Top 10 PubMed Studies on GLP-1 Drugs for Dementia and Alzheimer's (2026)

ELAD is the definitive randomized trial of a GLP-1 agonist in Alzheimer's disease published to date. It enrolled 204 participants with mild-to-moderate Alzheimer's and randomized them to subcutaneous liraglutide (up to 1.8 mg daily) or placebo for 52 weeks at 24 UK centers. The primary FDG-PET endpoint was not met, but liraglutide slowed brain-volume loss in key regions and reduced decline on the ADAS-Exec composite by roughly 18 percent versus placebo. Established proof-of-concept for the class and informed the design of the larger evoke and evoke+ semaglutide trials.

PMID 41326666 ↗DOI 10.1038/s41591-025-04106-7 ↗

evoke and evoke+ are the first phase 3 randomized placebo-controlled trials of any GLP-1 agonist for Alzheimer's disease, testing oral semaglutide 14 mg daily versus placebo in roughly 1,840 and 1,960 participants respectively with early symptomatic Alzheimer's (MCI due to AD or mild AD dementia). This baseline-characteristics paper describes the enrolled population — mean age 72, 49 percent APOE-e4 carriers, 56 percent on background symptomatic AD therapy. Topline efficacy results on the CDR-SB primary endpoint are expected in late 2026 and will define the class for the next decade.

PMID 41522368 ↗NCT04777396 ↗DOI 10.1002/trc2.70200 ↗

This prespecified exploratory analysis of the REWIND cardiovascular outcomes trial (9,901 type 2 diabetes patients followed a median 5.4 years on dulaglutide 1.5 mg weekly versus placebo) was the first randomized signal that a GLP-1 might slow cognitive decline. Incident substantial cognitive impairment occurred in 4.05 percent on dulaglutide versus 4.35 percent on placebo (HR 0.86, 95 percent CI 0.79-0.95, p=0.0018) after adjustment for baseline cognitive score. The absolute effect is small but statistically robust in the largest GLP-1 cognitive randomized data set to date.

PMID 32562683 ↗NCT01394952 ↗DOI 10.1016/S1474-4422(20)30173-3 ↗

Wang and colleagues at Case Western Reserve used the TriNetX network (more than 1 million type 2 diabetes patients across 60 US healthcare organizations) to emulate seven target trials comparing semaglutide to insulin, metformin, DPP-4 inhibitors, SGLT2 inhibitors, and other antidiabetics. Semaglutide was associated with a 40 to 70 percent lower hazard of first-time Alzheimer's diagnosis at 3 years (HR ranging from 0.30 to 0.60 depending on comparator), with the largest effect versus insulin. Results are hypothesis-generating, not causal, and motivated the evoke phase 3 program.

PMID 39445596 ↗DOI 10.1002/alz.14313 ↗

The Wang group extended their 2024 analysis to the broader Alzheimer's disease-related dementias (ADRD) outcome — which includes vascular dementia, frontotemporal dementia, and Lewy body dementia. Across the TriNetX network, semaglutide users had hazard ratios for first-time ADRD ranging from 0.40 to 0.70 versus comparators, with consistent effects across age, sex, and obesity strata. The signal is stronger for vascular and mixed dementias than for pure Alzheimer's, supporting a vascular-mechanism contribution alongside any direct amyloid effect.

PMID 40552638 ↗DOI 10.1177/13872877251351329 ↗

Sabbagh and colleagues pooled safety data from 14 phase 3 semaglutide trials covering 8,464 participants aged 65 and older — including 1,463 aged 75-plus — to characterize tolerability in the population that the evoke trials are enrolling. Serious adverse events occurred in 14.9 percent on semaglutide versus 14.5 percent on placebo; gastrointestinal events were higher on semaglutide (43 versus 25 percent) but rarely led to discontinuation. No signal for falls, fractures, or cognitive worsening. Provides the safety foundation for treating an older, frailer Alzheimer's population at the 14 mg oral dose.

PMID 40337158 ↗DOI 10.1002/trc2.70076 ↗

Younis and colleagues at Liverpool used the TriNetX global network to emulate three head-to-head target trials in type 2 diabetes: tirzepatide versus SGLT2 inhibitors, semaglutide versus SGLT2 inhibitors, and tirzepatide versus semaglutide. Both GLP-1 agents reduced all-cause dementia incidence versus SGLT2 inhibitors over a 3-year follow-up. The tirzepatide-versus-semaglutide head-to-head was numerically in favor of tirzepatide but not statistically significant. Importantly the SGLT2 comparator was itself dementia-protective in prior work, so the absolute size of the GLP-1 advantage may be larger versus other comparators.

PMID 41544899 ↗DOI 10.1016/j.diabres.2026.113083 ↗

da Silva and colleagues used the TriNetX network to compare tirzepatide and semaglutide head-to-head for cognitive outcomes in propensity-matched type 2 diabetes patients. Across roughly 50,000 matched pairs followed for up to 3 years, tirzepatide was associated with a modestly lower composite hazard of MCI, dementia, or Alzheimer's diagnosis than semaglutide (HR approximately 0.85, 95 percent CI overlap minimal). Results are consistent with the broader observation that the dual GIP/GLP-1 mechanism delivers slightly stronger metabolic and possibly cognitive effects than GLP-1 alone, though residual confounding cannot be excluded.

PMID 41825212 ↗DOI 10.1016/j.jdiacomp.2026.109306 ↗

Lee and colleagues focused on a high-risk subpopulation — patients with type 2 diabetes and chronic kidney disease, who carry roughly twice the baseline dementia risk of the general T2D population. Using the TriNetX network, GLP-1 users with diabetic CKD had a 35 to 45 percent lower hazard of all-cause dementia and Alzheimer's disease than non-GLP-1 users matched on age, sex, eGFR, albuminuria, and HbA1c. The effect was stronger in patients with more advanced CKD (eGFR <45), consistent with the broader hypothesis that the GLP-1 cognitive benefit is partly vascular and partly mediated by improved kidney-brain axis signaling.

PMID 41697144 ↗DOI 10.1093/ndt/gfag032 ↗

O'Mara and colleagues pooled all available randomized GLP-1 trials in nondiabetic MCI or Alzheimer's patients — a small evidence base dominated by ELAD and a handful of single-center exenatide and lixisenatide trials. Across roughly 350 randomized patients, GLP-1 therapy produced a small but statistically significant improvement in ADAS-Cog and executive-function composite scores versus placebo (standardized mean difference approximately 0.25). The authors emphasize that the effect size is modest and the evidence base will be transformed by the evoke and evoke+ readouts. Useful as a quantitative pre-evoke baseline.

PMID 41524953 ↗DOI 10.1007/s10072-025-08602-z ↗