Accelerometer-derived physical activity is associated with reduced stroke risk. The biological pathways underpinning this relationship, however, are not yet understood. Herein, we aim to identify metabolic signatures associated with accelerometer-measured PA and investigate their relationships with reduced stroke incidence.

Utilizing UK Biobank accelerometer data, we derived physical activity into total physical activity (TPA), moderate-to-vigorous physical activity (MVPA), and light physical activity (LPA) and linked them to 249 NMR-quantified plasma metabolites. The metabolomic signatures (TPA-/MVPA-/LPA-metabolomic signatures) were developed through internal validation followed by elastic-net regression modeling. Cox proportional hazards models evaluated activity-stroke associations (adjusted for sociodemographic/genetic factors), followed by mediation analysis to quantify metabolomic signature effects.

Through UK Biobank study (N = 29445; 14.1-year follow-up with 513 stroke events), we identified 195 TPA, 173 MVPA, and 164 LPA metabolite associations (FDR < 0.05), with 107, 92, and 15 validated, respectively. Elastic net-derived physical activity-metabolomic signatures (TPA-/MVPA-metabolomic signatures) correlated with physical activity intensities (r = 0.20−0.30, P < 0.001) and were associated with reduced stroke risk: TPA-metabolomic signatures (HR = 0.61, 95% CI: 0.44−0.87); MVPA-metabolomic signatures (HR = 0.50, 95%CI: 0.29−0.88). Mediation analyses showed TPA-metabolomic signatures and MVPA-metabolomic signatures explained 12.2% and 8.5% of physical activity-stroke associations ( P < 0.001), implicating specific lipoprotein subclasses and lipids as key mediators.

TPA-metabolomic signatures and MVPA-metabolomic signatures, particularly the 11 key metabolites included, significantly mediate the association between accelerometer-derived physical activity and stroke risk.