Phenotyping of Mitral Valve Prolapse Without Severe Mitral Regurgitation Using Electrocardiographic and...

Arrhythmic risk stratification in mitral valve prolapse (MVP) without significant mitral regurgitation (MR) remains elusive. Unsupervised machine learning may reveal phenotypic variation among arrhythmic MVP without severe MR.

In this study, the authors hypothesized that hierarchical clustering of echocardiographic and 12-lead electrocardiographic (ECG) parameters alone could identify MVP phenotypes without severe MR associated with sustained ventricular arrhythmia and excess mortality.

The authors identified 343 consecutive MVPs (58 ± 16 years; 51% female) with ≤moderate MR and comprehensive echocardiographic, 12-lead and ambulatory ECG data. They used hierarchical clustering analysis to identify distinctive MVP phenotypes and investigated their association with: 1) arrhythmic events (sudden cardiac arrest, ventricular fibrillation/tachycardia, or frequent ventricular ectopy); and 2) overall mortality (mean follow-up: 5.4 ± 2.7 years).

Three clusters were identified: Cluster 1 (83% of MVP cases), Cluster 2 (9%), and Cluster 3 (8%). Despite mostly trace/mild MR, Cluster 3 exhibited more abnormal parameters of left atrial (LA) and left ventricular structure and function compared with Clusters 1 and 2 (all P < 0.001). Top clustering features included ECG intervals, LA systolic strain, and LA function index. Arrhythmic presentations (n = 77) were identified in 19%, 38%, and 43% of Clusters 1, 2, and 3 (P < 0.001), respectively. Compared with Cluster 1, Clusters 2 (HR: 5.01; P < 0.001) and 3 (HR: 5.85; P < 0.001) had significantly increased mortality risk.

Hierarchical clustering based on standard ECG and echocardiographic data alone identifies 3 MVP clusters with distinct arrhythmic profiles and excess mortality, highlighting LA function as a novel risk parameter in MVP without significant MR.