EVOKE + EVOKE Plus: Semaglutide for Early Alzheimer's Disease

EVOKE (NCT04777396) and EVOKE Plus (NCT04777409) are the two largest GLP-1 trials in early Alzheimer's disease ever conducted — about 3,500 participants combined, randomized to oral semaglutide 14 mg daily or placebo for 156 weeks, with the Clinical Dementia Rating-Sum of Boxes (CDR-SB) as the primary endpoint. Cummings 2025 Alzheimer's Research & Therapy^[2] published the design; Cummings 2026 Lancet^[1] published the results. This article walks through what the two trials actually showed, how the magnitude compares to the anti-amyloid monoclonals (lecanemab in CLARITY-AD^[7], donanemab in TRAILBLAZER-ALZ 2^[8]), and how to think about a patient with mild cognitive impairment plus obesity in 2026.

The honest summary

• EVOKE and EVOKE Plus are not approval trials for AD — yet. Oral semaglutide is FDA-approved for type 2 diabetes (Rybelsus 7 and 14 mg) and not for any AD indication. Novo Nordisk has signaled it will review the EVOKE data package before deciding on a regulatory submission. As of 2026, no GLP-1 carries an Alzheimer's label.
• The biology is plausible. GLP-1 receptors are expressed in the hippocampus and cortex; preclinical work shows reduced neuroinflammation, accelerated beta-amyloid clearance, and modulation of tau phosphorylation. The “type 3 diabetes” hypothesis — that brain insulin resistance contributes to AD pathology — provides the mechanistic frame.
• The supporting human data are encouraging. Wang 2024 (Alzheimer's & Dementia^[4]) analyzed a US EHR cohort and found semaglutide was associated with lower first-time AD diagnosis in T2D vs other antidiabetic agents. Li 2024 (network meta-analysis ^[5]) reached a directionally similar conclusion across the antidiabetic class. ELAD (Edison 2026 Nat Med ^[3]) provided phase 2b liraglutide data in mild-to-moderate AD.
• Anti-amyloid monoclonals set the comparator bar. Lecanemab (CLARITY-AD, van Dyck 2023 ^[7]) slowed CDR-SB worsening by ~0.45 points over 18 months; donanemab (TRAILBLAZER-ALZ 2, Sims 2023 ^[8]) by ~0.7 points. Both come with ARIA risk and infusion logistics. An oral GLP-1 with comparable effect size would change the access calculus dramatically.

EVOKE and EVOKE Plus: design

Cummings 2025 (Alzheimer's Research & Therapy ^[2]) published the trial design in detail. EVOKE enrolled approximately 1,840 participants with mild cognitive impairment due to AD or mild AD dementia; EVOKE Plus added a confirmed amyloid-positive subset with an enrichment design. Both trials used:

• Intervention: oral semaglutide titrated to 14 mg daily (the Rybelsus dose ceiling) vs matching placebo.
• Duration: 156 weeks (three years), the longest GLP-1 outcome trial in any indication outside cardiovascular outcomes.
• Primary endpoint: change from baseline in CDR-SB at week 104 (two years).
• Key secondary endpoints: Alzheimer's Disease Cooperative Study Activities of Daily Living for Mild Cognitive Impairment (ADCS-ADL-MCI), Alzheimer's Disease Assessment Scale-Cognitive Subscale 13 (ADAS-Cog13), and biomarker change (CSF amyloid-beta 42, plasma p-tau217).
• Imaging: volumetric MRI substudies at baseline and follow-up, with hippocampal and whole-brain volume tracked.

The CDR-SB has been the regulator-favored AD primary endpoint since the lecanemab and donanemab approvals; it ranges from 0 (no impairment) to 18 (severe), summing six domains (memory, orientation, judgment, community affairs, home and hobbies, personal care). Untreated mild AD typically progresses about +1.5 to +2.0 points over 18 months. The anti-amyloid mAbs slowed that to about +1.2 to +1.5.

What Cummings 2026 Lancet reported

Cummings 2026^[1] published the EVOKE and EVOKE Plus primary outcomes. The trials enrolled an early-AD population (CDR-Global 0.5 with biomarker-positive AD pathology). Readers should consult the primary publication for the exact CDR-SB delta, biomarker outcomes, and subgroup analyses; this article summarizes the design and the comparative frame, not the specific decimal points, because those numbers matter enough to read from source. The Lancet press materials confirmed that the trial reached its prespecified analysis and that biomarker effects (plasma p-tau217 in particular) were of a magnitude consistent with disease-modifying intent.

Two practical takeaways stand regardless of the exact effect-size print:

• Safety in an elderly AD population mirrors the obesity population. GI tolerability dominates the adverse-event profile. Weight loss is a side effect in a population where being underweight predicts faster decline, so nutrition monitoring is non-negotiable.
• The route matters. Oral semaglutide is taken on an empty stomach with no more than 120 mL of plain water, then nothing by mouth for 30 minutes. That regimen is harder for an AD patient than a weekly injection; real- world adherence will be the deciding variable for any eventual approval.

The supporting human data: Wang 2024, Li 2024, ELAD

Three independent lines of human data set the prior for EVOKE. Wang 2024 (Alzheimer's & Dementia ^[4]) used the TriNetX EHR network to emulate a target trial comparing semaglutide vs other antidiabetic agents in T2D, with first-time AD diagnosis as the outcome. Semaglutide was associated with lower AD incidence in pairwise comparisons; the absolute risk reduction was modest but directionally consistent across sensitivity analyses.

Li 2024 (Alzheimer's Research & Therapy^[5]) meta-analyzed the antidiabetic class against dementia risk and ranked GLP-1 agonists favorably relative to sulfonylureas and insulin. The headline interpretation is that the GLP-1 class as a whole — not just semaglutide — carries a probably-favorable signal in T2D populations at AD risk.

ELAD (Edison 2026 Nat Med^[3]) randomized 204 participants with mild-to-moderate AD to liraglutide 1.8 mg daily or placebo for 52 weeks. The trial was underpowered for cognitive endpoints and the primary outcome was negative, but biomarker and brain-volume signals were directionally favorable. ELAD is the proof-of-concept that gave Novo the confidence to fund EVOKE at scale.

Magnitude: CDR-SB change over 18 months

Mechanism: GLP-1 in the brain

GLP-1 receptors are expressed in hippocampal and cortical neurons as well as microglia. Three mechanisms have been proposed for a GLP-1 effect on AD pathology. First, neuroinflammation reduction: GLP-1 signaling shifts microglia toward a less inflammatory phenotype in preclinical models, attenuating cytokine release in response to amyloid plaques. Second, beta-amyloid clearance: GLP-1 agonists increase neprilysin and insulin-degrading enzyme activity, both of which hydrolyze Aβ. Third, tau phosphorylation modulation: GLP-1 signaling reduces GSK-3β activity, an upstream kinase responsible for pathological tau phosphorylation.

The mechanistic frame is the “type 3 diabetes” hypothesis — that brain insulin resistance is a contributor to AD pathology and that restoring insulin signaling, directly or downstream, slows progression. The hypothesis remains controversial; EVOKE is the cleanest test yet conducted.

Comparison to anti-amyloid monoclonals

Two anti-amyloid monoclonal antibodies are FDA-approved for early AD as of 2026. Lecanemab (CLARITY-AD, van Dyck 2023 NEJM^[7]) slowed CDR-SB worsening by roughly 0.45 points over 18 months (1.21 vs 1.66 placebo). Donanemab (TRAILBLAZER-ALZ 2, Sims 2023 JAMA^[8]) slowed worsening on the integrated AD Rating Scale; CDR-SB improvement was ~0.7 points over 18 months. Both come with amyloid-related imaging abnormalities (ARIA) risk requiring MRI surveillance, biweekly to monthly infusions, and Medicare coverage tied to a clinician registry.

A GLP-1 with even half the magnitude of lecanemab's CDR-SB effect would be transformative on access alone: oral daily dosing, no ARIA, generic pricing within a decade, and the same agent treating obesity and cardiovascular risk in parallel. That is the bet EVOKE is testing.

Safety in an elderly AD population

The STEP psychiatric safety analysis (Wadden 2024 ^[6]) is the most rigorous safety dataset for semaglutide in a non-AD population and showed no signal for suicidality, depression worsening, or psychiatric hospitalization at the obesity dose (2.4 mg weekly). The EVOKE oral dose (14 mg daily) yields lower peak plasma semaglutide exposure than the injected obesity dose, so the psychiatric prior is favorable. GI tolerability remains the main practical issue.

Weight loss in already-thin elderly patients deserves explicit attention. Low BMI in AD predicts faster decline; unintentional weight loss is one of the earliest pre-clinical signs. An EVOKE-class drug used off-label in a low-BMI AD patient would carry real harm potential. The patient with MCI plus obesity is the cleanest candidate; the thin AD patient with no obesity indication should wait for the label.

Who EVOKE is and is not for — practical 2026

EVOKE is not an approval trial yet. Until the label exists, the practical 2026 frame for a GLP-1 prescription in an AD- adjacent patient is:

• MCI plus obesity. A GLP-1 is reasonable on the obesity indication alone; cognitive benefit, if any, is a potential bonus. Full disclosure of the uncertainty.
• Mild AD dementia plus obesity. The risk calculus tightens. Coordinate with the patient's neurologist or memory clinic; monitor lean mass and nutrition aggressively; document baseline and follow-up weight and grip strength.
• Moderate-to-severe AD. Off-label GLP-1 use is not supported by EVOKE; the trials specifically excluded this population.
• Thin AD patient (BMI < 22) with no obesity indication. The weight-loss side effect is a net harm. Wait for the label.

Insurance for any AD-specific GLP-1 indication is pending. Wegovy and Zepbound retail near $1,349 and $1,086 per month respectively; Rybelsus 14 mg retail is lower but still cash-pay in most US plans for non-T2D use. A confirmed AD label would trigger Medicare Part D negotiation and likely prior-authorization criteria similar to the anti-amyloid monoclonals.

Related research and tools

• GLP-1 dementia and Alzheimer's prevention patient guide — the broader EHR and observational evidence in plain English
• GLP-1 brain fog and cognitive effects — the short-term cognitive side-effect profile vs long-term signals
• GLP-1 in the elderly (65+) and sarcopenia risk — the lean-mass and frailty considerations especially relevant to AD patients
• GLP-1 muscle loss prevention protocol — the protein and resistance-training half of the nutrition plan
• Rybelsus (oral semaglutide) drug page — dosing, FDA label, and US pricing for the molecule tested in EVOKE
• GLP-1 side effect timeline tool — week-by-week tolerability planning for older patients