Applying two-sample drug-target Mendelian randomisation (MR) across eight antidiabetic drug targets, this analysis found no robust genetic evidence that currently licensed diabetes medications reduce vascular dementia (VaD) risk. Using HbA1c-based genetic instruments derived from 437,749 individuals and VaD outcome data from 7,009 cases against nearly 900,000 controls, none of the seven testable targets — including GLP-1 receptors, SGLT2, and metformin's target AMPK — reached conventional significance for VaD, white matter hyperintensity volume, or other cerebrovascular neuroimaging markers. DPP4 inhibition showed a suggestive protective signal (OR 0.68, 95% CI 0.46–1.02), but the confidence interval crossed the null. PPARG modulation showed a secondary signal on white matter integrity, though this was isolated.
This finding matters because vascular dementia has no approved disease-modifying therapy, and repurposing diabetes drugs — particularly GLP-1 agonists, which have attracted intense interest for neuroprotection — has been a high-profile research priority. These null MR results provide important genetic-level evidence that any observed epidemiological associations may reflect confounding rather than causality. That said, MR captures lifetime exposure to HbA1c-lowering via specific targets, which may not replicate the pharmacological effects of acute drug use, particularly for drugs with pleiotropic actions. The DPP4 signal, while sub-threshold, aligns with some observational literature and deserves replication in larger datasets. Critically, this is a preprint posted on medRxiv and has not yet been peer-reviewed — findings and their interpretation may change substantially after expert scrutiny. Currently, the evidence does not support initiating these drugs specifically for dementia prevention.