Ensembling Electrical and Proteogenomics Biomarkers for Improved Prediction of Cardiac-Related 3-Month Hospitalizations: A Pilot Study

Background Many risk models for predicting mortality, hospitalizations, or both in patients with heart failure have been developed but do not have sufficient discriminatory ability. The purpose of this study was to identify predictive biomarkers of hospitalizations in heart failure patients using omics-based technologies applied to blood and electrical monitoring of the heart. Methods Blood samples were collected from 58 heart failure patients during enrollment into this study. Each patient wore a 48-hour Holter monitor that recorded the electrical activity of their heart. The blood samples were profiled for gene expression using microarrays and protein levels using multiple reaction monitoring. Statistical deconvolution was used to estimate cellular frequencies of common blood cells. Classification models were developed using clinical variables, Holter variables, cell types, gene transcripts, and proteins to predict hospitalization status. Results Of the 58 patients recruited, 13 were hospitalized within 3 months after enrollment. These patients had lower diastolic and systolic blood pressures, higher brain natriuretic peptide levels, most had higher blood creatinine levels, and had been diagnosed with heart failure for a longer time period. The best-performing clinical model had an area under the receiver operating characteristic curve of 0.76. An ensemble biomarker panel consisting of Holter variables, cell types, gene transcripts, and proteins had an area under the receiver operating characteristic curve of 0.88. Conclusions Molecular-based analyses as well as sensory data might provide sensitive biomarkers for the prediction of hospitalizations in heart failure patients. These approaches may be combined with traditional clinical models for the development of improved risk prediction models for heart failure.