Semaglutide Fails to Slow Alzheimer's in Largest GLP-1 Brain Trial, but Biomarker Shifts Keep the Metabolic Hypothesis Alive
Full results from Novo Nordisk's EVOKE and EVOKE+ trials, published in The Lancet, show oral semaglutide had no effect on cognitive decline over two years despite reducing key cerebrospinal fluid markers of neuroinflammation and tau pathology by roughly 10 percent.
Overview
The full results from the two largest clinical trials ever to test a GLP-1 receptor agonist against Alzheimer’s disease have been published in The Lancet on March 19, 2026, confirming what topline data suggested in late 2025: oral semaglutide does not slow cognitive or functional decline in people with early-stage Alzheimer’s. The EVOKE and EVOKE+ Phase 3 trials, conducted by Novo Nordisk across 566 sites in 40 countries, enrolled 3,808 participants aged 55 to 85 with amyloid-confirmed disease.
Yet the story is not purely negative. Buried in the cerebrospinal fluid substudy data are statistically significant reductions in tau pathology and neuroinflammation markers — effects that, while too small to move the clinical needle, have kept the metabolic hypothesis of Alzheimer’s from being abandoned entirely.
What We Know
The trials tested oral semaglutide at doses up to 14 mg daily against placebo over 104 weeks. The primary endpoint was change in the Clinical Dementia Rating-Sum of Boxes (CDR-SB), a standard composite measure of cognitive and functional decline. According to The Lancet publication, mean CDR-SB changes from baseline to week 104 were 2.3 and 2.2 with semaglutide versus 2.3 and 2.1 with placebo in the two trials respectively (estimated difference: -0.08, 95% CI -0.35 to 0.20, p=0.57 in EVOKE; 0.10, -0.17 to 0.38, p=0.46 in EVOKE+). The secondary endpoint — change in the Alzheimer’s Disease Cooperative Study-Activities of Daily Living Scale for MCI — also showed no separation between treatment groups.
However, the CSF biomarker substudy told a different story. In participants treated with semaglutide, there were nominally significant reductions of approximately 8 to 10 percent across several markers tied to Alzheimer’s pathology. According to The Lancet data, phosphorylated tau-181 showed a treatment ratio of 0.92, phosphorylated tau-217 reached 0.91, and non-phosphorylated tau-181 and tau-205 both came in at 0.90 and 0.91 respectively. Markers of neuroinflammation (YKL-40: 0.93), neurodegeneration (total tau: 0.93), and synaptic function (neurogranin: 0.92) all moved in the same direction.
The safety profile was consistent with semaglutide’s established record. Treatment-emergent adverse events were reported in 91.2 percent of semaglutide participants versus 84.8 percent on placebo, according to The Lancet. Five fatalities were considered treatment-related by investigators — one in the semaglutide group and four in the placebo group.
Novo Nordisk discontinued the planned one-year extension period in both trials following the negative topline readout in November 2025 and has terminated the Alzheimer’s semaglutide program.
What We Don’t Know
The central puzzle is why the drug moved biomarkers without moving cognition. Professor Paul Edison of Imperial College London, who served as principal investigator at the London site, suggested in an analysis published by Imperial that the oral formulation’s chemical properties — optimized for gut absorption — may have limited its penetration into the brain. Under this interpretation, the concept of targeting GLP-1 receptors in the brain is not disproven; rather, the delivery mechanism may have been inadequate.
Plasma biomarkers of tau (p-tau181 and p-tau217) did not show the same reductions observed in the CSF substudy, raising further questions about how much of the drug’s effect was centrally versus peripherally mediated. It is worth noting that plasma p-tau217 has emerged as a powerful diagnostic tool in its own right — as previously reported by The Machine Herald, a Washington University study showed the biomarker can estimate Alzheimer’s symptom onset within three to four years.
Whether a more brain-penetrant GLP-1 agonist — or a combination approach pairing metabolic modulators with anti-amyloid therapies — could succeed where oral semaglutide failed remains untested. As Dr. Howard Fillit of the Alzheimer’s Drug Discovery Foundation noted, existing anti-amyloid drugs slow cognitive decline by approximately 30 percent, leaving the remaining 70 percent unaddressed and underscoring the need for therapies targeting alternative pathways.
Analysis
The EVOKE results represent a significant setback for the hypothesis that repurposed metabolic drugs could offer a shortcut to Alzheimer’s treatment. GLP-1 agonists had been surrounded by observational data suggesting lower dementia risk among users, and the prospect of leveraging a drug class already prescribed to millions carried obvious appeal.
But the biomarker signal — however modest — prevents a clean verdict. A roughly 10 percent reduction in CSF tau markers and neuroinflammation is meaningful in directional terms, even if it was not sufficient to produce clinical benefit at the tested dose and formulation. The field’s attention is now likely to turn to injectable or next-generation formulations that achieve higher central nervous system concentrations, as well as to combination strategies.
Dr. Fillit’s framing of Alzheimer’s as a condition ultimately requiring precision medicine — with treatment tailored to each patient’s biomarker profile — reflects the broader shift away from monotherapy approaches. More than 70 percent of the current Alzheimer’s pipeline now targets non-amyloid mechanisms, according to the Alzheimer’s Drug Discovery Foundation. The EVOKE data, paradoxically, may accelerate that diversification by demonstrating that metabolic pathways can be nudged pharmaceutically in Alzheimer’s patients, even if the first attempt fell short.