Development of a model to predict recurrence after bronchial artery embolization for non-cancer related hemoptysis

Background Relapse after effective bronchial arterial embolization (BAE) for controlling hemoptysis is not uncommon. Studies reported diverse predictors of recurrence. However, a model to assess the probability of recurrence in non-cancer related hemoptysis patients after BAE has not been reported. This study was to develop a model to predict recurrence after BAE for non-cancer related hemoptysis. Methods The study cohort included 487 patients who underwent BAE for non-cancer-related hemoptysis between January 2015 and December 2019. We derived the model's variables from univariate and multivariate Cox regression analyses. The model presented as a nomogram scaled by the proportional regression coefficient of each predictor. Model performance was assessed with respect to discrimination and calibration. Results One-month and 1-, 2-, 3- and 5-year recurrence-free rates were 94.5%, 88.0%, 81.4%, 76.2% and 73.8%, respectively. Risk factors for recurrence were underlying lung diseases and the presence of systemic arterial-pulmonary circulation shunts. This risk prediction model with two risk factors provided good discrimination (area under curve, 0.69; 95% confidence interval, 0.62-0.76), and lower prediction error (integrated Brier score, 0.143). Conclusion The proposed model based on routinely available clinical and imaging features demonstrates good performance for predicting recurrence of non-cancer-related hemoptysis after BAE. The model may assist clinicians in identifying higher-risk patients to improve the long-term efficacy of BAE.