Clustering in dilated cardiomyopathy at initial evaluation: An effective tool for clinical stratification

Area of Science:

• Cardiology
• Medical Informatics
• Genetics

Background:

• Dilated cardiomyopathy (DCM) presents variably, complicating risk stratification.
• Current stratification methods are complex and require extensive evaluations.
• There is a need for simpler, effective methods for early DCM patient assessment.

Purpose of the Study:

• To identify distinct subgroups within dilated cardiomyopathy (DCM) patients using machine learning (ML).
• To develop a simplified approach for DCM patient stratification based on initial cardiological evaluation.
• To correlate identified subgroups with genetic background and clinical outcomes.

Main Methods:

• Retrospective analysis of a large DCM patient cohort with baseline clinical, genetic, and outcome data.
• Unsupervised ML clustering to identify patient subgroups, followed by simplification using key electrocardiogram (ECG) variables.
• Validation of the simplified clustering model in an independent external cohort.

Main Results:

• Two DCM subgroups (CL1 and CL2) were identified, primarily differentiated by ECG characteristics.
• CL2 showed a lower prevalence of pathogenic/likely pathogenic variants compared to CL1 (15% vs. 47%).
• A simplified 3-variable ECG model effectively identified subgroups and predicted a lower risk of sudden cardiac death/major ventricular arrhythmias in CL2.

Conclusions:

• ML analysis of baseline ECG variables can effectively delineate DCM subgroups.
• These subgroups exhibit differences in genetic profiles and disease progression.
• This ML-driven ECG-based approach offers a valuable tool for early risk stratification in DCM patients.