Estimating a clinically normal individual’s position along a preclinical Alzheimer’s disease continuum using cognitive and amyloid trajectories
Diana Townsend, Michael J Properzi, Tobey J Betthauser, Hannah M Klinger, Rory Boyle, Gillian T Coughlan, Bernard J Hanseeuw, Hyun‐Sik Yang, Rebecca E. Amariglio, Michelle E. Farrell, Heidi I.L. Jacobs, Zahra Shirzadi, Wai‐Ying Wendy Yau, Julie C Price, Jasmeer P. Chhatwal, Dorene M. Rentz, Keith A. Johnson, Reisa A. Sperling, Aaron P. Schultz, Rachel F. Buckley- Psychiatry and Mental health
- Cellular and Molecular Neuroscience
- Geriatrics and Gerontology
- Neurology (clinical)
- Developmental Neuroscience
- Health Policy
- Epidemiology
Abstract
Background
Optimizing longitudinal cognitive and biomarker trajectories can distill multiple observations from one individual into a single metric. Relative to other individuals, this metric can represent an individual’s distance from an anchor‐point based on their rate and non‐linearity of change. We have recently developed a cognitive time (c‐time) based on the cognitive trajectories of clinically normal older adults. We examined the association between c‐time and a previously published ‘time‐to‐Aβ+ threshold’ and how these metrics align with demographics and other biomarkers.
Method
We identified 135 clinically normal older adults from the Harvard Aging Brain Study (Agemean:73years(±5.9); Female:61%) with ≥3 neuropsychological assessments and PiB‐PET, ≥1 Flortaucipir‐PET, ≥2 volumetric MRI, and diagnostic follow up. We defined c‐time using iterative non‐linear least‐squares optimization to define a curvilinear function that described the group‐level Preclinical Alzheimer Cognitive Composite (PACC) trajectory (Fig1D). Each participant’s PACC trajectory was subsequently located on the curve using the same optimization framework (Fig1E). We identified the anchor‐point of cognitive decline (c‐time) using piecewise linear mixed‐effects models. Time‐to‐Aβ+ was calculated using the published sampled iterative local approximation (SILA; Fig1A) algorithm with the anchor‐point indicating Aβ+ threshold (Fig1B). We examined associations between c‐time and time‐to‐Aβ+ using linear regression. Individuals were subsequently placed into groups depending on their position relative to the anchor‐point on each axis, as well as the line‐of‐best‐fit (Fig2). We compared the groups on demographics, and both cross‐sectional and longitudinal indices of medial temporal (MTL) Flortaucipir‐PET (entorhinal, parahippocampal, amygdala) and ICV‐adjusted hippocampal volume.
Result
C‐time and time‐to‐Aβ+ were significantly associated (r = 0.42,p<0.001). Only one participant (who progressed to MCI/dementia) was post‐c‐time and remained pre‐time‐to‐Aβ+, supporting the notion that time‐to‐Aβ+ occurs prior to cognitive inflection. Individuals post‐c‐time and post‐time‐to‐Aβ+ (Group 1) were more likely to be APOEε4 carriers, progressors to MCI/dementia, have significantly higher baseline MTL tau and lower hippocampal volume, and faster hippocampal atrophy (Fig3). Group 2 (post‐time‐to‐Aβ+/pre‐c‐time) were more likely APOEε4 carriers. Notably, no age or other effects were apparent between groups.
Conclusion
Optimizing longitudinal cognitive and biomarker data to estimate a preclinical disease continuum can provide unique, and potentially age‐independent, information about the distance an individual might be from disease‐relevant events.