Risk stratification is essential for identifying high-risk individuals and disease prevention. Researchers investigated the potential of nuclear magnetic resonance (NMR) spectroscopy-derived metabolomic profiles to provide information on multi-disease risk. They conducted the research in addition to conventional clinical predictors for the onset of 24 common conditions, including metabolic, vascular, respiratory, musculoskeletal, neurological, and cancer diseases.
The researchers specifically trained a neural network to learn disease-specific metabolomic states from 168 circulating metabolic markers measured in 117,981 UK Biobank participants with 1.4 million person-years of follow-up and validated the model in four independent cohorts.
Except for breast cancer, the researchers discovered that metabolomic states were associated with incident event rates in all the studied conditions. A combination of age, gender, and metabolomic state equaled or outperformed established predictors for 10-year outcome prediction for 15 endpoints, with and without established metabolic contribution.
Furthermore, the metabolomic state provided predictive information for eight common diseases, including type 2 diabetes, dementia, and heart failure. According to decision curve analyses, predictive improvements translated into clinical utility for a wide range of potential decision thresholds. The findings show the potential and limitations of NMR-derived metabolomic profiles as a multi-disease assay for simultaneously predicting the risk of multiple common diseases.