Abstract / Summary
Obesity is a major risk factor for cardiometabolic diseases, but individuals with obesity differ substantially in their future disease trajectories. We aimed to evaluate whether circulating proteins provide incremental and biologically interpretable risk information for incident type 2 diabetes, hypertension, and hyperlipidemia among individuals with an operationally defined pre-clinical obesity phenotype in UK Biobank. Among 13,931 UK Biobank participants with abnormal adiposity but without established obesity-related diseases at baseline, 2,919 circulating proteins were analyzed. Single-protein associations with incident type 2 diabetes, hypertension, and hyperlipidemia were assessed using Cox proportional hazards models. Protein-based risk scores (ProteinScores) were developed using elastic-net Cox regression for 5-, 10-, and 15-year prediction horizons. Predictive performance was evaluated using AUC, calibration, and category-based net reclassification improvement, and compared with baseline covariate models, extended clinical models, available UK Biobank polygenic risk scores, and metabolomic scores. Pathway enrichment analyses were performed to characterize the biological processes represented by ProteinScore proteins. We identified 973 Bonferroni-significant protein–disease associations involving 653 unique proteins across the three cardiometabolic outcomes. ProteinScores improved discrimination beyond age- and sex-adjusted and baseline covariate models across prediction horizons. Compared with extended clinical models that included disease-specific clinical indicators, incremental improvement was more outcome- and horizon-dependent, with significant AUC gains observed for selected longer-term predictions. Calibration and reclassification analyses showed that ProteinScores provided additional risk-stratification information in several disease–horizon combinations, although the magnitude of improvement varied. ProteinScores showed favorable performance compared with available PRS fields and MetaboScores in this UK Biobank framework, but the relative performance differed by outcome and prediction horizon. Pathway analyses implicated lipid transport, lipoprotein remodeling, extracellular matrix organization, peptide hormone processing, and immune-inflammatory signaling. Plasma proteomic profiling provides effective risk stratification for future cardiometabolic diseases in individuals with pre-clinical obesity. These findings highlight the potential of integrating dynamic proteomic signatures into early clinical frameworks to advance precision prevention strategies in obesity management.