Biomarker-driven prediction of airway obstruction severity and exacerbation in COPD using machine learning

Abstract

Background
Chronic obstructive pulmonary disease (COPD) represents a global health challenge, with acute exacerbations contributing to increased morbidity and mortality. Inflammation is a hallmark of COPD. This study employed machine learning techniques to assess the potential of serum levels of Clara cell secretory protein (CCSP-16), interleukin-8 (IL-8), and interleukin-6 (IL-6) for evaluating disease severity and acute exacerbations in COPD patients.


Methods
A cross-sectional study included 80 male COPD patients and 60 matched controls. Data collected included demographic, clinical, and spirometry results, along with serum biomarker levels measured via ELISA. Analyses combined conventional statistical methods with machine learning-based feature ranking to identify predictors of severe airway obstruction and exacerbation.


Results

COPD patients had significantly lower levels of CCSP-16 and IL-8 compared to healthy controls, while IL-6 levels were markedly higher in COPD patients. Using conventional statistical analyses, elevated CCSP-16 levels were significantly associated with severe airway obstruction (OR 1.87,
p
 = 0.04) and exacerbations (OR 1.72,
p
 = 0.01). Combining all three biomarkers improved the discrimination for severe airway obstruction (AUC = 69%,
p
 = 0.03) and exacerbation events (AUC = 73%,
p
 < 0.001). Feature ranking using multiple machine learning models further highlighted the integration of clinical factors with biomarker data. This approach achieved cross-validated ROC-AUC values of 0.94 and 1.0 for association with severe airway obstruction and exacerbations, respectively.



Conclusions
This study demonstrates the potential of CCSP-16, IL-6, and IL-8 as promising non-invasive biomarkers for COPD, as shown by both conventional statistical analyses and feature ranking using machine learning models. The integration of these biomarkers with clinical factors enhanced the predictive accuracy for severe airway obstruction and exacerbation risks. These findings highlight the critical role of combining clinical and inflammatory markers with advanced analytical techniques in improving the management of COPD.