Prognostic Impact of HFpEF in Hypertrophic Cardiomyopathy: A Machine Learning-Based Risk Stratification Study
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Background: Hypertrophic cardiomyopathy (HCM) is a genetic cardiac disorder that often progresses to heart failure with preserved ejection fraction (HFpEF), substantially impacting patient prognosis. However, the clinical characteristics and prognostic significance of HFpEF in HCM are not well understood. This study aims to identify key predictors of adverse outcomes in HCM patients with HFpEF and to develop a machine learning-based risk stratification model for improved patient management. Methods: This retrospective cohort study analyzed data from 2,615 HCM patients who were evaluated at four tertiary medical centers between October 1, 2009, and December 31, 2024. Of these, 1,152 patients were diagnosed with HFpEF based on the American Heart Association (AHA) criteria. Clinical characteristics, echocardiographic parameters, and laboratory biomarkers were assessed, with all-cause mortality and cardiac transplantation as the primary endpoints. Cox proportional hazards regression models were employed to identify independent risk factors. Additionally, machine learning algorithms, including XGBoost, were used to develop a predictive model. Results: HFpEF patients had significantly greater left ventricular septal thickness (19.01 ± 5.86 mm vs. 16.62 ± 5.65 mm, p < 0.001), higher preoperative LVOT gradients (52.97 ± 41.35 mmHg vs. 40.72 ± 38.08 mmHg, p < 0.001), and elevated BNP levels (2,281.63 ± 1,761.16 pg/mL vs. 365.46 ± 193.07 pg/mL, p < 0.001). Both univariate and multivariate Cox regression analysis identified BNP (HR: 1.028, 95% CI: 1.018–1.038, p < 0.001) and atrial fibrillation (HR: 4.028, 95% CI: 1.929–8.411, p < 0.001) as independent predictors of adverse cardiovascular events. The XGBoost-based predictive model outperformed traditional regression methods, achieving an area under the receiver operating characteristic curve (AUC) of 0.90, with an accuracy of 85.4%, sensitivity of 83.9%, and specificity of 86.8%. Conclusions: HFpEF significantly worsens the prognosis in HCM patients, emphasizing the importance of early identification and risk stratification. BNP levels and atrial fibrillation were identified as independent predictors of mortality. The XGBoost-based model demonstrated superior predictive performance compared to conventional methods, presenting a promising AI-driven approach for personalized risk assessment. Future studies should focus on prospective validation, novel HFpEF-targeted therapeutics, and the integration of AI in clinical decision-making.