GLP-1s for Alzheimer's & Dementia Prevention: Patient Guide

Headlines about “Ozempic for Alzheimer's” have outpaced the evidence. As of May 2026, no GLP-1 medication is FDA-approved for the prevention or treatment of Alzheimer's or any dementia. The first phase 3 trials in any GLP-1 for AD — evoke and evoke+, testing oral semaglutide 14 mg in ~3,800 participants with early symptomatic AD^[2] — reported topline results in November 2025 and did not meet their primary cognitive endpoint (CDR-SB at week 104), despite some improvement in Alzheimer's-related biomarkers (Novo Nordisk topline, November 2025; Cummings et al. 2026, Lancet). The strongest finished randomized trial is ELAD, a 204-participant phase 2b of liraglutide that missed its FDG-PET primary but slowed brain-volume loss and produced ~18% slowing of ADAS-Exec decline (Edison 2026 Nat Med^[1]). Most of the remaining human signal is observational: TTEs of US claims data show semaglutide and tirzepatide users with T2D have 30–70% lower hazards of first-time AD or dementia diagnoses than matched comparators^[4][5]. These are hypothesis-generating, not causal. This patient-facing guide walks through the mechanism, what each trial shows, the APOE-e4 question, and when a conversation with your neurologist or obesity-medicine clinician is appropriate.

The honest summary

• Not FDA-approved. Any GLP-1 tied to cognition is off-label (in a patient who qualifies for T2D, obesity, MASH, or CV risk reduction) or trial-only.
• ELAD. 204 participants with mild-to- moderate AD; primary FDG-PET endpoint not met; liraglutide slowed brain-volume loss and reduced ADAS-Exec decline by ~18% (Edison 2026^[1]).
• evoke + evoke+. First phase 3 trials of any GLP-1 in AD: ~3,800 participants on oral semaglutide 14 mg vs placebo, primary CDR-SB at week 104^[2]. Topline reported November 2025: primary endpoint missed — no significant slowing of cognitive decline vs placebo, despite some biomarker improvement (Novo Nordisk topline, November 2025; Cummings et al. 2026, Lancet).
• REWIND substudy. Dulaglutide vs placebo in 8,828 T2D patients, 5.4 yr median — HR ~0.86 for substantive cognitive impairment^[3].
• Observational TTEs. Wang 2024 / 2025 on TriNetX: HR 0.30–0.60 for first-time AD diagnosis in T2D semaglutide users vs comparators^[4][5].
• Safety in older adults. Sabbagh 2025 pooled 8,464 aged 65+: SAEs 14.9% vs 14.5% placebo; no signal for falls, fractures, or cognitive worsening^[6].

The proposed mechanism

GLP-1 receptors are expressed widely in the CNS — hippocampal and cortical neurons, microglia, and choroid plexus. Four candidate mechanisms could plausibly slow cognitive decline: (1) direct neuronal signaling (PKA/CREB pathways implicated in synaptic plasticity; mouse models show reduced amyloid-beta and tau); (2) neuroinflammation reduction (reduced microglial TNF-alpha, IL-1-beta, IL-6 in preclinical models; ELAD secondary outcomes consistent^[1]); (3) brain insulin resistance (the “type 3 diabetes” framing — whether GLP-1 agonists restore central insulin sensitivity in humans is unsettled); and (4) cardiometabolic/vascular contribution (SELECT, LEADER, SUSTAIN-6 CV benefit; Wang 2025 showed stronger effects on vascular and mixed dementias than pure AD^[5]).

The randomized evidence so far

ELAD (Edison 2026 Nat Med). Evaluating Liraglutide in Alzheimer's Disease randomized 204 participants with mild-to-moderate AD 1:1 to subcutaneous liraglutide (titrated to 1.8 mg daily) or placebo for 52 weeks across 24 UK centers. Primary FDG-PET endpoint was not met. Liraglutide slowed brain-volume loss in key regions on MRI and reduced ADAS-Exec decline by ~18% vs placebo, without a statistically significant CDR-SB effect at this sample size^[1]. Proof-of-concept, not approval-grade.

evoke + evoke+ (negative primary readout). Sponsored by Novo Nordisk, these are the first phase 3 randomized placebo-controlled trials of any GLP-1 in AD. They tested oral semaglutide 14 mg daily (the molecule in Rybelsus / Ozempic / Wegovy) vs placebo in ~1,840 and ~1,960 participants with early symptomatic AD (MCI due to AD or mild AD dementia) for roughly two years. Per Scheltens 2026^[2]: mean age 72, ~49% APOE-e4 carriers, ~56% on background symptomatic AD therapy. Primary endpoint: CDR-SB at week 104. Topline results, announced November 2025, were negative — oral semaglutide did not significantly slow cognitive decline vs placebo, despite some improvement in Alzheimer's-related biomarkers. Full results were presented at CTAD (late 2025) and AD/PD 2026 (Novo Nordisk topline, November 2025; Cummings et al. 2026, Lancet).

REWIND cognitive substudy (Cukierman-Yaffe 2020). The CV-outcomes trial of dulaglutide 1.5 mg weekly in 9,901 adults with T2D; 8,828 had cognitive assessments. Over a median 5.4 years, dulaglutide showed a smaller decline on modified MoCA, with HR ~0.86 (95% CI 0.79–0.95) for substantive cognitive impairment after adjustment^[3] — the original randomized cognitive signal and justification for the phase 3 AD program.

The observational signal: target-trial emulations

Alongside the limited randomized trial evidence — now including the negative November-2025 evoke readout — most of the remaining public discussion is observational. The dominant method is the target trial emulation (TTE), using real-world EHR data (primarily the TriNetX network: 60+ US healthcare organizations, 1M+ T2D patients) to simulate a randomized design. Wang 2024^[4] emulated seven target trials comparing semaglutide to insulin, metformin, DPP-4i, SGLT2i, and other antidiabetics for first-time AD diagnosis at 3 years; hazard ratios ranged 0.30–0.60 (largest effect vs insulin). The 2025 extension to broader ADRD showed similar magnitudes with stronger effects on vascular and mixed dementias^[5]. Younis 2026^[7] and da Silva 2026^[8] ran head-to-heads: both GLP-1 agents reduced dementia incidence vs SGLT2i over 3 years, with tirzepatide modestly favored over semaglutide (HR ~0.85). These are hypothesis- generating — residual confounding by indication and differential diagnosis-coding cannot be ruled out.

How the observational magnitudes compare

Observational dementia magnitudes (40–70%) look enormous next to the randomized REWIND signal (14%). Some of the gap is real biology; some is confounding. The negative evoke readout (November 2025) suggests confounding accounts for much of it — the randomized cognitive benefit predicted by the observational data did not materialize.

APOE-e4 and the T2D pathway

APOE-e4 is the strongest common genetic risk factor for sporadic AD (~3-fold risk with one copy, 8 to 12-fold with two) and interacts with the anti-amyloid antibody class (higher ARIA rates). For GLP-1 agonists, an APOE-e4 interaction is an open question — ELAD was not powered, evoke enrolled ~49% e4 carriers with a pre-specified subgroup^[2]. APOE genotyping should not be ordered to guide off-label GLP-1 prescribing. And note that the strongest observational signals all come from people with T2D — itself a dementia risk factor (PAF ~5–10%) via vascular injury and shared upstream factors. So the signal could reflect (1) a direct CNS effect, (2) glycemic-control benefit, (3) weight-loss cardiometabolic benefit, or (4) residual confounding. Whether (1) is true on its own in nondiabetic adults is the question O'Mara 2026^[10] examined and evoke was built to answer — and evoke's negative primary readout (November 2025) argues against a large stand-alone cognitive benefit of oral semaglutide in early symptomatic AD, tempering the observational optimism.

When this conversation makes sense

Off-label GLP-1 use for cognition alone — in a patient with no other indication — is not appropriate on current evidence. The drug has real side effects (GI in ~40%, rare pancreatitis and gallbladder risk), and the first randomized cognitive readout (evoke / evoke+, November 2025) was negative for cognition. The conversation does make sense in four scenarios:

• T2D plus MCI. A patient who meets the FDA-approved T2D indication and has documented MCI is a reasonable candidate to discuss semaglutide or tirzepatide with an endocrinologist, with cognition as a secondary factor.
• Obesity plus family history of AD. Meeting BMI 30+, or 27+ with a weight-related comorbidity, plus meaningful family history makes the observational signal one factor in the decision.
• CV risk plus cognitive concern. SELECT established CV benefit for semaglutide 2.4 mg in obesity-without-diabetes; vascular dementia shares risk factors. The strongest honest framing for now.
• Clinical trial enrollment. With evoke and evoke+ now completed, patients with early symptomatic AD can ask their neurologist about other ongoing trials — ClinicalTrials.gov is the canonical search.

A patient action plan

1. Define the question. GLP-1 for cognition alone, or for cognition plus an FDA-approved indication (T2D, obesity, MASH, CV risk)? The answer changes the conversation entirely.
2. See a neurologist first if you have documented MCI or an unworked-up memory complaint. Workup typically includes MoCA, MRI brain, labs (B12, thyroid), and for AD specifically CSF biomarkers, amyloid PET, or plasma p-tau217.
3. Pair with obesity medicine if you meet BMI criteria — a board-certified obesity-medicine clinician or endocrinologist is the right prescriber. See our GLP-1 provider directory .
4. Be honest about the evidence. The observational signal is genuine, but the first randomized phase 3 readout (evoke / evoke+, November 2025) was negative for cognition. The honest position is now “observational signal not confirmed by the randomized trial.”
5. Manage modifiable risk now. The 2024 Lancet Commission identified 14 modifiable risk factors (inactivity, untreated hearing loss, obesity, BP, diabetes, smoking, alcohol, depression, social isolation, air pollution, low education, TBI, untreated visual loss, high LDL) accounting for ~45% of dementia risk — actionable today regardless of GLP-1.

Safety in older adults

Sabbagh 2025^[6] pooled 14 phase 3 semaglutide trials covering 8,464 participants aged 65+ (1,463 aged 75+). SAEs 14.9% vs 14.5% placebo. GI events higher on drug (43% vs 25%) but rarely led to discontinuation. No signal for falls, fractures, or cognitive worsening. Age alone is not a barrier when an FDA-approved indication is met.

Bottom line

• No GLP-1 is FDA-approved for AD or dementia. Any use is off-label or trial-only.
• ELAD missed FDG-PET primary; ~18% ADAS-Exec decline slowing^[1].
• evoke / evoke+ reported negative topline in November 2025 — oral semaglutide did not significantly slow cognitive decline vs placebo (Novo Nordisk; Cummings et al. 2026, Lancet).
• Observational TTEs show 30–70% lower hazards in T2D users — hypothesis-generating, not causal^[4][5].
• Off-label use for cognition alone is not appropriate. Conversation makes sense when an FDA-approved indication coexists.

Related research and tools

• Top 10 PubMed studies on GLP-1 drugs for dementia and Alzheimer's (2026) — the ranked study list with PMIDs
• GLP-1 for fatty liver & MASH: the patient action plan — the parallel patient-facing walkthrough
• GLP-1s, depression, and suicidality — the parallel CNS-effect review
• Semaglutide drug page — the molecule in Ozempic, Wegovy, Rybelsus, and evoke
• Tirzepatide drug page — the dual GIP/GLP-1 molecule in Mounjaro and Zepbound
• GLP-1 provider directory — telehealth options when an FDA-approved indication is met