Residual decline in cognition after adjustment for common neuropathologic conditions

Abstract

Objective: Neurodegenerative and cerebrovascular conditions are common in old age and are associated with cognitive decline. However, considerable heterogeneity remains in residual decline (i.e., person-specific trajectories of cognitive decline adjusted for these common neuropathologic conditions). The present study aimed to characterize profiles of residual decline in late life cognition.

Method: Up to 19 waves of longitudinal cognitive data were collected from 876 autopsied participants from 2 ongoing clinical-pathologic cohort studies of aging. Uniform neuropathologic examinations quantified measures of Alzheimer's disease, cerebral infarcts, Lewy body disease, and hippocampal sclerosis. Random effects mixture models characterized latent profiles of residual decline in global cognition.

Results: We identified 4 latent groups, and each group demonstrated distinct residual decline profiles. On average, 44% of the participants had little or no decline, 35% showed moderate decline, 13% showed severe decline and the rest (8%) had substantial within-subject fluctuation of longitudinal cognitive measures. These latent groups differed in psychological, experiential and neurobiologic factors that have been previously shown to be associated with cognitive decline. Specifically, compared with nondecliners, decliners had more depressive symptoms, were more socially isolated; were less engaged in cognitive or physical activities; and had lower density of noradrenergic neurons in locus ceruleus.

Conclusions: After controlling for common dementia related pathologies, considerable residual variability remains in cognitive aging trajectories and this variability is not random but rather is related to markers of cognitive and neural reserve. The mixture modeling approach provides a powerful tool to identify latent groups with distinct cognitive trajectories.

Figure 1. Trajectory of cognitive decline from a model without mixture (K=1)

The figure shows the raw cognitive data (gray) and fitted values (blue) from a linear mixed model (k=1) for a random sample of 100 participants. The black curve corresponds to the fitted mean trajectory.

Figure 2. Comparison of models with different number of latent classes (K)

The figure shows the values of model fit statistics of AIC and BIC decrease and level off at k=4

Figure 3. Distinct profiles of cognitive decline by latent groups

The figure shows the observed longitudinal global cognitive trajectories (gray) for the participants, as well as model derived mean trajectories (black), by latent groups.