Pre-diabetes as a Critical Stage for Risk of Dementia and Stroke: Evidence from the UK Biobank and Mendelian Randomization

Background and Aims Type 2 diabetes (T2D) is a recognized risk factor for dementia and stroke, but whether risks begin during pre-diabetes remains unclear. We aimed to determine the risks of these neurological conditions with pre-diabetes and the glycemic thresholds at which risks emerge. Methods We studied 432,285 adults (mean age 57 years; 55% women) from the UK Biobank without prior dementia or stroke at baseline. Glycemic status (normoglycemia, pre-diabetes, T2D) was defined by HbA1c. Incident dementia, stroke, and MRI-derived brain markers, including brain volumes and white matter hyperintensity volumes for these conditions, were assessed. HbA1c was modeled continuously to assess non-linear effects. Outcomes were estimated using competing risks and linear regression models, adjusted for demographic, lifestyle, and clinical factors. Two-sample Mendelian randomization (MR) was used to test causality. Results During a median 13.7 years of follow-up, pre-diabetes was associated with higher risks of Alzheimer's disease (HR:1.30; 95% CI: 1.16-1.45), vascular dementia (HR:1.50; 95% CI: 1.28-1.7), and all-cause stroke (HR:1.24; 95% CI: 1.16-1.33), independent of other vascular risk factors. On brain MRI (N=3,723), total and grey matter volumes and hippocampal volumes were smaller, and white matter hyperintensity volume was larger in pre-diabetes. Spline models indicated that adverse outcomes emerged at HbA1c levels around 40-45 mmol/mol, below the diabetes threshold. Risks were further elevated in T2D. MR analyses supported causal effects of glycemia on Alzheimer's disease, stroke, and hippocampal atrophy. Conclusion Pre-diabetes is not a benign state: even modest HbA1c elevations are linked to dementia, stroke, and brain structural decline. These risks emerge below current diagnostic glycemic thresholds, underscoring pre-diabetes as a critical stage for early intervention to preserve brain health.