Diabetes and neurodegeneration in cognitively unimpaired older adults: implications for cognition

Type 2 diabetes mellitus is associated with cognitive impairment and greater Alzheimer's disease risk, with cross-sectional studies suggesting that this is driven by neurodegeneration and vascular changes. Longitudinal studies, however, report similar rates of total brain atrophy over time in diabetic and non-diabetic adults. We recently showed that diabetes-related neurodegeneration in cognitively unimpaired adults is regional and may not be evident in longitudinal studies of global brain atrophy. The mechanisms driving diabetes-related neurodegeneration are unknown; glycemic control may play an important role but the relative influence of Alzheimer's and cerebrovascular pathology is unclear. Here, we longitudinally examined regional patterns of cortical thinning associated with diabetes and glycemic control over nearly 3 years. We controlled for Alzheimer's disease neuropathology and white matter hyperintensity burden. We also examined whether subsequent changes in hemoglobin A1c (HbA1c) levels correlated with rates of cortical thinning in diabetes-associated regions and tested whether faster cortical thinning in diabetes-relevant regions mediated a relationship between diabetes and decline in cognitive performance. Among 1,298 cognitively unimpaired participants (mean age=65.0 years, 305 diabetic, 869 female) from the Health and Aging Brain Study-Health Disparities cohort who completed baseline and follow-up MRI scans, we used linear mixed-effects models to examine the relationship between diabetes and cortical thickness changes across 34 brain regions, controlling for socioeconomic factors and comorbidities. We further adjusted for amyloid-PET, tau-PET, white matter hyperintensities, and APOE ε4 carrier status. We also examined associations between baseline and longitudinal HbA1c levels and cortical thinning rates in diabetes-related regions in the whole sample and separately in diabetic and non-diabetic participants. P-values were corrected using the false discovery rate method. Path analysis tested whether diabetes-related cortical thinning mediated the relationship between diabetes and decline in cognitive performance. Diabetic participants exhibited faster cortical thinning in seven frontal, parietal, and occipital regions (-0.060≤βs≤-0.046, corrected Ps<0.017). Associations remained unchanged after accounting for socioeconomic factors, comorbidities, amyloid, tau, white matter hyperintensities, and APOE ε4. Higher baseline HbA1c predicted faster thinning in diabetes-vulnerable regions (corrected Ps<0.032), independent of subsequent HbA1c changes. Diabetes was associated with faster decline in processing speed (β=-0.022, P=0.033), with cortical thinning in diabetes-related regions mediating approximately 13% of this effect (indirect effect β=-0.034, P=0.019). Diabetes-related cortical thinning in cognitively unimpaired older adults follows a regional pattern independent of Alzheimer's and cerebrovascular pathology and partially mediates accelerated decline in processing speed. Chronic hyperglycemia may contribute to diabetes-related neurodegeneration, regardless of subsequent short-term changes in glycemic control.