Respiratory medicine · 2025
Predicting response to inhaled corticosteroid maintenance therapy in patients with chronic obstructive pulmonary disease using machine learning models.
СтатьяПоддержка исследования: вне правительства СШАНазвание и аннотация: Перевод пока не подготовлен
Обработка статьи · этапы, ошибки и управление
Загрузка состояния…
Ошибки и ограничения
История · было → стало
Состояние обновляется каждые 5 секунд, пока блок открыт. Процент показывается только там, где известен объём работы. Пауза останавливает продолжение этапа; текущий запрос к источнику может завершиться.
Wu Shan-Chieh, Wu Chih-Ying, Chien Jung-Yien, Dong Yaa-Hui, Lin Fang-Ju
Аннотация
BACKGROUND: Blood eosinophil count and exacerbation history are established predictors of inhaled corticosteroid (ICS) effectiveness in chronic obstructive pulmonary disease (COPD). However, treatment responsiveness is heterogeneous and influenced by additional clinical characteristics. This study aimed to develop a machine learning-based prediction model to identify predictors of response to ICS in COPD patients.
METHODS: Using a nationwide administrative database linked with individual laboratory results, we identified COPD patients initiating ICS between 2015 and 2019. Patients were stratified into low- and high-exacerbation-risk groups based on prior exacerbation frequency. Prediction models for favorable ICS response were developed using logistic regression, lasso regression, and extreme gradient boosting (XGBoost). Model performance was assessed by receiver operating characteristic (ROC) curves and calibration plots. Key predictors were identified using Shapley Additive exPlanations.
RESULTS: Among 23,587 ICS-naïve patients, favorable ICS response rates were 73.7 % in the low-risk group and 59.1 % in the high-risk group. XGBoost model outperformed other models in discriminative ability, achieving an area under the ROC curve of 0.72 (95 % CI, 0.70-0.74) for the low-risk group and 0.67 (95 % CI, 0.64-0.70) for the high-risk group in the validation dataset. Younger age, male sex, comorbid asthma, and lower prior use of COPD-related medications were significant predictors of ICS response. The relationship between prior exacerbations on ICS response varied between risk groups. Elevated blood eosinophil levels demonstrated relatively limited predictive ability.
CONCLUSIONS: Machine learning identified potential predictors of ICS response in COPD patients, which may inform future efforts to enhance personalized treatment strategies based on risk profile.
METHODS: Using a nationwide administrative database linked with individual laboratory results, we identified COPD patients initiating ICS between 2015 and 2019. Patients were stratified into low- and high-exacerbation-risk groups based on prior exacerbation frequency. Prediction models for favorable ICS response were developed using logistic regression, lasso regression, and extreme gradient boosting (XGBoost). Model performance was assessed by receiver operating characteristic (ROC) curves and calibration plots. Key predictors were identified using Shapley Additive exPlanations.
RESULTS: Among 23,587 ICS-naïve patients, favorable ICS response rates were 73.7 % in the low-risk group and 59.1 % in the high-risk group. XGBoost model outperformed other models in discriminative ability, achieving an area under the ROC curve of 0.72 (95 % CI, 0.70-0.74) for the low-risk group and 0.67 (95 % CI, 0.64-0.70) for the high-risk group in the validation dataset. Younger age, male sex, comorbid asthma, and lower prior use of COPD-related medications were significant predictors of ICS response. The relationship between prior exacerbations on ICS response varied between risk groups. Elevated blood eosinophil levels demonstrated relatively limited predictive ability.
CONCLUSIONS: Machine learning identified potential predictors of ICS response in COPD patients, which may inform future efforts to enhance personalized treatment strategies based on risk profile.