# Trajectories of Cognitive Decline and Subsequent Physical Activity

## TL;DR

People who experience faster cognitive decline tend to have lower and more variable physical activity later in life, even after accounting for initial cognitive levels.

## Contribution

This study is the first to show that faster cognitive decline is independently linked to reduced and more fragmented physical activity in older adults.

## Key findings

- Faster cognitive decliners had lower daily activity and higher fragmentation compared to stable cognition groups.
- Variability in physical activity was consistently lower in fast cognitive decliners across all cognitive domains.
- These associations remained significant after adjusting for baseline cognition and other covariates.

## Abstract

Physical activity (PA) is a modifiable risk factor for dementia, but whether a faster rate of cognitive decline adversely affects subsequent PA is unknown. We investigated the association of cognitive trajectories over 8.2((SD = 1.6), visits 5-9) years with late-life PA (visit 9) in 810 (mean age73.1(SD = 4.0)years, 60.4%, female) Atherosclerosis Risk in Communities study participants. PA was assessed using 24-hour/7-day wrist-worn accelerometry and defined using measures of daily activity volume, fragmentation, and variability. Overall and domain-specific cognitive function were derived using mixed-effects models and divided into tertiles of cognitive decline (fast, slow, stable(ref)). Associations between cognitive trajectories and PA were analyzed using multivariable linear regression adjusted for relavant covariates. At baseline, fast cognitive decliners were more likely to have low education, smoking, and ≥1 APOE ε4 allele compared to those in the slow-decline and stable groups. For overall cognition, compared to those in the stable group, those in the fast- and slow-decline groups had lower daily activity (β=-35.4(SE = 9.8)min/day, β=-24.3(SE = 8.6)min/day, respectively), higher fragmentation (β = 2.9%(SE = 0.7), β = 1.3%(SE = 0.6), respectively), and lower variability of activity (β=-127.6(SE = 18.4), β=-62.9(SE = 16.2), respectively) (p < 0.01 for all). In domain-specific analyses, results for executive function were similar to overall cognition, but significance for language and memory results varied. Across all domains, variability was consistently lower among those in the fast-decline group compared to the stable group (p < 0.05 for all). Collectively, these results suggest those with faster cognitive decline have lower and more variable daily PA, independent of baseline cognition. Notably, variability of PA may be an indicator of adverse cognitive change across domains.