Researchers developed a machine learning model using the gradient boosting decision tree algorithm to identify lung cancer patients at high risk for developing COPD, achieving moderate discrimination with eight key predictors including age, smoking history, and systemic coagulation-inflammation index.
- Researchers trained six machine learning models on 1,016 lung cancer patients to predict COPD risk, with GBDT showing the best performance (area under curve of 0.74)
- Eight key predictors identified: age, smoking index, systemic coagulation-inflammation index, eosinophils, bicarbonate, SES, lymphocytes, and hemoglobin
- COPD coexists with lung cancer in many patients, leading to worse clinical outcomes and diagnostic challenges
- The model offers a potential framework for early COPD screening and individualized risk stratification in routine oncology and respiratory care settings
COPD often coexists with lung cancer which naturally leads to worse clinical outcomes. A research team from China aimed to develop a machine learning model for early COPD screening in lung cancer patients using clinical variables and a novel systemic coagulation-inflammation index (SCI).
Their findings were reported in the paper, “Integrating Machine Learning for Early COPD Prediction in Lung Cancer Patients: A Focus on Systemic Coagulation-Inflammation Index,” which was published in the International Journal of Chronic Obstructive Pulmonary Disease.
“Despite its importance, recognizing COPD in patients with lung cancer remains challenging,” the researchers wrote. “COPD often presents insidiously, and symptoms such as cough, sputum production and dyspnea may be attributed to lung cancer itself or to smoking-related respiratory complaints.”
To build their model, the researchers retrospectively enrolled 1,016 patients in the study, extracting demographic, smoking, vitals and laboratory data. The patients came from the department of respiratory and critical care medicine at Enshi Tujia and Miao Autonomous Prefecture Central Hospital between July 2023 and December 2025.
Eligible patients had pathologically- or cytologically-confirmed primary lung cancer and were aged 18 years or older. They then conducted a feature selection with the Boruta algorithm and the least absolute shrinkage and selection operator (LASSO) The researchers then trained six models on the data:
- Logistic regression (LR)
- Decision tree (DT)
- Multilayer perceptron (MLP)
- Support vector machine (SVM)
- Gradient boosting decision tree (GBDT)
- Extreme gradient boosting (XGBoost)
Boruta and LASSO also identified eight key predictors of COPD in lung cancer patients. They are:
- Age
- Historical smoking index
- SCI
- Eosinophils
- Bicarbonate (HCO3)
- Ses
- Lymphocytes
- Hemoglobin
Of the six models, GBDT showed the best performance, with an area under the curve of 0.74 and moderate sensitivity and specificity. The researchers said a GBDT-based model showed moderate discrimination in identifying patients with lung cancer at high risk of concomitant COPD.
“Among the evaluated algorithms, GBDT showed a reasonable overall balance of discrimination, calibration and clinical interpretability,” the researchers wrote. “After external validation, this approach may offer a potential framework for early COPD screening and individualized risk stratification in routine oncology and respiratory care settings.”
