Domain specific cognitive functions predict neuropathological traits in the Framingham Heart Study.

Abstract

Clinical diagnosis of Alzheimer's disease depends on assessing cognition using domain specific neuropsychological (NP) tests. However, relationships between ante-mortem NP tests and detailed post-mortem neuropathological traits from the same participants are understudied. We analyzed 25 NP and 24 neuropathological traits measured in 159 brain donors from the Framingham Heart Study (FHS). Collectively, the NP tests assessed performance in memory, language, executive, and visual-spatial domains. We normalized NP tests after adjusting for sex and age at NP exam, while neuropathological traits for sex and age at death. We tested the association of NP test scores with neuropathological diagnosis of Alzheimer's disease (AD) as a binary outcome adjusting for sex, age at death, and interval age (age at death minus age at NP exam) in a logistic regression. Associations with covariate-adjusted neuropathological traits were performed under linear regression models. We also evaluated a model predicting AD including terms for domain-specific NP tests, age at death, sex, and APOE ɛ4 status RESULT: AD was significantly associated (P<0.05) with measures of memory, language, and executive function (best with executive function: P=0.0016). Braak stages and plaque scores showed a strong association with all cognitive domains (P<0.02). Neurofibrillary tangle (NFT) with memory domain (P=7.4x10-5 ) and amyloid beta plaque density with executive function (P=0.01) showed significant associations. Total tau level was strongly associated with executive function (P=5.3x10-6 ). Significant associations were also observed between pTau231 and language (P=0.001), as well as between pTau396 and visual-spatial domain (P=0.03). PTau181 (P=0.02), pTau202 (P=0.0089), and amyloid-β40 (P=5.9x10-4 ) were significantly associated with memory domains. For complement component proteins, the C4A level was nominally significant with language (P=0.03) and the C4B level with executive function (P=0.02). AD was predicted as area under the ROC (receiver operating characteristic) curve is 0.75 (confidence interval: 0.59-0.91), where memory and language domains were the top-ranked predictors with importance value>0.63. We confirmed association of memory and executive function with AD-related neuropathological traits. This study demonstrated for the first time associations of ante-mortem language and executive function with post-mortem complement component protein levels. AD in this study was largely predicted by language and memory performance.