DISTINCT COGNITIVE TRAJECTORIES IN LATE MIDDLE-AGE AND THEIR ASSOCIATIONS WITH BRAIN STRUCTURE AND ALZHEIMER'S DISEASE BIOMARKERS: FINDINGS FROM THE WISCONSIN REGISTRY FOR ALZHEIMER’S PREVENTION

Abstract

Late middle-age is an important period to identify mild cognitive decline expected to precede clinical diagnoses of mild cognitive impairment and dementia. This study aimed to characterize patterns of longitudinal cognitive change in late middle-age and to identify a subgroup at highest risk for dementia. 677 late middle-aged participants enriched for AD risk (mean age at baseline 53.5, 69% female, 77% parental history of AD) who were cognitively healthy at baseline completed 4 or 5 neuropsychological assessments over a period of 8-10 years. Using linear mixed-effects regression, unconditional growth models were conducted for six neuropsychological variables, and the fixed and random effect estimates were combined to obtain slope estimates for each participant. Hierarchical cluster analysis was conducted on slope estimates, and resulting clusters were compared on slopes, intercepts, demographics, AD risk, and clinical outcomes. A subset underwent MRI scans (n=235) and cerebrospinal fluid assays (n=88). ANCOVA models (covarying age and gender) compared clustered groups on brain structure and AD biomarkers. Cluster analysis resulted in three groups: 1) Multi-domain group declining on learning, memory, and executive function measures (n=188), 2) dysexecutive group with mild decline in verbal learning and executive functioning (n=152), and 3) group with normal age-related decline in processing speed only (n=337) (Table 1, Figure 1). The multi-domain cluster was older (F=19.32, p<.001), comprised of more men (X2 = 9.20; p=.01), and endorsed more depressive symptoms (F=3.51, p<.05) at baseline. The multi-domain group included a greater percentage of APOE ε4 carriers (X2 = 5.27, p=.07), individuals with mild clinical symptoms (CDR=0.5, X2 = 20.39, p<.001) and clinical diagnoses (X2 = 52.69, p<.001) at follow-up, and individuals classified as AD biomarker positive (18% total-tau/Aβ-42 positive; X2 = 7.26; p<.05). The dysexecutive cluster had less education (F=5.06, p<.01) and lower literacy estimates (F=14.06, p<.001). The normally aging cluster exhibited higher intercepts on neuropsychological measures (p'S < .02) and less global atrophy (F=6.80, p=.001) (Figure 2). Annual rates of change on neuropsychological measures included in cluster analysis projected over 20-year period (note: Increase in Trailmaking Test Part B = worse performance). Mean global atrophy in three clusters. Full model also included age, gender, and intracranial Cluster analysis of slopes reflecting annual rates of cognitive change provides a data-driven method to identify mild cognitive decline in late middle-age that may indicate increased risk for dementia due to AD.