Two GLP-1 trials in Alzheimer's report out in one week. Neither slowed the disease.

Within three days at the start of December, the two largest tests of the idea that diabetes and obesity drugs might slow Alzheimer's disease both came back short.

On 1 December, Nature Medicine published ELAD, a phase 2b trial of liraglutide in 204 people with mild to moderate Alzheimer's disease syndrome and no diabetes [s2]. Within days, detailed results from the far larger EVOKE and EVOKE+ programme — 3,808 adults aged 55 to 85 with mild cognitive impairment or mild dementia due to Alzheimer's, confirmed by amyloid biomarkers — were reported from the 18th Clinical Trials on Alzheimer's Disease conference [s1]. Neither trial slowed the disease.

What the semaglutide trials found

EVOKE and EVOKE+ tested once-daily oral semaglutide against placebo over a planned two-year treatment phase, with a 12-month extension [s1]. The primary endpoint was the Clinical Dementia Rating – Sum of Boxes, a composite of cognition and everyday function that is the standard currency of Alzheimer's trials.

Semaglutide did not produce a statistically significant slowing of progression on that measure in either trial [s1]. Analyses of secondary cognitive and functional outcomes did not show meaningful differences between the treated and placebo groups either [s1].

What did move were the biomarkers. Reductions of up to roughly 10% were seen in certain Alzheimer's-related markers [s1] — enough to show the drug was doing something biologically, not enough to register in how patients did. Novo Nordisk has discontinued the one-year extension period on the basis of the efficacy results [s1].

The topline had already been released on 24 November, and the Alzheimer's Association's statement at the time framed the same tension: biomarker improvement in both trials that "did not translate into a delay of disease progression" [s3].

What the liraglutide trial found

ELAD is a different kind of study — smaller, older in conception, and built around imaging rather than a clinical scale. It randomised 204 participants to daily injections of liraglutide or placebo for 52 weeks, and its primary outcome was change in cerebral glucose metabolic rate measured by fluorodeoxyglucose PET [s2].

That primary outcome was negative: a difference of −0.17 between groups (95% CI, −0.39 to 0.06; P = 0.14) [s2].

Among the secondary outcomes, the executive-function domain of the Alzheimer's Disease Assessment Scale favoured liraglutide (0.15; 95% CI, 0.03 to 0.28; unadjusted P = 0.01) [s2]. Two caveats sit on that number and the trial's authors flag both in the reporting: it is a secondary outcome in a trial whose primary outcome failed, and the P value is unadjusted for multiple comparisons. The other secondary clinical measures showed nothing. On the Alzheimer's Disease Cooperative Study–Activities of Daily Living scale the difference was −0.58 (95% CI, −3.13 to 1.97; unadjusted P = 0.65), and on the Clinical Dementia Rating–Sum of Boxes it was −0.06 (95% CI, −0.57 to 0.44; unadjusted P = 0.81) [s2]. Liraglutide was generally safe and well tolerated in this non-diabetic population [s2].

Why anyone expected otherwise

The GLP-1-in-dementia hypothesis was never arbitrary. It rested on a stack of observations — insulin signalling in the brain, neuroprotective effects in animal models of Alzheimer's, and epidemiology suggesting people on these drugs for metabolic reasons developed dementia at lower rates. ELAD's own rationale was explicitly the animal-model neuroprotection literature [s2].

That stack has now been tested twice at very different scales, and the answer in symptomatic Alzheimer's disease is the same both times.

What this does and does not settle

It settles less than the headline suggests. Both trials enrolled people who already had clinical symptoms — mild cognitive impairment or mild-to-moderate dementia [s1][s2]. Neither tested whether a GLP-1 drug given years earlier, to people who are metabolically at risk but cognitively intact, changes anything. That is a different trial and it has not been run.

It also does not settle the biomarker question in a satisfying way. Something in the semaglutide trials moved by up to about 10% [s1], and the field will spend some time arguing about what that means: a real but insufficient effect on pathology, an effect on the wrong pathology, or an effect arriving too late in the disease to matter.

For people currently taking these drugs for obesity or type 2 diabetes, nothing here speaks to those indications. Neither trial was designed to.

What to watch

Three things. Whether the full EVOKE and EVOKE+ datasets, when they reach peer review, contain subgroup or biomarker signals that survive scrutiny. Whether anyone funds a prevention trial in cognitively normal adults at metabolic risk — a study that would take years and cost a great deal. And how the Alzheimer's pipeline reallocates: the Alzheimer's Association counted 182 active clinical trials evaluating 138 novel drugs at the time of the topline release [s3], and two of the most heavily watched slots in that pipeline just came free.

Sources

1. Results of EVOKE and EVOKE+ trials show no effect of oral semaglutide on AD progression — Alzheimer Europe, 3 December 2025
2. Liraglutide in mild to moderate Alzheimer's disease: a phase 2b clinical trial — Nature Medicine, 1 December 2025
3. Alzheimer's Association Statement on Oral Semaglutide Phase 3 Topline Data Release — Alzheimer's Association, 24 November 2025