Antemortem Domain-Specific Cognitive Functions Predict Postmortem Neuropathological Traits in the Framingham Heart Study

Abstract

Abstract Background: Clinical diagnosis of Alzheimer’s disease is aided by neuropsychological (NP) tests. However, correlations between antemortem domain specific cognitive functions from NP tests and neuropathological outcomes at autopsy are not well known. Methods: We conducted association tests between antemortem NP tests and postmortem neuropathological traits in 159 participants from the Framingham Heart Study (FHS). The NP tests included memory, language, executive, and visuospatial domains. Neuropathological traits contained neuropathological diagnosis of AD (AD), Braak stage, neuritic plaque score, and quantitative measures of site-specific tau phosphorylation (pTau), density of microglia and synaptic protein density from the dorsolateral prefrontal cortex area of the autopsied brains in the FHS subjects. For the most significantly associated postmortem outcome (pTau) with antemortem cognitive domains, we compared prediction models containing significant cognitive domains with those including all cognitive domains, adjusting for interval age and APOE ɛ4 status as covariates. Results: We identified significant associations for AD pathology with memory, language, and executive domains (best P with memory=1.1x10-4). At least three distinct cognitive domains were nominally associated (P<0.05) with Braak stages, plaque scores, and phosphorylated paired helical filament tau (AT8). Language domain impairment was associated with microglia (Iba1) density, while language and executive function impairments were significantly associated with complement component C4a and C4b levels, respectively (P<0.05). Impairment in the memory domain was significantly associated with levels of pTau181, pTau202, pTau231, postsynaptic density protein 95 (PSD95), and ratio of Aβ42/40 (best P with pTau231=0.008). Impairment in the visuospatial domain was associated with pTau396, and the area under the curve (AUC) for a model predicting high pTau396 utilizing the block design was better (AUC=0.81, 95% confidence interval [CI]: 0.46-0.85) than for a model including tests for all domains (AUC=0.61, 95% CI=0.31-0.84). Conclusions: These findings confirm significant associations between antemortem domain specific cognitive tests and neuropathological traits at autopsy, especially postmortem measures of phosphorylated tau levels. Prediction models demonstrated that impairment in block design alone is sufficient to predict a high-level of pTau396 at autopsy.