Arnaud Pforr1, Pierre-Benoit Pagès2, Jean-Marc Baste3, Pascal Thomas4, Pierre-Emmanuel Falcoz5, Francoise Lepimpec Barthes6, Marcel Dahan7, Alain Bernard1. 1. Centre Hospitalier Universitaire (CHU) Dijon, Bocage Hospital, Dijon, France. 2. Centre Hospitalier Universitaire (CHU) Dijon, Bocage Hospital, Dijon, France. Electronic address: pierrebenoit.pages@chu-dijon.fr. 3. CHU Rouen, Charles Nicolle Hospital, Rouen, France. 4. CHU Marseille, North Hospital, Marseille, France. 5. CHU Strasbourg, Civil Hospital, Strasbourg, France. 6. Georges Pompidou European Hospital, Paris, France. 7. CHU Toulouse, Larrey Hospital, Toulouse, France.
Abstract
BACKGROUND: Bronchopleural fistula (BPF) remains a rare but fatal complication of thoracic surgery. The aim of this study was to develop and validate a predictive model of BPF after pulmonary resection and to identify patients at high risk for BPF. METHODS: From January 2005 to December 2012, 34,000 patients underwent major pulmonary resection (lobectomy, bilobectomy, or pneumonectomy) and were entered into the French National database Epithor. The primary outcome was the occurrence of postoperative BPF at 30 days. The logistic regression model was built using a backward stepwise variable selection. RESULTS: Bronchopleural fistula occurred in 318 patients (0.94%); its prevalence was 0.5% for lobectomy (n = 139), 2.2% for bilobectomy (n = 39), and 3% for pneumonectomy (n = 140). The mortality rate was 25.9% for lobectomy (n = 36), 16.7% for bilobectomy (n = 6), and 20% for pneumonectomy (n = 28). In the final model, nine variables were selected: sex, body mass index, dyspnea score, number of comorbidities per patient, bilobectomy, pneumonectomy, emergency surgery, sleeve resection, and the side of the resection. In the development data set, the C-index was 0.8 (95% confidence interval: 0.78 to 0.82). This model was well calibrated because the Hosmer-Lemeshow test was not significant (χ(2) = 10.5, p = 0.23). We then calculated the logistic regression coefficient to build the predictive score for BPF. CONCLUSIONS: This strong model could be easily used by surgeons to identify patient at high risk for BPF. This score needs to be confirmed prospectively in an independent cohort.
BACKGROUND:Bronchopleural fistula (BPF) remains a rare but fatal complication of thoracic surgery. The aim of this study was to develop and validate a predictive model of BPF after pulmonary resection and to identify patients at high risk for BPF. METHODS: From January 2005 to December 2012, 34,000 patients underwent major pulmonary resection (lobectomy, bilobectomy, or pneumonectomy) and were entered into the French National database Epithor. The primary outcome was the occurrence of postoperative BPF at 30 days. The logistic regression model was built using a backward stepwise variable selection. RESULTS:Bronchopleural fistula occurred in 318 patients (0.94%); its prevalence was 0.5% for lobectomy (n = 139), 2.2% for bilobectomy (n = 39), and 3% for pneumonectomy (n = 140). The mortality rate was 25.9% for lobectomy (n = 36), 16.7% for bilobectomy (n = 6), and 20% for pneumonectomy (n = 28). In the final model, nine variables were selected: sex, body mass index, dyspnea score, number of comorbidities per patient, bilobectomy, pneumonectomy, emergency surgery, sleeve resection, and the side of the resection. In the development data set, the C-index was 0.8 (95% confidence interval: 0.78 to 0.82). This model was well calibrated because the Hosmer-Lemeshow test was not significant (χ(2) = 10.5, p = 0.23). We then calculated the logistic regression coefficient to build the predictive score for BPF. CONCLUSIONS: This strong model could be easily used by surgeons to identify patient at high risk for BPF. This score needs to be confirmed prospectively in an independent cohort.
Authors: Andrew B Servais; Cristian D Valenzuela; Alexandra B Ysasi; Willi L Wagner; Arne Kienzle; Stephen H Loring; Akira Tsuda; Maximilian Ackermann; Steven J Mentzer Journal: Physiol Rep Date: 2018-05