Heart failure & cardiomyopathiesSecondary analysis
Machine Learning Approach to Identify Phenotypes in Patients with Ischaemic Heart Failure with Reduced Ejection Fraction
- Secondary analysis of the EPHESUS and CAPRICORN randomised trials (machine learning clustering)
- pooled trials
- external validation in COMMANDER-HF (n = 5022)
Authors' conclusion
Clustering reveals distinct risk subgroups in the heterogeneous array of ischaemic HFrEF patients.
Study context
- Population
- Ischaemic heart failure with reduced ejection fraction (HFrEF) patients pooled from the EPHESUS and CAPRICORN trials (64 ± 12 years; 28% women)validation n = 5022
- Participants
- 8,591 patients
- Comparison
- Machine learning-derived phenotype clusters
- Cluster 1n = 2,161
- Cluster 2n = 1,376
- Cluster 3n = 1,157
- Cluster 4n = 2,073
- Cluster 5n = 1,824
- Primary endpoint
- HF hospitalization or all-cause death; CV hospitalization or all-cause death; MACE
Primary result
As reported in the abstract
Association of clusters with outcomes
- Five clinically distinct clusters identified; membership efficiently predicted by a random forest algorithm
- Clusters significantly associated with outcomes in derivation and validation datasets; Cluster 1 highest risk, Cluster 4 lowest
Further findings
MRA benefit on cardiovascular (CV) hospitalization or all-cause death, by cluster
- MRA benefit magnified in lowest-risk clusters (Clusters 2 and 4)
Original publication
Machine Learning Approach to Identify Phenotypes in Patients with Ischaemic Heart Failure with Reduced Ejection Fraction. Eur J Heart Fail. Published 10 December 2024. doi:10.1002/ejhf.3547
Summarised from the published abstract; every number is checked against it