Risk factors associated with prolonged air leak after video-assisted thoracic surgery pulmonary resection: a predictive model and meta-analysis.

BACKGROUND

This study aimed to establish a predictive model for prolonged air leak (PAL) after video-assisted thoracic surgery (VATS) lung resection; and additionally, to present a meta-analysis of the relevant literature to estimate the association between various clinical factors and PAL.

METHODS

A retrospective, case-control study was conducted using univariate analysis and logistic regression based on 493 medical records from patients who underwent VATS lung resection between January 2015 and August 2017 at our institution. PAL was defined as air leak more than 5 days after lung surgery. Subsequently, a nomogram was established as a predictive model. Relevant studies were screened from PubMed, Embase and Cochrane for relevant studies and data was extracted from those enrolled. Pooled odds ratios or weighted mean differences with corresponding 95% confidence intervals were calculated to estimate the association between various clinical factors and PAL.

RESULTS

Incidence of PAL after VATS lung resection was observed in 54 (10.8%) of 493 patients. Logistic regression revealed that smoking (P=0.014), pulmonary function (P=0.011), pleural adhesion (P<0.001), stapling length (P<0.001), early postoperative drainage (P=0.002) were significantly associated with PAL. Our meta-analysis, including 17 eligible studies and 14 potential risk factors, further validating our findings. Upper lobectomy was determined to be a significant risk factor for PAL in Europeans and North Americans (OR =2.03, P<0.001), but not in Asians (OR =1.04, P=0.610). Importantly, the constructed nomogram demonstrated a good predictive ability (C-index =0.858).

CONCLUSIONS

Lung stapling length and early postoperative drainage are important indicators for the evaluation of PAL occurrence. Upper pulmonary resection is a factor with particular regional differences as its association with PAL is not significant within Asian populations. Our nomogram, incorporating multiple factors, provided a simple and practical predictive model with value for clinical application.