Amyloid and Tau Pathology Associations With Personality Traits, Neuropsychiatric Symptoms, and Cognitive Lifestyle in the Preclinical Phases of Sporadic and Autosomal Dominant Alzheimer's Disease

Abstract

Background: Major prevention trials for Alzheimer's disease (AD) are now focusing on multidomain lifestyle interventions. However, the exact combination of behavioral factors related to AD pathology remains unclear. In 2 cohorts of cognitively unimpaired individuals at risk of AD, we examined which combinations of personality traits, neuropsychiatric symptoms, and cognitive lifestyle (years of education or lifetime cognitive activity) related to the pathological hallmarks of AD, amyloid-β, and tau deposits.

Methods: A total of 115 older adults with a parental or multiple-sibling family history of sporadic AD (PREVENT-AD [PRe-symptomatic EValuation of Experimental or Novel Treatments for AD] cohort) underwent amyloid and tau positron emission tomography and answered several questionnaires related to behavioral attributes. Separately, we studied 117 mutation carriers from the DIAN (Dominant Inherited Alzheimer Network) study group cohort with amyloid positron emission tomography and behavioral data. Using partial least squares analysis, we identified latent variables relating amyloid or tau pathology with combinations of personality traits, neuropsychiatric symptoms, and cognitive lifestyle.

Results: In PREVENT-AD, lower neuroticism, neuropsychiatric burden, and higher education were associated with less amyloid deposition (p = .014). Lower neuroticism and neuropsychiatric features, along with higher measures of openness and extraversion, were related to less tau deposition (p = .006). In DIAN, lower neuropsychiatric burden and higher education were also associated with less amyloid (p = .005). The combination of these factors accounted for up to 14% of AD pathology.

Conclusions: In the preclinical phase of both sporadic and autosomal dominant AD, multiple behavioral features were associated with AD pathology. These results may suggest potential pathways by which multidomain interventions might help delay AD onset or progression.

Keywords: Alzheimer’s; PET; Prevention; Reserve; Resistance; Risk factors.

Figure 1.

Partial least squares analysis finds maximally correlated linear combinations of two input matrices, one with behavioral features (top matrix in A) and the other with Alzheimer’s disease pathology across defined cortical regions (bottom matrix in A). These two matrices are then correlated together, and this latter matrix (B) is decomposed into multiple latent variables using singular value decomposition. (C) An example of a latent variable. Briefly, each latent variable consists of a singular value (related to the covariance between the 2 input matrices) and 2 vectors of weights representing how much each behavioral feature and each brain region contribute the overall multivariate relationship. Aβ. amyloid-β; SUVR, standardized uptake value ratio.

Figure 2.

Correlations between behavioral features in both cohorts. Intercorrelation (Pearson correlation) between behavioral factors in PRe-symptomatic EValuation of Experimental or Novel Treatments for AD (PREVENT-AD) (A) and Dominant Inherited Alzheimer Network (DIAN) (B) study groups. White stars correspond to negative correlations and black stars to positive correlations that remained significant after false discovery rate correction. NPI-Q, Neuropsychiatric Inventory Questionnaire.

Figure 3.

Latent variables from partial least squares analysis relating behavioral features and Alzheimer’s disease (AD) pathology in both cohorts. Results from the different partial least squares analyses representing which combinations of behavioral features relate to amyloid-β (Aβ) pathology in PRe-symptomatic EValuation of Experimental or Novel Treatments for AD (PREVENT-AD) study group (A), tau pathology in PREVENT-AD study group (B), and Aβ pathology in Dominant Inherited Alzheimer Network (DIAN) study group (C). Bar graphs represent the weight of each behavioral feature to the multivariate relationship. Confidence intervals are derived from bootstrap resampling. All brain regions included in the partial least squares analyses are projected on the brains. Bootstrap ratios correspond to the importance of each region to the behavioral feature-pathology relationship. NPI-Q, Neuropsychiatric Inventory Questionnaire.

Figure 4.

Latent variable from partial least squares analysis relating personality facets and behavioral features with amyloid-β (Aβ) in Dominant Inherited Alzheimer Network (DIAN) study group. Result from the partial least squares analysis relating behavioral features including the 30 personality facets and AP pathology across brain regions in DIAN. Bar graphs represent the weight of each behavioral feature to the multivariate relationship. Confidence intervals are derived from bootstrap resampling. All brain regions included in the analysis are projected on the brain. Bootstrap ratios correspond to the importance of each region to the behavioral feature-pathology relationship. GDS, Geriatric Depression Scale; NPI-Q, Neuropsychiatric Inventory Questionnaire.