Abstract
BackgroundPhysical activity (PA) is known to benefit sleep health, and the detrimental effects of PM2.5 on sleep are also recognized. However, whether taking PA with PM2.5 exposure is still beneficial to sleep is not well-understood.MethodsWe analyzed two UK Biobank cohorts. Both cohorts included PM2.5 exposure and sleep behavior data. The full cohort (n = 293,980) collected walking and moderate-to-vigorous PA (MVPA) data via International Physical Activity Questionnaire from 2006 to 2010. The accelerometer-measured subsample (n = 63,474) provided light PA (LPA) and MVPA data from 2013 to 2015. Using linear regression and generalized additive models, we assessed independent associations and evaluated effect modification by PA on the PM2.5-sleep relationship using interaction terms. We further characterized nonlinear dose-response patterns to identify optimal PA levels.ResultsLong-term PM2.5 exposure was associated with reduced sleep quality and sleep duration. For every 10 MET-hours/week increase in MVPA, the negative association between PM2.5 and sleep quality was reduced by 0.0651 (0.0096, 0.1205), mitigating 24.7% of the detrimental effect of PM2.5 on sleep (derived from the ratio of the interaction coefficient to the main effect coefficient). Walking offset 19.6% of the negative impact. A threshold effect was observed in the dose-response relationship; optimal sleep improvements occurred at 37 MET-hours/week for walking (11.2 h/week) and 50 MET-hours/week for MVPA (8.33 h/week). Notably, these modifying effects were particularly robust in vulnerable populations, including individuals with obesity, males, and older adults.ConclusionsLong-term exposure to PM2.5 compromises sleep health, whereas PA can partially offset these adverse effects. This study identifies granular dose-response thresholds for Walking and MVPA, offering a quantitative target for intervention. PA serves as an economical strategy to preserve sleep health under PM2.5 exposure, with potential implications for tailoring exercise guidelines.