Multiple Plasma Biomarkers for Risk Stratification in HFpEF

Study Questions:

How does a multimarker panel perform in predicting risk of death and hospitalization for heart failure of patients with heart failure and preserved ejection fraction (HFpEF)?


The authors measured 48 plasma biomarkers in two cohorts of patients with HFpEF using a Luminex bead-based multiplexed assay. The biomarkers each reflect different pathways including angiogenesis, atherothrombosis, myocardial injury, extracellular matrix turnover, tissue remodeling, inflammation, adipocyte biology, mineral metabolism, and neurohormonal regulation. The biomarkers were measured in 379 participants of TOPCAT (Treatment of Preserved Cardiac Function Heart Failure With an Aldosterone Antagonist Trial), in addition to 156 patients enrolled in the PHFS (Penn Heart Failure Study) as a validation cohort. The authors examined the association between the biomarker levels and the primary outcome of all-cause death or hospitalization for HF using a machine-learning tree-based pipeline optimizer platform.


The TOPCAT cohort consisted of 379 patients with a median age of 70 years, of whom 54% were male, 92% were white, and 58% were enrolled in the United States and Canada. During a median follow-up of 2.86 years, 94 patients experienced the primary outcome. In adjusted analyses using the Bonferroni correction, only FGF-23, FABP-4, and IL-6 were independently predictive of death or hospitalization for HF. The machine learning incorporating all biomarkers was strongly associated with the primary outcome (standardized hazard ratio, 2.61; 95% confidence interval [CI], 1.84-3.71), independently of a clinical risk score, and improved the C-statistic from 0.62 (95% CI, 0.56-0.68) to 0.73 (95% CI, 0.67-0.79). The association did not differ between spironolactone and placebo arms. The findings were replicated in the PHFS cohort.


A multimarker model reflecting various pathophysiologic processes was significantly associated with outcomes and improved risk discrimination in patients with HFpEF.


This study examines the association of a preselected set of biomarkers with outcomes in patients with HFpEF and uses a machine learning algorithm to derive the optimal combination for risk prediction. The results are unsurprising, in that biomarkers of inflammation and kidney dysfunction generally are associated with worse outcomes, and that machine learning can optimize the risk models for the cohort. There was no significant interaction between randomization arms in TOPCAT, suggesting that these markers are not useful to identify patients who would benefit from spironolactone; although this remains to be shown prospectively. The most interesting finding is the relationship between biomarkers and their clustering pattern; consistent with the hypothesis that chronic inflammation is the major pathophysiologic process underlying HFpEF.

Clinical Topics: Geriatric Cardiology, Heart Failure and Cardiomyopathies, Prevention, Acute Heart Failure, Heart Failure and Cardiac Biomarkers

Keywords: Adipocytes, Aldosterone, Biomarkers, Pharmacological, Geriatrics, Heart Failure, Inflammation, Interleukin-6, Mineralocorticoid Receptor Antagonists, Risk Assessment, Secondary Prevention, Spironolactone, Stroke Volume

< Back to Listings