HEART failure with preserved ejection fraction (HFpEF) in hypertrophic cardiomyopathy (HCM) was identified as a major independent predictor of adverse outcomes in a large new machine learning–driven study, highlighting an under-recognised high-risk patient subgroup.
HCM, a genetic cardiac disorder characterised by abnormal thickening of the heart muscle, has long been associated with variable clinical trajectories. While some patients remain stable, others experience progressive heart failure, arrhythmias, or sudden cardiac death.
Increasingly, HFpEF in HCM has been recognised as a common but poorly understood phenotype, prompting efforts to refine risk stratification.
HFpEF in HCM Signals Higher Risk
In this multicentre retrospective cohort study of 2,802 patients with HCM, nearly half (47.8%) were found to have HFpEF. After propensity score matching to balance baseline characteristics, HFpEF in HCM remained strongly associated with worse event-free survival (hazard ratio [HR]: 2.612; 95% confidence interval [CI]: 2.188–3.118; p<0.001).
Further stratification using the H₂FPEF score revealed a clear gradient of risk. Patients classified as high-risk demonstrated significantly poorer outcomes (HR: 2.925; 95% CI: 2.210–3.701; p<0.001), reinforcing the heterogeneity within this population.
Machine Learning Enhances Risk Prediction
To improve individualised prognostication, researchers developed four machine learning models.
Among these, a random forest model achieved the highest predictive accuracy (area under the curve: 0.856). Importantly, model interpretability analysis identified HFpEF status and B-type natriuretic peptide (BNP) levels as the most influential predictors of adverse outcomes.
A non-linear relationship between BNP and risk was also observed, with event rates accelerating at higher biomarker concentrations. This finding suggests that traditional linear risk models may underestimate risk in patients with markedly elevated BNP levels.
Implications for Precision Cardiology
These findings position HFpEF in HCM as a prevalent and clinically significant phenotype that warrants closer attention in routine practice. By integrating clinical scoring systems, biomarker analysis, and machine learning, the study offers a more nuanced framework for risk stratification.
As the analysis was retrospective and conducted across tertiary centres, the authors noted that external validation is needed before widespread clinical adoption.
Nonetheless, the results underscore the potential of data-driven approaches to refine prognostic assessment and guide personalised management strategies in HCM.
As precision cardiology continues to evolve, identifying high-risk subgroups such as HFpEF in HCM may be key to improving long-term outcomes.
Reference
Zhang W et al. Machine learning–based risk stratification identifies heart failure with preserved ejection fraction as an independent predictor of adverse outcomes in hypertrophic cardiomyopathy. Sci Rep. 2026;DOI:10.1038/s41598-026-46573-z.
Featured image: Dai Yim on Adobe Stock
