Harnessing cognitive trajectory clusterings to examine subclinical decline risk factors

Abstract

Cognitive decline in Alzheimer’s disease and other dementias typically begins long before clinical impairment. Identifying people experiencing subclinical decline may facilitate earlier intervention. This study developed cognitive trajectory clusters using longitudinally-based random slope and change point parameter estimates from a Preclinical Alzheimer’s disease Cognitive Composite and examined how baseline and most recently available clinical/health-related characteristics, cognitive statuses and biomarkers for Alzheimer’s disease and vascular disease varied across these cognitive clusters. Data were drawn from the Wisconsin Registry for Alzheimer’s Prevention, a longitudinal cohort study of adults from late midlife, enriched for a parental history of Alzheimer’s disease and without dementia at baseline. Participants who were cognitively unimpaired at the baseline visit with >= 3 cognitive visits were included in trajectory modeling (n=1068). The following biomarker data were available for subsets: positron emission tomography (PET) amyloid (Amyloid: n = 367; [C-11]PiB: Global PiB distribution volume ratio); PET tau (Tau: n = 321; [F-18]MK-6240: primary regions of interest Meta-Temporal composite); MRI neurodegeneration (Neurodegeneration: n = 581; hippocampal volume and global brain atrophy); T2-FLAIR MRI white matter ischemic lesion volumes (Vascular: White matter hyperintensities; n=419); and plasma pTau217 (n=165). Posterior median estimate person-level change points, slopes pre- and post- change point, and estimated outcome (intercepts) at change point for cognitive composite were extracted from Bayesian Bent-Line Regression modeling and used to characterize cognitive trajectory groups (K-means clustering). A common method was used to identify Amyloid/Tau/Neurodegeneration/Vascular biomarker thresholds. We compared demographics, last visit cognitive status, health-related factors and Amyloid/Tau/Neurodegeneration/Vascular biomarkers across the cognitive groups using ANOVA, Kruskal-Wallis, Chi-square, and Fisher’s exact tests. Mean(SD) baseline and last cognitive assessment ages were 58.4(6.4) and 66.6(6.6) years, respectively. Cluster analysis identified 3 cognitive trajectory groups representing Steep: n = 77(7.2%); Intermediate: n = 446(41.8%); and Minimal: n = 545(51.0%) cognitive decline. The Steep decline group was older, had more females, APOE e4 carriers, and Mild cognitive impairment/dementia at last visit; it also showed worse self-reported general health-related and vascular risk factors and higher Amyloid, Tau, Neurodegeneration and White matter hyperintensities positive proportions at last visit. Subtle cognitive decline was consistently evident in the steep decline group and was associated with generally worse health. In addition, cognitive trajectory groups differed on etiology-informative biomarkers and risk factors, suggesting an intimate link between preclinical cognitive patterns and Amyloid/Tau/Neurodegeneration/Vascular biomarkers differences in late middle-aged adults. The result explains some of the heterogeneity in cognitive performance within cognitively unimpaired late middle-aged adults.