Identification and validation of ventricular rhythmic risk subgroups in hypertrophic cardiomyopathy by clustering analysis including left ventricular strain analysis

Abstract Background The excess mortality in hypertrophic cardiomyopathy (HCM) patients is mainly attributed to the occurrence of ventricular arrhythmia (VA).The prediction of VA remains challenging and could be improved. Purpose The study aimed to characterize the VA risk profile in HCM patients through clustering analysis combining conventional parameters with information derived from left ventricular longitudinal strain analysis (LV-LS). Methods 434 HCM patients (65% men, mean age 56 years) were retrospectively included from two referral centers and followed longitudinally (mean duration 6 years). The validation cohort included 177 HCM patients from an other tertiary French center. Mechanical and temporal parameters were automatically extracted from the LV-LS segmental curves of each patient in addition to conventional clinical and imaging data. A total of 287 features were analyzed using a clustering approach (k-means). The first endpoint for VA included suspected SCD, aborted cardiac arrest, appropriate ICD therapy (AIT), sustained ventricular tachycardia (SVT) and non-sustained ventricular tachycardia (NSVT) during follow-up. Results From the derivation cohort, 4 clusters were identified with a higher rhythmic risk for clusters 1 and 4 (VA rates of 26%(28/108), 13%(13/97), 12%(14/120), and 31%(34/109) for cluster 1,2,3 and 4 respectively) (Figure 1). These 4 clusters differed mainly by the LV mechanics. Patients from cluster 4 had a severe and homogeneous decrease of myocardial deformation (global longitudinal strain (GLS) -11%, normal value ≥ -20%) associated with LV and left atrium (LA) remodeling (left atrial volume index (LAVI) 46.6 vs. 41.5 ml/m², p=0.04 and LVEF 59.7 vs. 66.3%, p < 0.001) and impaired exercise capacity (% predicted peak work 58.6 vs. 69.5%; p= 0.025). Patients from cluster 1 showed a marked deformation delay and temporal dispersion on the segmental analysis associated with a moderate decrease of the GLS (-15.3%; p <0.0001 for GLS comparison between clusters). Patients from clusters 2 and 3 had a small GLS decrease (GLS -17.2 and -16.3%, respectively) with a less severe phenotype for cluster 2 and older patients for cluster 3 (mean age 62.1 vs. 54.1 years; p < 0.0001). The clustering analysis in the validation population resulted in the creation of 4 clusters overlapping the derivation cohort with similar characteristics (Figure 2). Clusters 1’ and 4’ had a higher incidence of ventricular arrhythmia with a VA rate of 22%(5/23), 18%(16/88), 0%(0/17), and 20%(10/49) for clusters 1’,2’,3’ and 4’ respectively. Conclusion Machine-learning-based cluster analysis processing LV-LS parameters in HCM patients identified 4 clusters with specific LV-strain patterns and different rhythmic risk levels. In the future, the automatic extraction and analysis of LV strain parameters could help to improve the risk stratification for SCD in HCM patients.Clustering analysisExternal validation