Abstract
BACKGROUND AND OBJECTIVES: Cerebral amyloid angiopathy (CAA) is common in older adults and is associated with dementia. Less is known whether this association is mediated by Alzheimer disease (AD) neuropathologic changes, the examination of which was the objective of this study. METHODS: This was a retrospective cross-sectional examination of the Kaiser Permanente Washington database of the Adult Changes in Thought (ACT) autopsy cohort with information on CAA, dementia, the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) (amyloid neuritic plaques), and Braak (tau neurofibrillary tangles). CAA was diagnosed by immunohistochemistry and dementia by ACT Consensus Diagnostic Conference. AD neuropathology was categorized by CERAD scores and Braak stages. Multivariable logistic regression models were used to estimate odds ratios (ORs) and 95% CIs of the associations of CAA with dementia, adjusting for age at death and sex, and with additional adjustments separately for CERAD scores (moderate-severe vs mild-absent), Braak stages (V-VI vs 0-IV), APOE ε4, and stroke. Formal mediation analyses were conducted to estimate age-sex-adjusted OR (95% CI) for natural indirect effects (NIEs) of CERAD scores and Braak stages. RESULTS: The 848 participants had a mean age of 86.7 ± 4.6 years at death, and 57.6% were female. CAA was present in 322 participants (38.0%), of whom 152, 145, and 25 had mild, moderate, and severe CAA, respectively. Dementia was present in 384 participants (45.3%), of whom 317 had AD. Dementia was more common in those with CAA than without (53.7% vs 40.1%; age-sex-adjusted OR 1.57, 95% CI 1.18-2.10). This association remained significant after separate adjustment for other covariates but lost significance when adjusted for CERAD scores (OR 1.27, 95% CI 0.93-1.71) and Braak stages (OR 0.96, 95% CI 0.69-1.33). Findings from our formal mediation analyses show that ORs (95% CIs) for NIE of CERAD scores and Braak stages were 1.25 (1.13-1.37) and 1.63 (1.38-1.88), respectively, and CERAD scores and Braak stages mediated 53% and 111% of the total association, respectively. DISCUSSION: We observed a significant association between CAA and dementia that disappeared when adjusted for CERAD or Braak stages. Findings from our mediation analyses suggest that the CAA-dementia association may be potentially mediated by AD neuropathologic changes. This hypothesis needs to be tested in future mechanistic studies in AD accounting for unmeasured confounders.