Semaglutide Alzheimer's Trial Results: What EVOKE Showed
No. In two phase 3 trials of 3,808 people with early Alzheimer's, oral semaglutide did not slow disease progression on the primary CDR-SB endpoint or any secondary measure, though some biomarkers improved. Topline results were announced 24 November 2025.
Semaglutides.org provides information only, not medical advice. Talk to a licensed provider about your own care before making changes. Some telehealth links are affiliate links and are labeled where they appear.
No. Semaglutide did not slow Alzheimer’s disease in the EVOKE and EVOKE+ phase 3 trials. On 24 November 2025, Novo Nordisk announced that the trials “did not confirm superiority of semaglutide versus placebo” on the primary endpoint — change in the Clinical Dementia Rating Sum of Boxes (CDR-SB) score from baseline to week 104 [3]. Alzheimer’s-related biomarkers did improve in both trials, but that did not translate into any delay of disease progression [3][4].
The picture was the same across every measure. For the primary endpoint and each secondary cognitive and functional outcome, the curves for the semaglutide and placebo groups were indistinguishable over the three-year trials [2]. A summary of the results on Alzforum called them “flat-out negative” [2]. Based on the efficacy results in the overall study population, Novo Nordisk discontinued the planned one-year extension period [3]. Semaglutide is marketed as Ozempic and Rybelsus for type 2 diabetes and Wegovy for chronic weight management; it has no Alzheimer’s indication anywhere [3].
What were the EVOKE and EVOKE+ trials?
EVOKE (NCT04777396) and EVOKE+ (NCT04777409) were two randomized, double-blind, placebo-controlled phase 3 trials testing once-daily oral semaglutide against placebo in early-stage symptomatic Alzheimer’s disease [1][11]. Eligible participants were men or women aged 55 to 85 with mild cognitive impairment (MCI) or mild dementia due to Alzheimer’s, with the diagnosis confirmed by an abnormal amyloid result on a PET scan or cerebrospinal fluid (CSF) analysis [1][3]. MCI was defined as a CDR global score of 0.5, and mild dementia as a CDR global score of 1.0 [3].
The trials screened nearly 10,000 people across 40 countries and randomized 3,808 participants — 1,855 in EVOKE and 1,953 in EVOKE+ — in a 1:1 ratio to oral semaglutide or placebo [2][3]. Enrollment ran from May 18, 2021 to September 8, 2023, with the main phase expected to finish in September 2025 [1]. The drug was given on top of standard of care [3].
The dosing schedule matched the type 2 diabetes regimen for oral semaglutide: 3 mg for weeks 0 to 4, 7 mg for weeks 4 to 8, then 14 mg from week 8 through week 156 [1][2]. Participants who had adverse events on the maintenance dose could drop to a lower dose for the rest of the trial [2]. The full treatment period was 156 weeks — a 104-week main phase plus a 52-week extension [1][3].
The primary endpoint was the difference between semaglutide and placebo in CDR-SB change from baseline to week 104 [1][3]. The CDR-SB covers six areas of cognition and function, each scored 0 to 3, for a maximum of 18 [3].
Notably, Novo Nordisk had not run a prior dose-finding phase 2 trial of semaglutide in Alzheimer’s before launching the phase 3 program [2]. The decision to pursue an Alzheimer’s indication rested on real-world evidence studies, preclinical models, and post-hoc analyses of diabetes and obesity trials [3][11].
What was the difference between EVOKE and EVOKE+?
The only design difference: EVOKE excluded people with cerebrovascular or small vessel pathology seen on CT or MRI scans, while EVOKE+ included them [2]. In practice, the split did not matter. Fewer than 3% of EVOKE+ enrollees had small vessel pathology — just 54 participants in total — so the two study populations were essentially identical, according to Novo Nordisk’s Peter Johannsen [2].
Who participated in the trials?
Participants averaged about 72 years old — 71.8 in EVOKE and 72.6 in EVOKE+ [2][11]. Women made up 53.0% of EVOKE and 51.8% of EVOKE+, and 76.6% of participants were White [11]. Half were overweight or obese, 47% were at a healthy weight, and 2.5% were underweight [2]. Nearly 14% had type 2 diabetes [2].
Baseline disease severity was early: mean CDR-SB was 3.7 in both trials, ADCS-ADL-MCI scores were 39.4 and 38.9, and 72.5% and 68.7% of participants had a CDR global score of 0.5 [11]. Participants could take acetylcholinesterase inhibitors, and anti-amyloid antibodies were also permitted — 67 participants started lecanemab, 11 started donanemab, and 1 started aducanumab during the trials [2]. Biomarker data from those participants was screened out of the biomarker analysis [2].
About 85% of participants completed the second year, and roughly 30% had reached the end of the third year when data were locked [2].
What did the results show at the CTAD conference?
The topline results were presented at the Clinical Trials on Alzheimer’s Disease (CTAD) meeting in San Diego on 3 December 2025, during a conference that ran December 1 to 4 [2][3][7]. Jeffrey Cummings of the University of Nevada, Las Vegas, presented the efficacy results [2].
On the primary endpoint, the CDR-SB decline rate was identical in both arms, and the rate matched what would be expected in early Alzheimer’s disease [2]. There was no difference between groups on any secondary measure either: ADCS-ADL-MCI, time to progression from MCI to dementia, MoCA, ADAS-Cog13, MMSE, or ADCOMS [2].
Subgroup analyses were not presented at CTAD. When attendees asked whether any particular group showed a benefit, Johannsen said, “When you look at the curves, they are exactly on top of each other,” adding that any benefiting subgroup would have to be very small [2]. Full results are slated for the AD/PD conferences in March 2026 [3].
Why did biomarkers improve if cognition didn’t?
A CSF substudy — planned at 210 participants — ended up with about 100 participants per group undergoing lumbar puncture at baseline and about 60 per group at week 78 [1][2]. Seven CSF biomarkers showed nominally significant reductions of 10% or less with semaglutide: p-tau181, p-tau217, np-tau181, np-tau205, YKL-40 (a neuroinflammation marker), total tau, and neurogranin (neurodegeneration markers) [2].
Those shifts were small, and none of the seven appeared in plasma [2]. In fact, some plasma markers moved the other way: GFAP rose about 4% on semaglutide in both trials, and NfL rose about 5% in EVOKE+ only — results Cummings said he could not explain [2]. Plasma eMTBR-243, a newer tau marker, was below the limit of detection in two-thirds of participants [2]. Semaglutide is also not expected to reduce amyloid [2].
One plasma marker moved substantially: hs-CRP, a marker of inflammation, fell about 30% on semaglutide in both trials [2]. That suggests the drug quelled peripheral inflammation, but it produced no cognitive benefit [2].
| Measure | Result on semaglutide |
|---|---|
| 7 CSF biomarkers (tau, YKL-40, neurogranin) | Nominally significant reductions of 10% or less [2] |
| Same 7 markers in plasma | No shift [2] |
| Plasma GFAP | Rose ~4% in both trials [2] |
| Plasma NfL | Rose ~5% in EVOKE+ only [2] |
| Plasma hs-CRP | Fell ~30% in both trials [2] |
| CDR-SB and all secondary outcomes | No difference vs placebo [2] |
Was the drug safe in these older participants?
Filip Knop of Novo Nordisk presented pooled safety results [2]. People on semaglutide had more total adverse events than those on placebo, but there were no group differences in serious, severe, or fatal adverse events [2]. More people on semaglutide reduced their dose or discontinued treatment, largely due to mild-to-moderate gastrointestinal issues [2].
More than a third of participants on semaglutide lost weight; about 25% reported nausea, 14% diarrhea, and 12% vomiting [2]. Over the first two years, mean body weight fell 5.8% on semaglutide versus a 0.6% gain on placebo [2]. Novo Nordisk said the safety profile was consistent with previous semaglutide trials, noting more than 37 million patient-years of exposure to date across diverse populations [3].
Why did observational studies suggest semaglutide lowers dementia risk?
Several large studies in people with type 2 diabetes pointed in a more hopeful direction, and they are worth understanding — including their limits.
A target trial emulation by Wang and colleagues used a nationwide US electronic health records database of 1,710,995 patients with type 2 diabetes, 64,267 of whom were on semaglutide, with 3-year follow-up [5][6]. Semaglutide was associated with lower risk of Alzheimer’s-related dementia overall, with the strongest association for vascular dementia:
| Comparison | Overall ADRD hazard ratio | Vascular dementia hazard ratio |
|---|---|---|
| vs insulin | 0.54 (0.49–0.59) | 0.48 (0.39–0.59) |
| vs metformin | 0.67 (0.61–0.74) | 0.55 (0.45–0.68) |
| vs older GLP-1RAs | 0.80 (0.72–0.89) | 0.67 (0.54–0.84) |
Figures from [5][6]. These are associations from retrospective data, not proof of cause and effect.
The study found no evidence of an association with frontotemporal or Lewy body dementia [5][6]. The authors themselves flagged the limitations: a retrospective observational design, possible overdiagnosis or misdiagnosis, and limited follow-up — and they noted that establishing causal relationships would require randomized clinical trials [5][6].
Other data pointed the same way. A pooled analysis of a Danish nationwide health registry plus cardiovascular outcome trial data found people on semaglutide or liraglutide were less than half as likely to develop dementia as non-users [2]. An earlier target trial emulation found semaglutide associated with a 40% to 70% lower risk of first-time Alzheimer’s diagnosis, with a hazard ratio of 0.33 (0.21–0.51) versus insulins and 0.59 (0.37–0.95) versus other GLP-1 receptor agonists [11].
A key limitation was flagged at CTAD: the supporting studies diagnosed Alzheimer’s clinically, not with biomarkers, and type 2 diabetes is tied to dementia risk generally but less strongly to Alzheimer’s neuropathology specifically [2]. In short, the observational work measured who develops dementia, while EVOKE tested whether semaglutide could slow biomarker-confirmed Alzheimer’s that was already underway [2][3]. Those are different questions, and the trials answered only the second one.
Is the GLP-1 and brain health idea finished?
Not necessarily — the question has narrowed rather than closed. Some scientists believe semaglutide could still prevent Alzheimer’s or other dementias if started earlier, before symptomatic disease takes hold [2]. Johannsen defended the decision to run the trials in his keynote: “It was a scientific question that needed an answer, although we didn’t get the answer we had hoped for” [2].
The Alzheimer’s Drug Discovery Foundation’s Howard Fillit noted that existing anti-amyloid drugs slow cognitive decline by around 30%, so therapies targeting other pathways are needed for the remaining 70% — and he framed the completion of two phase 3 trials targeting a non-amyloid, biology-of-aging pathway as momentum toward combination therapy [7]. More than 70% of the Alzheimer’s pipeline is now focused on novel targets [7]. The ADDF had invested nearly $1 million starting in 2011 in a phase 2 liraglutide study by Paul Edison at Imperial College London that preceded these trials [7].
The Alzheimer’s Association noted that an annual pipeline review published in early 2025 counted 182 active clinical trials evaluating 138 novel drugs [4]. Chief Science Officer Maria Carrillo said the field will keep investigating this class of drugs, since different GLP-1s may act differently [4]. CEO Joanne Pike said data from every trial, regardless of outcome, informs the next generation of trials [4].
The Parkinson’s field offers a parallel story. A phase 3 trial of exenatide — 194 people with moderate Parkinson’s at six UK research hospitals, injected for 96 weeks, published in The Lancet on 4 February 2025 — found no evidence the drug slowed disease progression [9]. The Michael J. Fox Foundation said those results “effectively close the door” on exenatide as a Parkinson’s treatment, while noting other GLP-1s may act differently [9]. By contrast, a phase 2 lixisenatide trial published in the New England Journal of Medicine in April 2024 showed the treated group held largely steady while the placebo group progressed — a small difference, and one that leaves conflicting results across GLP-1 Parkinson’s trials [9]. MJFF has funded more than a dozen GLP-1 projects since 2007, totaling roughly $5 million [9].
What does this mean if you take Ozempic, Wegovy, or Rybelsus?
No regulator has approved semaglutide for Alzheimer’s disease or any cognitive indication; its approved uses are type 2 diabetes and chronic weight management [3]. It was approved in the EU for stroke prevention earlier in 2025 [11]. Novo Nordisk’s Martin Holst Lange said the evidence supporting semaglutide continues to benefit people with type 2 diabetes, obesity, and related comorbidities [3][11].
Nothing in these results is a reason to start, stop, or change a GLP-1 medication on your own — that is a conversation for your clinician. For people with early Alzheimer’s, approved treatments do exist, and the Alzheimer’s Association urges people to talk with their doctor about those options and about clinical trial participation, including through its TrialMatch service [4]. Alzheimer’s disease accounts for 60 to 80% of all dementia cases worldwide [3].
Sources
- PubMed. Evoke and evoke+: design of two large-scale, double-blind, placebo-controlled, phase 3 studies evaluating efficacy, safety, and tolerability of semaglutide in early-stage symptomatic Alzheimer’s disease.
- Alzforum. Semaglutide does not treat Alzheimer’s. Could it prevent dementia? December 16, 2025.
- Novo Nordisk A/S. Evoke phase 3 trials did not demonstrate a statistically significant reduction in Alzheimer’s disease progression. November 24, 2025.
- Alzheimer’s Association. Statement on oral semaglutide phase 3 topline data release. 2025.
- PubMed Central. Associations of semaglutide with Alzheimer’s disease-related dementias in patients with type 2 diabetes: a real-world target trial emulation study.
- Neurology Advisor. Semaglutide use tied to lower dementia risk in patients with type 2 diabetes.
- Alzheimer’s Drug Discovery Foundation. Readout of phase 3 semaglutide trials marks critical moment in Alzheimer’s research and suggests potential for combination therapies. November 24, 2025.
- The Lancet. Exenatide once a week versus placebo as a potential disease-modifying treatment for people with Parkinson’s disease in the UK: a phase 3, multicentre, double-blind, parallel-group, randomised, placebo-controlled trial.
- Michael J. Fox Foundation. Exenatide, a GLP-1 drug, shows no impact on Parkinson’s symptoms. February 5, 2025.
- The Lancet. Efficacy and safety of oral semaglutide 14 mg (flexible dose) in early-stage symptomatic Alzheimer’s disease (evoke and evoke+): two phase 3, randomised, placebo-controlled trials.
- NeurologyLive. GLP-1 semaglutide fails to outperform placebo in phase 3 EVOKE trial of Alzheimer disease. November 24, 2025.
- GoodRx. Can semaglutide prevent dementia? What the research says.
Frequently asked questions
Did semaglutide slow Alzheimer's in the EVOKE trials at all?
No. Semaglutide did not outperform placebo on the primary endpoint, CDR-SB change from baseline to week 104, or on any secondary cognitive or functional measure. The semaglutide and placebo curves were indistinguishable across the three-year trials, and the results were described as flat-out negative.
What dose of oral semaglutide was tested, and for how long?
Participants took once-daily oral semaglutide, escalating from 3 mg for weeks 0 to 4, to 7 mg for weeks 4 to 8, then 14 mg from week 8 onward — the same regimen used in type 2 diabetes. Treatment was planned for 156 weeks: a 104-week main phase plus a 52-week extension.
How many people were in EVOKE and EVOKE+, and who qualified?
The trials randomized 3,808 people — 1,855 in EVOKE and 1,953 in EVOKE+ — after screening nearly 10,000 across 40 countries. Participants were 55 to 85 years old with mild cognitive impairment or mild dementia due to Alzheimer's, confirmed by amyloid PET or CSF testing.
What is the difference between EVOKE and EVOKE+?
EVOKE excluded people with cerebrovascular or small vessel pathology on CT or MRI; EVOKE+ included them. In practice it barely mattered: fewer than 3% of EVOKE+ enrollees had small vessel pathology — only 54 people — making the two populations essentially identical.
Why did biomarkers improve if cognition didn't?
Seven CSF biomarkers fell by 10% or less — a nominally significant change in a small substudy of roughly 100 people per group at baseline. None of those markers shifted in plasma, some plasma markers rose instead, and no change in fluid markers translated into slower decline on any clinical measure.
Does Ozempic or Wegovy lower dementia risk if you don't have Alzheimer's?
That question was not tested in EVOKE, which studied people who already had early symptomatic Alzheimer's. Observational studies in type 2 diabetes have found associations with lower dementia risk, but those are retrospective data, not proof of prevention, and the authors themselves said randomized trials are needed to establish causality.
Why did the big observational studies suggest a benefit?
Those studies measured who develops dementia over time in routine health records, and several found semaglutide associated with lower risk. But they diagnosed Alzheimer's clinically rather than with biomarkers, and type 2 diabetes is tied to dementia risk generally more than to Alzheimer's pathology specifically — a distinction emphasized at CTAD.
Was the one-year extension cancelled, and why?
Yes. Novo Nordisk announced on 24 November 2025 that the 1-year extension period in EVOKE and EVOKE+ would be discontinued based on the efficacy results observed in the overall study population. The main 104-week analysis showed no slowing of progression.
This article summarizes FDA labeling, published research and company information current as of September 14, 2026. It is not medical advice and does not replace a conversation with your own healthcare provider. How we research and verify.