A six-metabolite signature characterizes metabolically unhealthy obesity and reveals hidden metabolic risk within metabolically healthy obesity
Metabolically healthy obesity (MHO) is commonly defined by the absence of metabolic syndrome-related abnormalities despite obesity. However, conventional clinical definitions may overlook substantial metabolic heterogeneity and hidden cardiometabolic risk. We aimed to identify metabolomic signatures distinguishing MHO from metabolically unhealthy obesity (MUHO), evaluate their discriminatory performance, and determine whether metabolomic profiling could further characterize heterogeneity within conventionally defined MHO. We analyzed 13215 UK Biobank adults with obesity and available clinical biomarker and NMR-based metabolomic data. Metabolic health was defined using triglycerides, HDL cholesterol, hypertension, fasting glucose, type 2 diabetes, and lipid-lowering medication use. Univariable logistic regression and LASSO regression were used for metabolite selection. Logistic regression and XGBoost models were developed using clinical variables, metabolomic markers, and their combination. A weighted metabolic signature score was applied within the MHO group to characterize cross-sectional metabolic and clinical heterogeneity, and proteomic analyses were performed in approximately 1408 participants. A six-metabolite signature comprising HDL_size, S_HDL_CE, XL_HDL_TG, GlycA, M_VLDL_C, and Omega_3 was selected. The combined clinical-metabolomic model showed better discrimination than clinical variables alone in the test set, with AUCs of 0.78 and 0.69, respectively. Within MHO, higher metabolomic score was associated with higher triglycerides, HbA1c, waist-to-hip ratio, lower HDL cholesterol, and greater metabolic and cardiovascular comorbidity burden. Proteomic analyses identified 10 metabolite-associated core proteins implicating lipoprotein remodeling, adipokine signaling, inflammation, and vascular-related pathways. A six-metabolite signature distinguished MHO from MUHO and revealed hidden metabolic risk within conventionally defined MHO, provides a metabolomic framework for refining obesity phenotyping and warrants further validation before clinical translation.