Abstract
BACKGROUND: Many studies have reported that a single measure of physical activity in older adults is associated with incident dementia. In this study, we examined repeated measures of physical activity to determine if this association with incident dementia varied. METHODS: Participants were from a community-based longitudinal study of older adults (the Rush Memory and Aging Project [MAP]). MAP participants were recruited from retirement centres and subsidised long-term care facilities across northeastern Illinois (USA). Physical activity level was extracted from biennial multiday wrist-wearing sensor recordings. Dementia diagnosis was based on neuropsychological test scores, clinical history, and examination. Joint modelling integrating a linear mixed-effects model and a Cox model was used to prospectively test the association of repeated measures of physical activity and dementia, and a time-varying effects model was used to retrospectively examine the association of dementia with repeated measures of physical activity before dementia diagnosis. FINDINGS: Our analyses included 972 older adults (mean age 80·5 years, SD 7·3) with frequent (mean 4·9, SD 2·6) biennial measurements of physical activity. 745 (77%) of 972 participants were female and 227 (23%) were male. 286 (29%) of the 972 participants developed dementia, which was clinically diagnosed as Alzheimer's disease. Although the joint model indicated an association between physical activity and incident dementia (hazard ratio [HR] 0·78, 95% CI 0·69-0·88; p<0·0001), the model accuracy was stronger for physical activity closer to dementia diagnosis. Baseline physical activity was not related to the risk of dementia more than 7 years after baseline (HR 1·00, 95% CI 0·69-1·45; p=0·99), whereas physical activity at year 6 was (0·55, 0·37-0·80; p=0·0021). In a time-varying effects model, repeated physical activity was not associated with incident dementia except during the last 2 years before dementia diagnosis. INTERPRETATION: Physical activity might be a modifiable protective factor for dementia in old age. However, analyses of repeated measures of physical activity suggest that further work is needed to define the timing of the beneficial effects of physical activity relative to the onset of dementia. FUNDING: National Institutes of Health.