Diabetes drugs show early brain signals but no proven benefit in Alzheimer’s, review finds

Diabetes medications that have captured global attention for their effects on weight and the heart are now being tested against one of medicine’s most feared diseases. A systematic review published in Discover Neuroscience on 27 October 2026 examined whether glucagon-like peptide-1 receptor agonists (GLP-1 RAs) and sodium-glucose co-transporter-2 inhibitors (SGLT-2i) — the drug classes behind liraglutide, dulaglutide, exenatide and dapagliflozin — can protect the brain in Alzheimer’s disease and cognitive impairment. The verdict from the randomized evidence available so far is cautious: intriguing biological signals, but no consistent clinical cognitive benefit.

The review, led by Marwa A. Dewer and colleagues and registered in PROSPERO (CRD420251083128), followed PRISMA guidelines and searched PubMed, Embase, Cochrane Central, ClinicalTrials.gov and the WHO ICTRP from inception to 30 August 2025, with supplementary searches of Google Scholar and SpringerLink. After screening 8,714 records, the team identified nine eligible publications: seven small randomized controlled trials totalling 317 participants and two large post hoc analyses of cardiovascular outcome trials, REWIND and EXSCEL, covering 24,653 participants — 24,970 people in all.

The rationale for testing these drugs rests on the deep metabolic entanglement of Alzheimer’s disease and type 2 diabetes. Individuals with poorly controlled or severe type 2 diabetes face a markedly increased risk of Alzheimer’s disease and other dementias, especially vascular dementia, compared with normoglycemic individuals. Epidemiological studies and meta-analyses have consistently shown that people with type 2 diabetes have an approximately 50% higher risk of developing dementia, including Alzheimer’s, than those without diabetes. Globally, more than 537 million adults live with diabetes, and dementia affects over 57 million people, with Alzheimer’s accounting for roughly 60–70% of those cases.

The mechanisms linking the two conditions are thought to include diabetes-induced microvascular and macrovascular damage that impairs cerebral blood vessel integrity, insulin resistance that disrupts cerebral energy metabolism, and promotion of the accumulation of amyloid-β and hyperphosphorylated tau, the hallmark proteins of Alzheimer’s pathology. Some researchers have proposed Alzheimer’s disease as a brain-specific insulin-resistant state, sometimes called ‘type 3 diabetes’, although the review’s authors stress this remains a conceptual framework rather than a formal diagnosis. Preclinical work has shown that GLP-1 RAs exert anti-inflammatory and antioxidant effects, reduce amyloid-β aggregation and tau hyperphosphorylation, decrease neuronal apoptosis and neurotoxicity, and promote neurogenesis in animal models — findings that provide biological plausibility but do not establish cognitive efficacy in humans.

What did the trials actually show? For GLP-1 RAs, the picture is one of exploratory signals rather than proven treatment effects. The post hoc REWIND analysis by Cukierman-Yaffe and colleagues suggested a reduction in incident cognitive impairment with dulaglutide over 5.4 years, but cognition was not a primary endpoint of the parent cardiovascular trial, and no consistent changes were detected on standard cognitive tests. The authors of the review emphasize that such dementia-related outcomes were derived from post hoc analyses and warrant considerable interpretive caution.

Liraglutide fared best among the GLP-1 RAs in the smaller trials. Cheng and colleagues reported increased odour-induced hippocampal activation on functional MRI alongside improvements in delayed memory, attention and executive function compared with dapagliflozin and acarbose. Vadini and colleagues observed improvements in short-term memory and composite memory scores despite comparable weight loss between treatment groups, hinting at effects beyond metabolic improvement. However, Gejl and colleagues found no significant difference in total cognition between liraglutide and placebo, although cerebral glucose metabolism appeared preserved despite stable cognition and unchanged amyloid burden. The review’s authors note that these imaging and metabolic findings represent surrogate or mechanistic signals rather than direct evidence of clinically meaningful cognitive improvement, and that all three liraglutide trials were small and of short duration.

Exenatide told a more sobering story on the clinical side. In the randomized study by Mullins and colleagues, exenatide treatment in patients with Alzheimer’s produced no significant cognitive improvements despite a reduction in neuronal extracellular vesicle amyloid-β42 levels; that trial was terminated early by the sponsor, shrinking the sample and reducing statistical power. Cas and colleagues found no improvement in the primary cognitive outcome, ADAS-Cog11, among individuals with mild cognitive impairment treated with exenatide, and even noted a possible sex-specific adverse effect in women — an exploratory, hypothesis-generating finding from a small subgroup. Meanwhile, Koychev and colleagues demonstrated significant modulation of circulating inflammatory and vascular-related proteins, including reductions in ficolin-2 and plasminogen activator inhibitor-1 and increases in soluble vascular cell adhesion molecule-1, without assessing cognition at all. Together, these studies illustrate the review’s central distinction: exploratory biomarker signals do not equal clinically meaningful cognitive outcomes.

Evidence for SGLT-2 inhibitors was considerably thinner. Zhao and colleagues reported improved Mini-Mental State Examination scores in individuals with type 2 diabetes and mild cognitive impairment receiving dapagliflozin, but a concurrent cognitive behavioural training intervention made it impossible to attribute the improvement to the drug alone. In a small exploratory pilot trial, Burns and colleagues found that dapagliflozin was well tolerated in probable Alzheimer’s disease and associated with favourable trends in cognitive performance and brain metabolic activity, including increased brain glutathione suggesting possible modulation of oxidative stress pathways — yet the study was not powered to detect definitive clinical effects, and plasma amyloid-β and tau biomarkers and the primary neuroimaging marker N-acetylaspartate showed no significant change.

The certainty of the evidence, graded with the GRADE framework, was low to very low across outcome domains, reflecting heterogeneity, small sample sizes, short follow-up and exploratory analyses. Risk of bias, assessed with the Cochrane RoB 2 tool, was mostly low or raised some concerns, with one study judged at high risk due to missing outcome data after early termination. The review’s limitations are substantial: only three GLP-1 RAs (dulaglutide, liraglutide, exenatide) and a single SGLT-2 inhibitor (dapagliflozin) were represented, so findings cannot be generalized to other agents in these classes. Populations ranged from established Alzheimer’s disease and mild cognitive impairment to prediabetes, obesity and type 2 diabetes, and improvements in metabolically defined populations cannot be assumed to reflect disease-modifying effects in Alzheimer’s pathology, since better glycemic control, weight loss and reduced vascular risk may influence cognition independently.

Where does this leave the field? The authors conclude that current evidence suggests early mechanistic signals for GLP-1 RAs, whereas evidence for SGLT-2 inhibitors remains preliminary, and that consistent clinical cognitive benefit has not been demonstrated. The available data should be regarded as hypothesis-generating rather than practice-changing, and are insufficient to inform clinical practice or support therapeutic recommendations in Alzheimer’s disease or cognitive impairment. Two major trials loom large over the next chapter: the phase IIb ELAD trial of liraglutide in mild Alzheimer’s disease and the phase III EVOKE and EVOKE+ trials of semaglutide in early symptomatic Alzheimer’s disease, whose full results were not available at the time of the final literature search on 30 August 2025. Those studies, with prospectively defined cognitive endpoints, adequate sample sizes and longer follow-up, will determine whether the exploratory signals reported in smaller trials translate into clinically meaningful cognitive outcomes — or whether the promise of repurposing diabetes drugs for the brain dissolves under the weight of rigorous testing.

Subject of Research: Randomized trial evidence for GLP-1 receptor agonists and SGLT-2 inhibitors in Alzheimer's disease and cognitive impairment

Article Title: Systematic review of the neuroprotective potential of GLP1 receptor agonists and SGLT2 inhibitors in Alzheimer’s disease and cognitive impairment

Article References: Dewer, M. A., Dawod, S. G., Elbadawy, N. M., & Bassiouny, E. A. (2026). Systematic review of the neuroprotective potential of GLP1 receptor agonists and SGLT2 inhibitors in Alzheimer’s disease and cognitive impairment. Discover Neuroscience, 21(1), Article 115. https://doi.org/10.1186/s13064-026-00345-7

Image Credits: AI Generated

DOI: 10.1186/s13064-026-00345-7

Keywords: Alzheimer's disease, GLP-1 receptor agonists, SGLT-2 inhibitors, liraglutide, dulaglutide, exenatide, dapagliflozin, systematic review, cognitive impairment, type 2 diabetes, biomarkers, neuroimaging