INTRODUCTION: To study the relationship of late-life hemoglobin A1C (A1C) with postmortem neuropathology in older adults with and without diabetes mellitus (DM). METHODS: A total of 990 participants from five cohort studies of aging and dementia with at least two annually-collected A1C measures, who had autopsy. Neuropathologic evaluations documented cerebrovascular disease, Alzheimer's disease (AD), and other pathologies. To evaluate the association of A1C mean and variability (standard deviation [SD]) with neuropathology, we used a series of adjusted regression models. RESULTS: Participants (mean age at death = 90.8 years
education = 15.8 years
76% women) had six A1C measurements on average. Mean A1C was associated with greater odds of macroinfarcts (estimate = 0.14
p = 0.04) and subcortical infarcts (estimate = 0.16
p = 0.02). A1C variability was not associated with cerebrovascular pathology. A1C mean and variability were inversely associated with AD pathology. DISCUSSION: The A1C average over time was associated with infarcts, and the A1C average and variability were inversely associated with AD pathology. Future studies should explore the underlying mechanisms linking A1C to dementia-related neuropathologies. HIGHLIGHTS: Hemoglobin A1C (A1C), a measure of peripheral insulin resistance, is used to assess glycemic control. Higher A1C mean was associated with greater odds of macroscopic subcortical infarcts. A1C variability was not associated with cerebrovascular pathology. Both A1C mean and variability had inverse associations with AD pathology. None of the associations varied by diabetes mellitus status.