Different association of HDLC, apoA-I, and HDL-C/apoA-I with multiple outcomes in HFrEF patients

Int J Cardiol. 2024 Dec 23:132929. doi: 10.1016/j.ijcard.2024.132929. Online ahead of print.

Abstract

Background: High-density lipoprotein-cholesterol (HDLC) has been considered a cardioprotective factor for several decades. However, its association with outcomes in patients with heart failure with reduced ejection fraction (HFrEF) remains controversial. We aimed to investigate the association of HDLC, apolipoprotein A-I (apoA-I), and the HDL-C/apoA-I ratio with multiple outcomes of HFrEF patients and establish prognostic models using machine learning methods.

Methods and results: This was a retrospective, single-center study. The associations between lipid levels and multiple outcomes were examined using logistic regression analysis. Prognostic models for multiple outcomes were further established using four machine learning methods. A total of 352 HFrEF patients were visited successfully. In the multivariable-adjusted logistic regression analysis, HDL-C did not show a significant association with any of the studied outcomes; apoA-I was marginally unassociated with all-cause rehospitalization (adjusted odds ratio [aOR] = 0.62, p = 0.063) but was significantly negatively associated with all-cause death (aOR = 0.53, p = 0.038), rehospitalization for cardiovascular or cerebrovascular disease (aOR = 0.43, p < 0.001), and rehospitalization for heart failure (aOR = 0.55, p = 0.024); apoA-I was also significantly positively associated with left ventricular ejection fraction (LVEF) improvement (aOR = 2.00, p = 0.039). Although several p-values were not statistically significant, both the first and third HDL-C/apoA-I groups showed an increased incidence rate for all adverse outcomes compared with the middle group and a decreased incidence rate for LVEF improvement. In the machine learning analysis, the support vector machine (SVM) and extreme gradient boosting (XGBoost) models demonstrated better predictive performance. For each outcome prognosis, apoA-I and logarithmic N-terminal pro-B-type natriuretic peptide (lg NT-proBNP) were automatically selected.

Conclusion: Among HFrEF patients, apoA-I may be a better marker for predicting outcomes than HDLC. Both low and high levels of HDL-C/apoA-I may indicate a poor prognosis of HFrEF patients.

Keywords: Apolipoprotein A-I; Heart failure; High-density lipoprotein cholesterol; Machine learning; Outcome.