Among 6,328 adults aged 55+ with type 2 diabetes and no prior dementia, initiating GLP-1 receptor agonists was associated with a statistically significant 0.21-month increase in Alzheimer's disease-type dementia-free survival over 48 months compared to sulfonylureas (95% CI: 0.07–0.35; NNT≈202). The highest-benefit subgroup gained 0.45 months. Against SGLT2 inhibitors (n=3,070), no average effect reached significance, though a high-benefit subgroup gained 0.83 months, with older age, insulin use, lower HbA1c, and lower BMI predicting greater benefit.
This preprint, not yet peer-reviewed, adds nuance to a rapidly expanding body of observational evidence suggesting GLP-1 agonists may confer neuroprotective effects beyond glycemic control — possibly through anti-inflammatory pathways, amyloid clearance modulation, or improved cerebrovascular health. The effect sizes, however, are clinically modest: less than a month of dementia-free time over four years, with number-needed-to-treat around 200. Target trial emulation with doubly robust causal machine learning is methodologically rigorous for observational data, but residual confounding, reliance on electronic health records for dementia ascertainment, and the predominance of patients already engaged in specialty care limit generalizability. Crucially, the comparison drug matters enormously — sulfonylureas carry known hypoglycemic risk, potentially inflating GLP-1 benefits by contrast. The SGLT2 inhibitor comparison is the more clinically meaningful benchmark and remains inconclusive. These are hypothesis-generating findings; treatment decisions should still prioritize glycemic, cardiovascular, and renal outcomes until randomized trial data emerge.
