Metabolomic signatures predict seven-year mortality in clinically stable COPD patients

Abstract

Chronic Obstructive Pulmonary Disease (COPD) is a complex condition with high mortality. Early identification of patients at increased risk of death remains a major clinical challenge. This pilot study aimed to explore whether plasma metabolomic profiling could aid in the prediction of long-term (7-year) mortality and provide insight into potential underlying mechanisms. Plasma samples from 54 randomly selected stable COPD patients were analyzed using both untargeted and semi-targeted LC-MS approaches. After excluding patients with unclear death data, non-COPD-related deaths and metabolomic outliers, 41 individuals were included in the final analysis. During follow-up, 13 patients (32%) died, and 28 survived. Univariate analysis identified 12 metabolites—mainly amino acids—that differed significantly between the two groups. Functional analysis suggested a significant disruption in energy production pathways. Predictive models developed using machine learning algorithms, consisting of either ten metabolites alone or nine metabolites plus FEV1, achieved high accuracy for 7-year mortality prediction, with the latter model performing slightly better. Internal validation was conducted using five-fold cross-validation. While exploratory, these findings support the hypothesis that early metabolic alterations, particularly in energy pathways, may contribute to long-term mortality risk in stable COPD patients, and could complement traditional prognostic markers such as FEV1.