Development of Age- and Sex-Specific Metabolomics-Based Biological Ageing Clocks for 10-Year Mortality Prediction.

Metabolite concentrations vary by age and sex, yet age- and sex-specific metabolomic risk scores and biological ageing clocks for mortality prediction remain undeveloped. Nuclear magnetic resonance (NMR)-based metabolomic profiling is conducted in 209144 UK Biobank participants (12347 deaths) and 6820 from the German ESTHER study (804 deaths). Mortality risk scores are derived using least absolute shrinkage and selection operator (LASSO)-regularized Cox regression, and metabolomics-based mortality risk clocks (MetaboMR clocks) are constructed using elastic net regression in sex- and age-stratified subgroups (50-59 and 60-69 years). Models are trained in 70% of UK Biobank and validated internally (30%) and externally in ESTHER. 68 metabolites are significantly associated with 10-year all-cause mortality in both cohorts. 20, 18, 12, and 13 metabolites improved 10-year mortality prediction in younger and older men, younger and older women. Metabolite-augmented models improved c-statistics by 0.036-0.084 across subgroups. In the external validation set, each year of age acceleration is associated with an 8% and 9% higher 10-year mortality risk for MetaboMR clock1 and clock2. Sex- and age-specific metabolomic risk scores significantly enhance 10-year mortality prediction beyond traditional models. The MetaboMR clocks may serve as measures of biological ageing and support personalized risk stratification in clinical settings.