Dilated cardiomyopathy (DCM) shows wide variation in presentation and disease course, and current stratification strategies require complex multimodality evaluation. A retrospective cohort study, published in the European Heart Journal, applied machine learning (ML) to baseline data from the first cardiology evaluation to identify clinically distinct subgroups.

The study enrolled 409 patients with DCM (mean age 46 years; 71% male). ML clustering defined two subgroups: CL1 (82%) and CL2 (18%). Electrocardiogram (ECG) features mainly separated the groups. CL2 carried fewer pathogenic or likely pathogenic variants compared with CL1 (15% vs. 47%, p<0.001). A simplified model used three ECG variables: QRS duration, presence of left bundle branch block, and intrinsicoid deflection greater than 50 ms. This model reproduced the subgroup classification. An external cohort of 160 patients (mean age 54 years; 68% male) validated the results. CL2 showed a lower risk of sudden cardiac death or major ventricular arrhythmias in both the derivation cohort (HR 0.29; 95% CI 0.13–0.67) and the validation cohort (p=0.017).

The study demonstrated that ECG-based clustering can define two DCM subgroups with different arrhythmic risk and genetic background, offering a practical approach to early risk stratification.