Cluster analysis of extracellular matrix biomarkers predicts the development of impaired systolic function within 1 year of acute myocardial infarction

Abstract

The clinical utility of combining extracellular matrix (ECM) biomarkers to predict the development of impaired systolic function following acute myocardial infarction (AMI) remains largely undetermined. A combination of ELISA and multiplexing assays were performed to measure matrix metalloproteinase (MMP)-2, MMP-3, MMP-8, MMP-9, periostin, N-terminal type I procollagen (PINP) and tissue inhibitor of matrix metalloproteinase-1 (TIMP-1) in plasma samples from 120 AMI patients. All patients had an echocardiogram within 1 year of AMI, and were divided into impaired (n = 37, LVEF < 50%) and preserved (n = 83, LVEF ≥ 50%) systolic function groups. Exploratory factor analysis was performed on log-transformed biomarkers using principle axis analysis with Oblimin rotation. Cluster analysis was performed on log-transformed and normalised biomarkers using Ward's method of minimum variance and the squared Euclidean distance metric. Upon univariate analysis, current smoking, prescription of ACE inhibitors at discharge, peak hsTnT > 610 ng/L (median), MMP-8 levels, Factor 1 scores and Cluster One assignment were predictive of impaired systolic function. Upon multivariate analysis, Cluster One assignment (odds ratio [95% CI], 2.74 [1.04-7.23], p = 0.04) remained an independent predictor of systolic dysfunction in combination with clinical variables. These observations support the usefulness of combining ECM biomarkers using cluster analysis for predicting the development of impaired systolic function in AMI patients.

Keywords: Acute myocardial infarction; Biomarkers; Cluster analysis; Combined biomarker analysis; Extracellular matrix.

Fig. 1

Exploratory factor analysis results in cohort of AMI patients a The rotated factor matrix of log-transformed ECM biomarkers in 120 AMI patients using EFA. The large ovals represent each latent factor with the percentage of variance for each factor described in bold. The small ovals represent the variables included within each latent factor, and the loading factors are displayed in bold below. b Mann–Whitney U testing demonstrated a significant increase in Factor 1 scores in patients with impaired systolic function compared to preserved systolic function. c No differences were observed in Factor 2 scores between systolic function groups. Median and interquartile range are plotted, and graphs were created using GraphPad Prism software, version 7.04 for Windows. Abbreviations: MMP matrix metalloproteinase, TIMP tissue inhibitor of matrix metalloproteinase, PINP N-terminal type I procollagen

Fig. 2

Dendrogram of cluster analysis performed on the AMI population. Hierarchical cluster analysis performed using MMP-3, MMP-8, MMP-9 and TIMP-1 biomarker levels separated patients into two distinct groups. Cluster One (n = 83) is shown in blue, and Cluster Two (n = 37) is shown in orange. This image was created using the factoextra package in R version 4.0.2, www.R-project.org