Frontiers in Endocrinology · Nov 15, 2026

Phenotypic heterogeneity within operationally defined cardiovascular kidney metabolic syndrome stage 2 across three cohorts and an independent clinical sample

By Shengyi Lei, Jinyuan Shen, Xu Wu et al.

Background: We examined phenotypic heterogeneity within cardiovascular-kidney-metabolic (CKM) stage 2 and assessed structural correspondence and clinical associations in independent populations. Methods: An eight-indicator, three-class latent profile model was derived in NHANES (2003–2018; n=8,654), applied as a fixed classifier in CHARLS (n=5,320), and evaluated structurally in LASI-DAD (n=1,891). Frozen-classifier application in 197 clinical patients used 50 multiply imputed datasets to examine cross-sectional carotid plaque associations. Results: Higher HDL-C/lower BMI, adiposity-related dyslipidemia, and hyperglycemia profiles represented 34.6%, 51.8%, and 13.6% of the weighted NHANES sample and showed nonparallel indicator patterns. Hyperglycemia was associated with NHANES all-cause mortality (HR 1.24, 95% CI 1.04–1.49) and CHARLS incident cardiovascular disease (HR 1.43, 95% CI 1.15–1.78). LASI-DAD frozen-classifier profile correlations with NHANES were 0.717, 0.775, and 0.951. Independent LASI-DAD analysis retained the hyperglycemia configuration (r=0.926), although fixed-versus-independent classification agreement was limited in both cohorts (ARI: CHARLS=0.113; LASI-DAD=0.183). Among 162 clinical patients with interpretable ultrasound, 73 had plaque. Age-and sex-adjusted odds ratios were 2.90 (95% CI 1.02–8.26) for adiposity-related dyslipidemia and 3.56 (1.02–12.35) for hyperglycemia (both nominal P=0.046; Holm-adjusted P=0.091). Complete-case estimates were directionally consistent but attenuated and imprecise. Conclusions: CKM stage 2 contains nonparallel cardiovascular, metabolic, and kidney-related configurations. Cross-population correspondence was strongest for the hyperglycemia-dominant profile. Profiles provided negligible incremental prediction beyond HbA1c and conventional covariates; clinical plaque associations remain exploratory.

The paperJournal siteBiomarkers

AgingResearch.News