Abstract
BACKGROUND: Metabolic dysregulation in patients with diabetes mellitus (DM) may contribute to pulmonary vascular remodeling, yet biomarkers for pulmonary hypertension (PH) risk in diabetes remain unclear. OBJECTIVES: To assess the association between DM and incident PH, quantify mediation by circulating metabolites, and develop a metabolite-based PH risk model in diabetes. METHODS: In a prospective cohort of 459,714 UK Biobank participants without PH at baseline, Cox proportional hazard models estimated the association between DM and incident PH. Nuclear magnetic resonance (NMR) metabolomics, mediation analysis, and LASSO regression were used to identify metabolic mediators and construct a predictive model. RESULTS: During a median follow-up of 13.25 years, 2,250 PH cases occurred. DM was associated with a 39.6% higher PH risk (HR = 1.396, 95%CI:1.255–1.553; P < 0.001), and earlier DM onset was associated with stronger risk (P < 0.001). Seven NMR-derived metabolites—related to hyperglycemia, inflammation (glycoprotein acetyls), amino acid and lipid metabolism—mediated 37.26% of the total effect. Including these metabolites improved prediction in diabetes (C index from 0.692 to 0.735; P < 0.001). CONCLUSIONS: DM is an independent risk factor for PH. Integration of NMR-derived metabolites improves risk prediction and suggests hyperglycemia and inflammation as key mechanisms, supporting the role of metabolomics in PH stratification and prevention among diabetes. SUPPLEMENTARY INFORMATION: The online version contains supplementary material available at 10.1186/s12931-026-03499-4.