Abstract
Physical activity (PA) is associated with memory and cognitive health, but large-scale population studies exploring the potential role of brain structural markers in the relationship between device-measured PA and memory function remain scarce. This cross-sectional study aimed to examine how device-measured PA affects memory function through the brain structure using data from 19,721 participants (aged 45-82 years) in the UK Biobank accelerometer subsample who completed memory assessments. Magnetic resonance imaging scans assessed total brain, white matter, gray matter, hippocampal, and white matter hyperintensity volumes (total brain volume [TBV], white matter volume [WMV], gray matter volume [GMV], hippocampal volume [HV], and white matter hyperintensity volume [WMHV]). Multiple linear regression and logistic regression revealed that PA was related to better short-term numeric memory performance (β: 0.045, 95% CI: 0.026, 0.064). PA was associated with greater TBV (β: 0.072, 95% CI: 0.058, 0.086), WMV (β: 0.044, 95% CI: 0.028, 0.059), GMV (β: 0.073, 95% CI: 0.059, 0.086), and HV (left: β: 0.058, 95% CI: 0.043, 0.074; right: β: 0.068, 95% CI: 0.053, 0.084) and smaller WMHV (β: -0.079, 95% CI: -0.094, -0.065). Pathway analyses demonstrated that TBV, WMV, GMV, and WMHV explained the relationship between PA and numeric memory, accounting for 25.91%, 26.36%, 10.45%, and 29.09% of the total effect, respectively. These findings suggest that PA may preserve short-term numeric memory by mitigating brain structural degeneration, highlighting its potential as a preventive strategy against memory decline in middle-aged and older adults.