Estimating the preclinical Alzheimer's disease course with multimodal data

In observational studies of preclinical AD, an arbitrary "baseline" can obscure where an individual is located along a theoretical continuum. Optimizing longitudinal trajectories can distill multiple, non-linearly distributed observations into a single metric and inform where an individual may be along the disease course.

We developed a cognitive time (c-time) metric based on longitudinal cognitive data (mean = 7.4 years, range = 1.7-11.6) from 316 participants from the Harvard Aging Brain Study using non-linear least-squares optimization. We examined path analyses between a published time-to-amyloid beta (Aβ)+ metric (longitudinal Aβ positron emission tomography [PET]: mean = 6.3 years, range = 4.2-9.7) and c-time, including demographics, brain volume, cortical thickness, tau PET, and cardiovascular risk.

Time-to-Aβ+ and c-time were positively correlated, with time-to-Aβ+ possessing direct and indirect associations with c-time through regional tau PET, hippocampal volume, and cortical thickness.

Optimizing longitudinal multimodal data to estimate a theoretical continuum can provide unique and age-independent information about the distance an individual might be from disease-related events.

Cognitive time (c-time) represents an individual's distance from a hinge point of cognitive decline. C-time and time-to-amyloid beta (Aβ)+ were moderately positively correlated. Inferior temporal tau and cortical thickness mediated the effect between c-time and time-to-Aβ+. C-time more closely associates with tau, brain atrophy, and cortical thickness.