weak signals. Semaglutide is designed to stay in the bloodstream for a long time (its half-life is roughly seven days), and that same property means it does not cross the blood-brain barrier easily^[2]. Novo Nordisk's own data showed that brain exposure reached only about 0.4% of plasma levels after three months of treatment^[1]. If the drug barely reaches the target organ, modest and delayed effects are what you would expect.

The paper suggests that newer GLP-1 receptor agonists engineered for better brain penetration could produce stronger neuroprotective effects. Several such molecules are in preclinical or early clinical development, though none has reached the scale of the EVOKE programme^[2].

What this does not change

The re-analysis does not turn a negative trial into a positive one. The primary endpoint was missed, and it was missed convincingly. Novo Nordisk discontinued the extensions based on that result. The later-timepoint signals Holscher highlights are secondary and exploratory measures, not pre-specified wins. In Alzheimer's trial history, secondary signals that fail to replicate in subsequent studies are common.

For readers following semaglutide for its approved metabolic indications (type-2 diabetes and obesity), none of this changes the drug's regulatory status or safety profile. The Alzheimer's programme was a separate research question, and the answer was, at best, inconclusive.

What we do not yet know

Whether a GLP-1 agonist with genuinely high brain penetration would do better remains untested in a Phase 3 setting. Whether the later-timepoint signals in EVOKE represent a real, slow-onset treatment effect or statistical noise is not something this data can resolve. And whether the modest biomarker changes matter for patients over years of treatment is an open question that only longer, larger trials could answer.

For broader context on where semaglutide fits in the Alzheimer's research picture, see our earlier explainer on anti-dementia drug advances in 2026. For the drug's approved uses in metabolic disease, see the semaglutide overview.

This article is for informational purposes only and does not constitute medical advice. If you or someone you know has concerns about cognitive decline, consult a qualified healthcare provider.