A multi‐omics risk score enables early and personalized screening for type 2 diabetes: A population‐based cohort study

OBJECTIVE: This study aims to develop a multi-omics model to predict type 2 diabetes onset using multi-omics data, beyond established clinical and genetic factors. METHODS: In the UK Biobank cohort (N = 21,312,893 incident type 2 diabetes cases), Ridge Cox regression models were constructed based on clinical factors defined by the Finnish Diabetes Risk Score (CliS), plasma metabolomics (MetS), proteomics (ProS) and polygenic risk score (PRS). Model performance was evaluated by C-index and net reclassification improvement (NRI). Clinical utility was assessed through net benefit, risk stratification and screening initiation ages. Independent validation of protein markers was performed in the Liyang Cohort (N = 10,056). RESULTS: The ProS was the top-performing single-omics predictor (C-index: 0.80). ComS (Combined risk score) achieved the highest discriminative ability (C-index: 0.84), significantly outperforming the CliS (C-index: 0.76, NRI: 0.328, p < 0.001). ComS reclassified 73 additional type 2 diabetes cases per 1000 participants not receiving intervention, without increasing false positives. Kaplan-Meier analysis confirmed superior risk stratification of ComS (Log-rank p < 0.0001). Critically, both MetS and ProS identified high-risk individuals missed by the conventional CliS. Screening initiation before age 40 was warranted for individuals in the top risk quintile of MetS, ProS, or ComS. In the Liyang Cohort, plasma FGF23 levels were significantly elevated in type 2 diabetes cases (p < 0.05), corroborating its role as a key proteomic contributor to risk prediction. CONCLUSION: The combined multi-omics model enables more precise, earlier type 2 diabetes risk stratification, supporting personalized screening strategies years before clinical onset.