Machine learning algorithms for predicting arrhythmic events in Hypertrophic Cardiomyopathy: limited enhancement beyond late gadolinium enhancement.

We aimed to develop and assess the performance of a Machine learning (ML) model integrating common clinical features to predict arrhythmic events in patients with Hypertrophic Cardiomyopathy (HCM). Post-hoc analysis of an international multicenter registry of 531 HCM patients (49 years (IQR 35-61), 57% male) who underwent cardiac magnetic resonance (CMR). The dataset comprised clinical, echocardiographic, and CMR variables, including quantification of late gadolinium enhancement (LGE) using the + 6 SD method. The endpoint was a composite of sudden cardiac death (SCD), aborted SCD, and sustained ventricular tachycardia (VT). A total of 28 events occurred over a median follow-up of 4.1 (IQR 1.8-7.3) years. Several ML models were developed and the predictive performance of the best model was compared to the ESC HCM risk score and to the amount of LGE. The Random Forest (RF) was the most effective method showing a good performance for predicting arrhythmic events [AUC of 0.78 (95% CI: 0.76-0.82, p < 0.001)], substantially outperforming the ESC HCM risk score [AUC of 0.64 (95% CI 0.62-0.67; p < 0.001), p < 0.001 for comparison]. However, when compared to LGE alone [AUC of 0.76 (95% CI: 0.73-0.84, p < 0.001)], the RF model did not provide significant improvement in predicting the endpoint (p = 0.817 for comparison). A ML model using available clinical variables significantly outperformed the ESC HCM risk score in predicting arrhythmic events in HCM. However, its incremental value over LGE alone was weak, underscoring the strong predictive value of this imaging marker. This findings should be interpreted as exploratory and hypothesis-generating.