Purpose: The objective of our study was to develop a clinical prediction model for prolonged air leak (PAL) after lobectomy for lung cancer using preoperative variables in a large patient dataset from the National Clinical Database (NCD) in Japan.
Methods: The preoperative characteristics of 57,532 and 30,967 patients who had undergone standard lobectomy for lung cancer were derived from the 2014 to 2015 and 2016 NCD datasets, respectively. PAL was defined as air leak persisting ≥ 7 days postoperatively or requiring postoperative interventional treatment, such as pleurodesis or reoperation. Risk models were developed from the 2014 to 2015 dataset and validated using the 2016 dataset. When performing model derivation, the least absolute shrinkage and selection operator (LASSO) method were applied for parameter selection.
Results: The rate of PAL was 4.5% in 2014-2015 and 5.3% in 2016. The age, sex, body mass index, comorbid interstitial pneumonia, smoking habits, forced expiratory volume in 1 s, tumor histology, multiple lung cancer, and tumor location were selected as important variables for PAL. Our risk model for predicting PAL was fair with a concordance index of 0.6895.
Conclusion: The LASSO-based risk model for PAL after lobectomy provided important preoperative variables for PAL and risk weighting for each variable.
Keywords: Lobectomy; Lung cancer; Prolonged air leak; Risk model.
© 2021. Springer Nature Singapore Pte Ltd.