Biological age construction for prediction of mortality in the Chinese population

Efforts to increase health span bring to light the necessity of constructing biological age (BA) for measuring aging. However, universally adaptive BA needs further investigation, especially among the Chinese population. Therefore, this study aimed to construct BA using routine clinical markers for the Chinese population. Included were two Chinese prospective cohorts, the Kailuan Study I (n = 83,571) for developing BA and the Kailuan Study II (n = 21,229) for validation. Leveraging baseline age-related clinical markers, we developed phenotypic BA (Pheno-Age) using Levine's methods and Klemera-Doubal BA (KDM-Age) using KDM methods and calculated the residuals of regressions of the two BA measured at baseline and during follow-up on chronological age, namely BA acceleration. The predictive performance of baseline, cumulative average, and updated BAs on mortality was evaluated using the area under the curve (AUC) and calibration plots. COX regressions were used to estimate hazard rations (HRs) and 95% confidence intervals (CIs) for the BA acceleration and risk of mortality. During 1,443,857 person-years of follow-up, 12,679 deaths were recorded in the two cohorts. Baseline Pheno-Age and KDM-Age produced desirable predictions for mortality in both the Kailuan Study I (AUC, 0.810 and 0.806, respectively) and the Kailuan Study II (AUC, 0.867 and 0.819, respectively). Calibration plots showed reasonable agreement between predicted and observed probabilities. The pooled multivariable-adjusted HRs (95% CIs) for per standard deviation increment of baseline Pheno-Age acceleration and mortality was 1.24 (1.18, 1.30), and for KDM-Age acceleration was 1.16 (1.10, 1.21). Similar predictive performance and association were observed when using cumulative average or updated BA. The associations were stronger in the adults aged ≤60 years, smokers, and drinkers, relative to their counterparts (P for interaction <0.05 for all). Pheno-Age and KDM-Age, developed and validated in the two large prospective cohorts, could predict mortality, independent of chronological age and other potential confounders, in Chinese populations.