Comparing the performance of risk stratification scores in Brugada syndrome: a multi-centre study

Abstract

AbstractIntroductionThe management of Brugada Syndrome (BrS) patients at intermediate risk of arrhythmic events remains controversial. The present study evaluated the predictive performance of different risk scores in an Asian BrS population and its intermediate risk subgroup.MethodsThis is a retrospective territory-wide cohort study of consecutive patients diagnosed with BrS from January 1st, 1997 to June 20th, 2020 in Hong Kong. The primary outcome is sustained ventricular tachyarrhythmias. A novel predictive score was developed. Machine learning-based nearest neighbor and Gaussian Naïve Bayes models were also developed. The area under the receiver operator characteristic (ROC) curve (AUC) was compared between the different scores.ResultsThe cohort consists of 548 consecutive BrS patients (7% female, age at diagnosis: 50±16 years old, follow-up duration: 84±55 months). For risk stratification in the whole BrS cohort, the score developed by Sieira et al. showed the best performance with an AUC of 0.805, followed by the Shanghai score (0.698), and the scores by Okamura et al. (0.667), Delise et al. (0.661), Letsas et al. (0.656) and Honarbakhsh et al. (0.592). A novel risk score was developed based on variables and weighting from the best performing score (the Sieira score), with the inclusion of additional variables significant on univariable Cox regression (arrhythmias other than ventricular tachyarrhythmias, early repolarization pattern in the peripheral leads, aVR sign, S-wave in lead I and QTc ≥436 ms). This score has the highest AUC of 0.855 (95% CI: 0.808-0.901). The Gaussian Naïve Bayes model demonstrated the best performance (AUC: 0.97) compared to logistic regression and nearest neighbor models.ConclusionThe inclusion of investigation results and more complex models are needed to improve the predictive performance of risk scores in the intermediate risk BrS population.