Abstract / Summary
Aims: Cardiovascular-kidney-metabolic (CKM) syndrome represents an integrated continuum of metabolic, kidney, and cardiovascular abnormalities. However, the stage-specific metabolic heterogeneity underlying this clinical framework remains incompletely characterised.
Materials and methods: We performed plasma metabolomic profiling in 1374 participants from the China Multi-Ethnic Cohort across CKM stages 0-4. Participants from Chengdu and Chongqing provinces comprised the discovery cohort (n = 969), whereas those from Yunnan province comprised the validation cohort (n = 405). Stage-associated metabolites were identified using prespecified pairwise comparisons with OPLS-DA and covariate-adjusted limma analyses. Weighted correlation network analysis identified coordinated metabolic modules. Machine-learning models were developed to evaluate discrimination of advanced CKM (stages 3-4) using metabolite signatures independent of conventional CKM-defining variables.
Results: Among 858 endogenous metabolites, 261 unique metabolites were associated with CKM stages, revealing distinct metabolic patterns from early to advanced CKM. Stage 1 was characterised by altered lipid- and bile acid-related metabolites, stage 2 by broader lipid and amino-acid remodelling, and stage 3 by additional carbohydrate, aromatic amino-acid, secondary bile acid, host-microbial, and renal-handling signals. Metabolic separation between stages 3 and 4 was comparatively weak. WGCNA identified a CKM-associated turquoise module, with γ-glutamylvaline as a hub metabolite. A model incorporating age, sex, and eight metabolites derived from the discovery cohort showed favourable performance for distinguishing advanced CKM from earlier stages and achieved an AUROC of 0.874 (95% CI, 0.802-0.931) in the validation cohort.
Conclusions: CKM stages exhibit distinct and non-linear metabolic signatures. A compact metabolomic panel independent of conventional CKM-defining variables may provide complementary molecular information for advanced CKM phenotyping and future risk stratification.