Metabolomics-Defined Subtypes of Prediabetes and Risk of Cardiovascular-Kidney-Metabolic Outcomes

Abstract

Prediabetes is highly prevalent and biologically heterogeneous, yet current glycemic definitions do not adequately capture differences in cardiometabolic and kidney risk. We aimed to identify metabolically defined subtypes of prediabetes using circulating metabolites and to examine their associations with cardiovascular-kidney-metabolic outcomes. We analyzed 24,638 participants with prediabetes from the UK Biobank who had metabolomics data available. Metabolomic biomarkers related to type 2 diabetes, cardiovascular disease, and chronic kidney disease were identified using machine learning–based feature selection methods, and unsupervised clustering was applied to derive metabolic subtypes. Associations between subtypes and cardiometabolic outcomes were evaluated, and interactions between dietary patterns and metabolic subtypes were explored. Mendelian randomization analyses were conducted to investigate potential causal roles of key metabolomic biomarkers. Three metabolically distinct subtypes of prediabetes were identified, representing low-, intermediate-, and high-risk metabolic profiles. These subtypes showed progressively higher risks of developing type 2 diabetes, cardiovascular disease, and chronic kidney disease during follow-up, and the associations between diet quality and disease outcomes differed across subtypes. Several metabolomic biomarkers demonstrated potential causal links with cardiometabolic outcomes. These findings highlight the metabolic heterogeneity of prediabetes and suggest that metabolomics-based subtypes may improve risk stratification and support precision prevention strategies. Article Highlights Previous studies have identified heterogeneity among prediabetes subgroups using clinical characteristics; however, biological and metabolic heterogeneity remains insufficiently captured. This study examined whether data-driven clustering based on metabolomic biomarkers could define distinct prediabetes subtypes with differential type 2 diabetes, cardiovascular disease, and chronic kidney disease risk. Using 16 metabolomic biomarkers, we identify three metabolically distinct clusters showing progressively higher risks of incident type 2 diabetes, cardiovascular disease, and chronic kidney disease. Differential diet-cluster associations across clusters were obtained, and Mendelian randomization supported potential causal roles for several metabolomic biomarkers. Metabolomics-based stratification may improve risk prevention and enable cluster-specific dietary interventions in prediabetes.