Toward a multifactorial model of Alzheimer disease

Abstract

Relations among antecedent biomarkers of Alzheimer disease (AD) were evaluated using causal modeling; although correlation cannot be equated to causation, causation does require correlation. Individuals aged 43 to 89 years (N = 220) enrolled as cognitively normal controls in longitudinal studies had clinical and psychometric assessment, structural magnetic resonance imaging (MRI), cerebrospinal fluid (CSF) biomarkers, and brain amyloid imaging via positron emission tomography with Pittsburgh Compound B (PIB) obtained within 1 year. CSF levels of Aβ(42) and tau were minimally correlated, indicating they represent independent processes. Aβ(42), tau, and their interaction explained 60% of the variance in PIB. Effects of APOE genotype and age on PIB were indirect, operating through CSF markers. Only spurious relations via their common relation with age were found between the biomarkers and regional brain volumes or cognition. Hence, at least 2 independent hypothesized processes, one reflected by CSF Aβ(42) and one by CSF tau, contribute to the development of fibrillar amyloid plaques preclinically. The lack of correlation between these 2 processes and brain volume in the regions most often affected in AD suggests the operation of a third process related to brain atrophy.

Figure 1

Histograms showing distributions of PIB and the CSF biomarkers (Aβ₄₂, tau, ptau)

Figure 2

Scatter plots of relations among PIB, Aβ_42, and tau

Figure 3

Interaction between CSF Aβ₄₂ and tau as related to amyloid plaque burden (PIB)

Figure 4

Model showing relations among all of the variables. The beta weights relating APOE and age to Aβ₄₂ are from the simultaneous regression of those three variables on Aβ₄₂. The beta weights relating Aβ₄₂ and tau to PIB are from the final step of the hierarchical regression analysis reported in Table 2. The beta weights relating age to hippocampus volume are from the second step of the analysis reported on the first line of Table 3.