Development and external validation of a novel model for predicting new clinically important atrial fibrillation after thoracoscopic anatomical lung cancer surgery: a multicenter retrospective cohort study.

ABSTRACT

BACKGROUND: New clinically important postoperative atrial fibrillation (POAF) is the most common arrhythmia after thoracoscopic anatomical lung cancer surgery and is associated with increased morbidity and mortality. The full spectrum of predictors remains unclear, and effective assessment tools are lacking. This study aimed to develop and externally validate a novel model for predicting new clinically important POAF. METHODS: This retrospective study included 14 074 consecutive patients who received thoracoscopic anatomical lung cancer surgery from January 2016 to December 2018 in Shanghai Chest Hospital. Based on the split date of 1 January 2018, we selected 8717 participants for the training cohort and 5357 participants for the testing cohort. For external validation, we pooled 2941 consecutive patients who received this surgical treatment from July 2016 to July 2021 in Shanghai Ruijin Hospital. Independent predictors were used to develop a model and internally validated using a bootstrap-resampling approach. The area under the receiver operating characteristic curves (AUROCs) and Brier score were performed to assess the model discrimination and calibration. The decision curve analysis (DCA) was used to evaluate clinical validity and net benefit. New clinically important POAF was defined as a new-onset of POAF that causes symptoms or requires treatment. RESULTS: Multivariate analysis suggested that age, hypertension, preoperative treatment, clinical tumor stage, intraoperative arrhythmia and transfusion, and operative time were independent predictors of new clinically important POAF. These seven candidate predictors were used to develop a nomogram, which showed a concordance statistic (C-statistic) value of 0.740 and good calibration (Brier score; 0.025). Internal validation revealed similarly good discrimination (C-statistic, 0.736; 95% CI 0.705-0.768) and calibration. The decision curve analysis showed positive net benefits with the threshold risk range of 0-100%. C-statistic value and Brier score were 0.717 and 0.028 in the testing cohort, and 0.768 and 0.012 in the external validation cohort, respectively. CONCLUSIONS: This study identified seven predictors of new clinically important POAF, among which preoperative treatment, intraoperative arrhythmia, and operative time were rarely reported. The established and externally validated model has good performance and clinical usefulness, which may promote the application of prevention and treatment in high-risk patients, and reduce the development and related adverse outcomes of this event.