Abstract 12519: Biomarker Characterisation of Heart Failure Subtypes: A Population-Based Bioresource Linked Study

Abstract

Background: Deep learning (DL)-based interpretation of echocardiographic images can help create pragmatic cohort studies to validate novel biomarkers. In this study, we have characterised biomarkers implicated in heart failure (HF) pathophysiology across the ejection fraction (EF) range in a DL-supported pragmatic cohort study. Methods: We used electronic health record data from Tayside, representing 20% of the Scottish population, between 1993 and 2021. Utilising record linkage to DL automated reading of echocardiographic DICOM images, we identified patients with HFrEF or HFpEF and patients without HF. We then linked identified patients to available plasma samples from the Go-SHARE BioBank. We analysed multi-panel biomarkers, such as NT-proBNP, high-sensitivity cardiac troponin-T (hs-cTnT), growth differentiation factor-15 (GDF-15), insulin-like growth factor-binding protein-7 (IGFBP-7), bone morphogenetic protein-10 (BMP-10), Dickkopf WNT Signaling Pathway Inhibitor-3 (DKK-3), fibroblast growth factor-23 (FGF-23), angiopoietin-2 (Ang2), and fatty acid-binding protein-3 (FABP-3), all with electrochemiluminescence immunoassays (Roche Diagnostics, Germany). Results: We identified 236 patients with HFpEF, 156 with HFrEF, and 185 patients without HF with available plasma samples from 3680 patient records linked to echocardiographic DICOMs and blood samples. Patients with HFrEF and HFpEF had significantly higher median values of hs-cTnT, GDF-15, NT-proBNP, IGFBP-7, DKK-3, and FGF-23 than controls. The diagnostic accuracy (AUC) of NT-proBNP, hs-cTnT, GDF-15, IGFBP-7, and FABP-3 was >0.8 for identifying patients with HFrEF or HFpEF from controls. DL-algorithm-based interpretation of DICOM images provided further coverage of echocardiographic parameters relevant for diagnosis of HF subtypes. All biomarkers were associated with all-cause mortality and HF-related hospitalisation during a median follow-up of 1089 days. Conclusions: In this population-based study, we demonstrated a novel approach utilising a bioresource linked to a DL-supported pragmatic study to explore biomarkers in HF and demonstrated that multiple biomarkers to have diagnostic and prognostic significance for patients with HFrEF and HFpEF.