Construction and validation of a predictive model for postoperative respiratory failure in esophageal cancer patients

Postoperative respiratory failure (PRF) is one of the most severe complications following esophageal cancer (EC) surgery, closely associated with high mortality and poor prognosis. Early diagnosis and intervention are crucial. This study aimed to explore the risk factors for PRF in EC, develop a predictive model, and validate its performance.

The clinical data of 265 EC patients who underwent surgery at the Sun Yat-sen University Cancer Center Gansu Hospital between January 2020 and June 2024 were retrospectively analyzed. The patients were randomly divided 7:3 into a training set (n=185) and an internal validation set (n=80). Another 80 EC patients who underwent surgery at the Sun Yat-sen University Cancer Center between January 2024 and June 2024 were employed as an external validation set. Feature selection was optimized using least absolute shrinkage and selection operator (LASSO)-logistic regression, and a predictive model was constructed and internally and externally validated.

Smoking index ≥400, forced expiratory volume in one second (FEV1), preoperative serum albumin level, surgical time, and postoperative anastomotic fistula were identified as risk factors for PRF in EC patients. The area under the curve (AUC) values of the predictive model were as follows: training set (0.856), internal validation set (0.839), and external validation set (0.773), indicating that the model had good discriminatory power. A calibration curve and Hosmer-Lemeshow test demonstrated that the model had favorable predictive accuracy and decision curve analysis (DCA) showed that the model had considerable clinical utility.

The predictive model developed using LASSO-logistic regression exhibited strong performance and clinical applicability in both internal and external validations, with the potential to assist clinicians in identifying high-risk patients for early individualized intervention.