Combining Circulating MicroRNA and NT-proBNP to Detect and Categorize Heart Failure Subtypes

Abstract

Background: Clinicians need improved tools to better identify nonacute heart failure with preserved ejection fraction (HFpEF).

Objectives: The purpose of this study was to derive and validate circulating microRNA signatures for nonacute heart failure (HF).

Methods: Discovery and validation cohorts (N = 1,710), comprised 903 HF and 807 non-HF patients from Singapore and New Zealand (NZ). MicroRNA biomarker panel discovery in a Singapore cohort (n = 546) was independently validated in a second Singapore cohort (Validation 1; n = 448) and a NZ cohort (Validation 2; n = 716).

Results: In discovery, an 8-microRNA panel identified HF with an area under the curve (AUC) 0.96, specificity 0.88, and accuracy 0.89. Corresponding metrics were 0.88, 0.66, and 0.77 in Validation 1, and 0.87, 0.58, and 0.74 in Validation 2. Combining microRNA panels with N-terminal pro-B-type natriuretic peptide (NT-proBNP) clearly improved specificity and accuracy from AUC 0.96, specificity 0.91, and accuracy 0.90 for NT-proBNP alone to corresponding metrics of 0.99, 0.99, and 0.93 in the discovery and 0.97, 0.96, and 0.93 in Validation 1. The 8-microRNA discovery panel distinguished HFpEF from HF with reduced ejection fraction with AUC 0.81, specificity 0.66, and accuracy 0.72. Corresponding metrics were 0.65, 0.41, and 0.56 in Validation 1 and 0.65, 0.41, and 0.62 in Validation 2. For phenotype categorization, combined markers achieved AUC 0.87, specificity 0.75, and accuracy 0.77 in the discovery with corresponding metrics of 0.74, 0.59, and 0.67 in Validation 1 and 0.72, 0.52, and 0.68 in Validation 2, as compared with NT-proBNP alone of AUC 0.71, specificity 0.46, and accuracy 0.62 in the discovery; with corresponding metrics of 0.72, 0.44, and 0.57 in Validation 1 and 0.69, 0.48, and 0.66 in Validation 2. Accordingly, false negative (FN) (81% Singapore and all NZ FN cases were HFpEF) as classified by a guideline-endorsed NT-proBNP ruleout threshold, were correctly reclassified by the 8-microRNA panel in the majority (72% and 88% of FN in Singapore and NZ, respectively) of cases.

Conclusions: Multi-microRNA panels in combination with NT-proBNP are highly discriminatory and improved specificity and accuracy in identifying nonacute HF. These findings suggest potential utility in the identification of nonacute HF, where clinical assessment, imaging, and NT-proBNP may not be definitive, especially in HFpEF.

Keywords: biomarker; diagnosis; heart failure; microRNA.