Two randomised trials of glucagon-like peptide-1 (GLP-1) receptor agonists in Alzheimer's disease reported sharply different headline results within weeks of each other in late 2025. The larger — Novo Nordisk's Phase 3 EVOKE programme of oral semaglutide — failed to slow cognitive decline. The smaller — an academic Phase 2b trial of injectable liraglutide led by Imperial College London — reported signals on secondary measures of brain structure and cognition, while missing its own primary endpoint.

Both drugs belong to the same class and are licensed in the UK for diabetes and weight management, not for dementia. Read together, the two datasets do not support a therapeutic indication. They do suggest a testable hypothesis about disease stage, and they illustrate how easily a class-level "neuroprotective" claim can outrun the evidence.

What the two trials measured

The EVOKE and EVOKE+ trials tested oral semaglutide in early-stage Alzheimer's disease. Together they enrolled 3,808 participants across 566 sites in 40 countries, with a primary endpoint of change from baseline on the Clinical Dementia Rating Scale Sum of Boxes (CDR-SB) over 104 weeks, using quadruple-blind methodology. Semaglutide was titrated from 3 mg to 7 mg at four weeks and up to 14 mg at eight weeks — a dose comparable to that used in type 2 diabetes.

The Imperial ELAD study was a much smaller academic trial: 204 non-diabetic participants with mild-to-moderate Alzheimer's disease, testing injectable liraglutide. Its primary endpoint was a change in cerebral glucose metabolism — an exploratory, mechanistic measure. Cognitive and imaging outcomes were secondary, and the study was not powered to detect changes in cognition.

What EVOKE and EVOKE+ found

Both Phase 3 trials failed their primary endpoint. In EVOKE, CDR-SB worsened by 2.3 points with semaglutide versus 2.3 points with placebo (difference −0.08; p=0.57). In EVOKE+, it worsened by 2.2 points with semaglutide versus 2.1 points with placebo (difference 0.10; p=0.46). On the primary and each secondary cognitive and functional outcome, the semaglutide and placebo curves were effectively indistinguishable.

There was a notable biomarker paradox: treatment moved some Alzheimer's-related fluid biomarkers in the expected direction, but this did not translate into a clinical benefit. On the strength of the results, Novo Nordisk discontinued the planned one-year extension periods. The full results were published in The Lancet in March 2026 and presented at the AD/PD 2026 conference.

What the Imperial ELAD trial of liraglutide found

Professor Paul Edison's team at Imperial College London reported that liraglutide was associated with roughly 50% less brain-volume loss and an 18% slower decline in cognitive function on secondary measures, with the findings published in Nature in 2025. These are the results that generated the more hopeful headlines.

The caveats matter as much as the numbers. The trial missed its primary endpoint — the change in cerebral glucose metabolism — and, by design, was not powered to detect cognitive change. Positive signals on secondary, underpowered endpoints are hypothesis-generating; they are not confirmation that the drug modifies the disease.

Why the two results are not a straightforward contradiction

The two trials differ in ways that make a direct comparison misleading:

  • Different molecules. Semaglutide and liraglutide are both GLP-1 receptor agonists, but they differ in half-life, receptor pharmacology and brain exposure.
  • Different route and dose. EVOKE used an oral formulation at diabetes-range doses; ELAD used injectable liraglutide.
  • Different populations. EVOKE enrolled early-stage disease; ELAD enrolled mild-to-moderate disease.
  • Different endpoints and power. EVOKE was powered for a clinical endpoint; ELAD was an exploratory mechanistic study with cognition as a secondary measure.

A confident negative Phase 3 result in established symptomatic disease does not disprove a preventive or earlier-stage effect — a point the Imperial authors themselves make, arguing for trials in preclinical disease using biomarkers to select those at highest risk. Equally, an encouraging secondary signal in a small, underpowered trial is not evidence of efficacy. Both statements can be true at once.

The UK regulatory position

No GLP-1 receptor agonist is licensed in the UK for Alzheimer's disease. Semaglutide and liraglutide remain prescription-only medicines (POM) authorised for their licensed indications — type 2 diabetes and/or weight management. Any use outside the marketing authorisation is a clinical decision governed by the Medicines Act 1968 and the Human Medicines Regulations 2012, and sits outside this site's scope.

The MHRA has repeatedly warned about GLP-1 medicines used outside their licensed indications and about illegal online supply of weight-loss products. This does not change with a research result in a different disease area: a negative or mixed trial in Alzheimer's has no bearing on the licensed use of these medicines, and no bearing on the legal status of research peptides.

What the evidence does and does not support

It does not support any claim that GLP-1 receptor agonists treat, cure or prevent Alzheimer's disease, and Peptide Data makes no such claim. The honest summary of the current picture is: one well-powered, confidently negative Phase 3 programme; one small, exploratory Phase 2b trial with a missed primary endpoint and positive secondary signals; and an epidemiological signal that would need properly powered prevention trials to test. Until those exist, GLP-1s remain a metabolic medicine class with an unresolved and unproven neurological hypothesis — not a dementia treatment.

This article is AI-researched and editorially reviewed. It is provided for research and educational purposes only and is not medical advice. Research peptides are not licensed for human consumption in the UK.