Predictors of cough resolution following endoscopic minimally invasive treatment in patients with GERD and chronic cough: a retrospective study.

OBJECTIVES: To investigate factors influencing cough resolution following endoscopic minimally invasive treatment in patients with gastroesophageal reflux disease (GERD) complicated by chronic cough, and to establish a predictive model based on these findings.

METHODS: Clinical data from 200 patients with GERD and cough undergoing endoscopic minimally invasive treatment at our hospital between January and December 2024 were included. All patient data were extracted from the electronic medical record system. Patients were divided into a cough non-relief group (n = 35) and a cough-relief group (n = 165) based on validated cough score reduction criteria (≥50% reduction in total cough score) post-treatment. Clinical data were collected and compared. Two machine learning methods-Least Absolute Subtraction and Selection (LASSO) regression and Extreme Gradient Boosting (XGBoost)-were employed for "intersection screening" of risk factors. Multivariate logistic regression analyzed risk factors for persistent cough after endoscopic minimally invasive treatment in GERD patients with chronic cough. A regression model was established using R software, validated internally via Bootstrap resampling (1,000 iterations), and evaluated for discriminatory power and consistency using receiver operating characteristic (ROC) curve area under the curve (AUC) and calibration curves.

RESULTS: Compared to the cough-relief group, the non-relief group exhibited significantly higher age, smoking history, cough duration, ineffective esophageal motility, DeMeester score, esophageal dysfunction, pepsin, and exhaled nitric oxide (FeNO), along with lower secretory immunoglobulin A (SIgA) levels (P < 0.05). LASSO regression and XGBoost with "overlapping coverage" identified seven common risk factors: age, cough duration, ineffective esophageal motility, DeMeester score, pepsin, FeNO, and SIgA. Multivariate logistic regression revealed that ineffective esophageal motility, cough duration, DeMeester score, pepsin, and FeNO were risk factors for persistent cough after endoscopic minimally invasive therapy (OR = 27.090, 2.639, 1.193, 1.106, 1.295, P < 0.05), while SIgA was a protective factor (OR = 0.891, P < 0.05). The Hosmer-Lemeshow test yielded χ2 = 1.535, P = 0.992 > 0.05, indicating good model calibration. A nomogram model was constructed based on multivariate regression analysis results. Model validation was performed using Bootstrap sampling (1,000 iterations), yielding a ROC-AUC of 0.967 (95% CI 0.936-0.998). The calibration curve demonstrated good agreement between predicted and actual values.

CONCLUSION: Independent risk factors influencing persistent cough after endoscopic minimally invasive treatment in GERD patients with chronic cough include ineffective esophageal motility, cough duration, DeMeester score, SIgA, pepsin, and FeNO. The nomogram model constructed based on these factors demonstrates good predictive efficacy for clinical treatment outcomes.