Predictive Value of Biological Age for All-Cause Mortality in Patients with COPD

Although biological age has emerged as a robust predictor of mortality across multiple chronic diseases, its prognostic value in patients with COPD remains insufficiently defined. This study aimed to compare the predictive performance of major biological age biomarkers and to develop a mortality risk prediction model for COPD based on the optimal biomarker.

We included 19,882 patients with COPD who had complete survival data extending beyond 10 years, recruited between 2006 and 2010 from a large biobank cohort. Biological aging was quantified using phenotypic age (PhenoAge) and Klemera-Doubal Method biological age (KDM-BA). Time-dependent ROC analyses were performed to compare the discriminatory ability of these biomarkers for predicting all-cause mortality. Cox regression was subsequently developed to construct mortality prediction models, with time-dependent ROC, net reclassification index (NRI) and integrated discrimination improvement (IDI) analyses evaluating model performance. Decision curve analysis(DCA) was then performed to evaluate the clinical benefit of each model.

Among 19,882 patients with COPD, the mean chronological age was 59.9 ± 7.1 years, with a mean PhenoAge of 53.4 ± 9.4 years and a mean KDM-BA of 58.9 ± 12.8 years. Time-dependent ROC analyses showed that PhenoAge consistently outperformed chronological age, PhenoAge acceleration, and KDM-BA in predicting all-cause mortality at 5-, 10-, and 15-year follow-up. Based on univariate analyses and assessment of collinearity among pulmonary function variables, two Cox models were constructed. Models incorporating PhenoAge demonstrated higher AUCs, as well as positive NRI and IDI values, compared with chronological age-based models. DCA curves further indicated that PhenoAge-derived models provided greater net clinical benefit across most threshold probabilities.

PhenoAge outperformed chronological age and KDM-BA in predicting all-cause mortality in COPD. PhenoAge-based models provided modest but consistent improvements in risk prediction over models based on chronological age and may complement individualized risk stratification.