Exercise intensity-dependent metabolic signatures and cardiovascular protection in type 2 diabetes: evidence from NMR metabolomics in the UK Biobank

Background Physical activity reduces cardiovascular risk in type 2 diabetes mellitus (T2DM), yet the intensity-specific metabolic mechanisms underlying this protection remain poorly characterized. Methods Using the UK Biobank prospective cohort, we included 6,887 T2DM patients with complete physical activity and plasma NMR metabolomics data. Three intensity-specific metabolic signature scores (walking, moderate-, and vigorous-intensity) were constructed via elastic net regression. Restricted cubic spline Cox regression was used to model dose–response relationships between total physical activity and major adverse cardiovascular events (MACE). Multi-level mediation analyses quantified the proportion of the exercise–MACE association statistically accounted for by each intensity-specific metabolic pathway. Results Over a median follow-up of 15.17 years, 974 MACE events occurred. Total physical activity was inversely associated with MACE risk in a dose–response manner (P overall = 0.007), plateauing at approximately 3,500–4,000 MET-min/week. Compared with the low activity group, moderately active and highly active groups showed 17% (HR = 0.829, 95% CI: 0.711–0.966) and 27% (HR = 0.734, 95% CI: 0.620–0.870) reductions in MACE risk, respectively. All three intensity-specific metabolic signature scores were independently associated with reduced MACE risk (all p < 0.001). GlycA and albumin—both selected into all three intensity-specific scores—showed the strongest nominal mediation signals among the 41 candidate metabolites tested; glucose, also selected into all three scores, showed a smaller but nominally significant signal, together consistent with a shared contribution from systemic inflammation, hepatic synthetic capacity, and glycemic control across exercise intensities. In joint mediation modeling, the three metabolic scores together accounted for 26.1% (95% CI: 15.1%–48.8%) of the exercise–MACE association, led by the vigorous-intensity pathway (proportion mediated = 22.7%). Because the three scores were highly correlated ( r = 0.78–0.84), we additionally modeled each individually: all three intensities, including walking (proportion mediated = 18.9%), showed significant metabolic mediation (all p < 0.001), indicating that exercise at every intensity engages overlapping metabolic pathways. Conclusions Physical activity is inversely associated with MACE risk in T2DM through multidimensional, overlapping metabolic pathways spanning inflammation, glycemic control, and lipid remodeling. The vigorous-intensity pathway, plausibly driven by lipid remodeling, statistically accounted for the largest share of this association, while walking retained the strongest independent direct association with MACE risk alongside a meaningful metabolic contribution of its own. These findings advance the mechanistic understanding of exercise-induced cardioprotection in T2DM and provide a metabolomics-informed basis for intensity-stratified exercise guidance.