SELECT trial: semaglutide and dementia risk

A post hoc analysis of the SELECT trial published in August 2026 reports that semaglutide slowed the progression of a validated blood-based dementia risk score in older adults with overweight or obesity and established cardiovascular disease^[1]. The finding does not mean semaglutide prevents or treats dementia, but it adds a fresh signal to a research area that has attracted sustained interest since the SELECT cardiovascular outcomes trial.

The SELECT trial and who this analysis covers

SELECT enrolled 17,604 adults aged 45 and over with a BMI of at least 27 and pre-existing cardiovascular disease but no diabetes^[2]. They were randomised to weekly subcutaneous semaglutide 2.4 mg or placebo. The trial's primary result, published in the New England Journal of Medicine in 2023, showed a 20% relative reduction in major adverse cardiovascular events (hazard ratio 0.80, 95% CI 0.72 to 0.90, P less than 0.001)^[2].

The dementia-biomarker sub-analysis drew on a subset of 2,970 participants aged 65 and over from that trial^[1]. Blood samples collected at baseline and at week 104 were analysed using the Dementia SomaSignal Test (dSST), a validated panel of 25 proteins that produces predicted 5-year and 20-year all-cause dementia risk scores. The 25 proteins in the dSST include markers spanning inflammation, neurodegeneration, and metabolic signalling pathways.

What the biomarker results showed

By week 104, the semaglutide group had a 2.5-fold smaller increase in predicted 5-year dementia risk compared to placebo^[1]. In absolute terms, that translated to a 26% lower predicted event rate (odds ratio 0.74, 95% confidence interval 0.65 to 0.85). For 20-year dementia risk, the difference was a 1.67-fold smaller increase, corresponding to an 8.8% lower predicted event rate (OR 0.91, 95% CI 0.88 to 0.94)^[1].

The analysis also found that semaglutide reduced the odds of being classified into a higher dementia risk category by 36%^[1]. These are proteomics-based predictions, not observed dementia events. The study was not designed, and does not have sufficient duration, to measure whether actual dementia diagnoses differed between groups.

What the dSST measures and why it matters

The Dementia SomaSignal Test is a 25-protein blood panel that was validated before this analysis. It is designed to predict who among a healthy older population will develop any-cause dementia within 5 or 20 years. Using a validated score rather than an ad hoc biomarker is an important methodological point: the dSST was not developed or tuned on SELECT data, which reduces the risk that the finding is a statistical artifact of post hoc exploration.

That said, post hoc analyses of large trials carry well-understood limitations. The selection of participants with blood samples available at both timepoints may not perfectly represent the full SELECT cohort. And the dSST predicts dementia risk from protein concentrations; it does not directly measure neurodegeneration or cognitive function. A slowing of the risk score is not the same as preventing dementia^[1].

How this fits the broader research picture

This analysis arrives against a complicated background. Novo Nordisk ran two large Phase 3 trials of oral semaglutide in Alzheimer's disease, EVOKE and EVOKE+, which enrolled roughly 1,840 adults each. The primary endpoint, change in Clinical Dementia Rating at week 104, was negative in both trials. Novo Nordisk discontinued the planned extensions based on those results.

The SELECT proteomics analysis is a different kind of question: not whether semaglutide slows Alzheimer's pathology in people already diagnosed with mild cognitive impairment, but whether it changes a risk trajectory in older adults who are metabolically and cardiovascularly at-risk. The populations and the outcomes are distinct, and neither result cancels the other.

Mechanistic hypotheses in the literature

Researchers have proposed several overlapping mechanisms. GLP-1 receptors are expressed in brain regions involved in energy balance and, at lower levels, in regions involved in memory and cognition. Semaglutide reduces systemic inflammation, improves insulin sensitivity, and lowers cardiovascular risk markers, any of which could plausibly reduce the metabolic burden on an aging brain. The 25-protein dSST panel captures inflammatory and neurodegeneration markers, so the signal seen in this study is consistent with a systemic anti-inflammatory or metabolic mechanism rather than direct neuronal action.

What this analysis does not establish

Several things the study does not show are worth being explicit about. It does not show that semaglutide prevents dementia. It does not establish that the biomarker changes translate into fewer dementia diagnoses over 5 or 20 years; that would require a longer follow-up study with clinical endpoints. It does not apply to people under 65, people with diabetes, or people without cardiovascular disease, because the SELECT sub-cohort was defined by those characteristics. And it does not change any approved indication for semaglutide in any jurisdiction.

This article is for informational purposes only and does not constitute medical advice. Any decision about starting, adjusting, or stopping a medication should be made with a clinician who knows your individual history.