Physical activity has been suggested as modifiable factor that might contribute to improving cognitive and brain function during aging. However, previous studies were mainly of cross-sectional design and did not consider effects of time or potential reverse causality. We aimed to investigate the bidirectional associations of physical activity with brain structure in middle-aged and older adults. Overall, 4365 participants (64.01 ± 10.82 years; 56% women) from the Rotterdam Study had physical activity and brain structure assessed on at least one of 2 timepoints ('baseline': 2006-2012 or 'follow-up': 2012-2017, median duration between visits: 5 years). Physical activity was assessed through the LASA Physical Activity Questionnaire. T1-weighted MRI and diffusion tensor imaging were used to quantify brain volumes and white matter microstructure, respectively. Cross-lagged panel models were performed to estimate bidirectional associations, and linear mixed-effects models to investigate the consistency of findings. Larger total brain volume (β = 0.067, 95%-confidence interval 0.035;0.099, pFDR = 0.001), gray matter volume (β = 0.063, 0.031;0.096, pFDR = 0.002), and white matter volume (β = 0.051, 0.020;0.083, pFDR = 0.013) at baseline were associated with higher levels of sports at follow-up. Lower global mean diffusivity at baseline was associated with higher levels of walking at follow-up (β = -0.074, -0.111;-0.037, pFDR = 0.001). No associations were found between physical activity levels at baseline and brain metrics at follow-up. In conclusion, larger brain volumes and white matter microstructure at baseline were associated with individuals remaining more physically active at follow-up. Overall, this study identified older adults with potentially advanced brain aging status as being at higher risk of physical inactivity over time, and therefore as potential target group for prevention and novel intervention strategies.

Physical activity levels and brain structure in middle-aged and older adults: a bidirectional longitudinal population-based study