Sudden cardiac death risk prediction in arrhythmogenic right ventricular cardiomyopathy: a practical approach to navigating the challenges of prediction models.

Abstract

This commentary refers to ‘Arrhythmic risk prediction in arrhythmogenic right ventricular cardiomyopathy: external validation of the arrhythmogenic right ventricular cardiomyopathy risk calculator’, by P. Jordà et al., https://doi.org/10.1093/eurheartj/ehac289 and the discussion piece ‘Sudden cardiac death risk prediction in arrhythmogenic right ventricular cardiomyopathy: the challenge of complex statistical modelling and its impact in clinical practice’, by E. Monda et al., https://doi.org/10.1093/eurheartj/ehac538. We appreciate the thoughtful critique by Monda et al.1 of our recently published external validation study2 of the arrhythmogenic right ventricular cardiomyopathy (ARVC) risk calculator for the prediction of incident sustained ventricular arrhythmias (VA).3 The authors mention three aspects of the study that they perceive as potentially problematic: the predicted outcome, the included population, and the absence of a recommended threshold for implantable cardioverter defibrillator (ICD) use. First, we agree with Monda et al. that the predicted outcome, any sustained VA, is an imperfect one. Since this study’s aim was to validate the published ARVC risk calculator, we had to use the same outcome. The chosen outcome follows the most accepted definition of ventricular tachycardia and usually warrants the consideration of an ICD in structural heart disease as recommended in European and American guidelines. A universally accepted definition for severe events has not been established. Nonetheless, we published another model specifically predicting events >250 bpm, ventricular fibrillation (VF) and (aborted) sudden cardiac death (SCD).4 While this offers some additional information, it comes with significant shortcomings such as the inclusion of patients with a prior history of sustained VA in the development cohort and that accelerating events that could be fatal if left untreated can be missed by the model. Validation is also yet to come.