Identifying distinct clinical clusters in heart failure with mildly reduced ejection fraction

Area of Science:

• Cardiology
• Clinical Research
• Data Science

Background:

• Heart failure (HF) is a complex condition with diverse patient profiles.
• Heart failure with mildly reduced ejection fraction (HFmrEF) is a recently defined category (EF 41-49%).
• Patient heterogeneity poses challenges for clinical trials and prognostication.

Purpose of the Study:

• To identify distinct patient clusters within the HFmrEF population.
• To compare the prognostic implications of these identified clusters.
• To assess the utility of cluster analysis for HFmrEF stratification.

Main Methods:

• Latent class analysis was employed to cluster HFmrEF patients from the Swedish HF registry (n=7316).
• Clusters were validated using the CHECK-HF registry (n=1536).
• Cox proportional hazard models and Fine-Gray sub-distribution models were used to compare mortality and hospitalization risks across clusters, adjusting for age and sex.

Main Results:

• Six robust clusters were identified with varying prevalence: low-comorbidity (17%), ischaemic-male (13%), atrial fibrillation (20%), device/wide QRS (9%), metabolic (19%), and cardio-renal phenotype (22%).
• Significant differences in mortality and hospitalization were observed between clusters.
• The device/wide QRS, metabolic, and cardio-renal phenotype clusters exhibited substantially higher hazard ratios for adverse outcomes compared to the low-comorbidity reference cluster.

Conclusions:

• Robust and clinically meaningful clusters were identified in HFmrEF patients.
• These clusters demonstrate differential prognosis regarding mortality and hospitalization.
• The developed clustering model shows promise as a tool for clinical differentiation and prognostic assessment in HFmrEF, aiding clinical trial design.