Development of a multivariable prediction model for prolonged air leak after lung resection

Objectives: Prolonged air leak (PAL) is a common complication of lung resection. Research on predictors of PAL using a digital drainage system (DDS) remains insufficient. In this study, we investigated the predictive factors of PAL to establish a novel early postoperative prediction model for PAL.

Methods: A retrospective cohort study and validation study were conducted. We examined patients who underwent lung resection with DDS at our institute. The relationship between the clinical factors and measurements of the DDS, including the difference between the set and measured intrapleural pressure (named: additional negative pressure [ANP]) at postoperative hour (POH) 3, with PAL was analyzed.

Results: A total of 494 patients were enrolled, 29 of whom had PAL. Percent forced expiratory volume in 1 s <60%, ANP <1 cmH₂O, air leak flow >20 mL/min and pleural adhesion findings at surgery were independent predictors of PAL according to a multivariable analysis. The PAL rate was clearly stratified according to our novel risk scoring system, which simply notes the presence of the above four factors, that is, the rate increases when the score increases. The area under the curve (AUC) of the receiver operating characteristic (ROC) analysis for this scoring system was 0.818. Analysis of the validation cohort (n = 133) revealed that this scoring system showed a sufficient ability to predict PAL.

Conclusions: ANP at POH 3 is an independent predictor of PAL. Thus, the risk-scoring system proposed in this study is useful for predicting PAL in the early postoperative period.