Postoperative pulmonary complications after esophagectomy: risk factors and prediction model.

BACKGROUND

Postoperative pulmonary complications (PPCs) affect up to one-third of patients undergoing esophagectomy and remain a major contributor to postoperative morbidity. This study aimed to identify pre- and perioperative risk factors for PPCs and to develop a predictive model.

METHODS

This retrospective cohort study included patients who underwent esophagectomy for esophageal or gastroesophageal junction cancer at Amsterdam UMC between 2013 and 2023. PPCs included pneumonia, pleural effusion, pneumothorax, atelectasis, respiratory failure, aspiration, acute respiratory distress syndrome, tracheobronchial fistula, and persistent air leakage. Univariable and multivariable logistic regression with backward selection were used to identify predictors. Model performance was assessed with the area under the receiver operating characteristic curve (AUC). Statistical significance was set at P < 0.05.

RESULTS

Among 960 patients, 254 (26.5%) developed at least one PPC. Independent predictors were smoking status (former: OR 1.45, 95% CI 1.03-2.06; current: OR 1.74, 95% CI 1.15-2.63), non-epidural analgesia (paravertebral: OR 1.50, 95% CI 1.03-2.17; other: OR 1.57, 95% CI 0.88-2.82), and cervical versus intrathoracic anastomosis (OR 1.64, 95% CI 1.17-2.29). Drain configuration also influenced the risk of PPC: one-sided double drains were protective (OR 0.48, 95% CI 0.25-0.91), whereas bilateral drains increased the risk (OR 2.60, 95% CI 1.38-4.87), compared with one-sided single drains. The model demonstrated modest discrimination after validation of AUC: 0.598.

CONCLUSION

Smoking, paravertebral analgesia, cervical anastomosis, and bilateral drains were independently associated with increased PPC risk. Although predictive performance was modest, these modifiable and structural factors inform perioperative management. Future models should incorporate intraoperative and physiological variables to improve risk stratification.