Background: Heart failure (HF) is a complex syndrome associated with significant morbidity and mortality globally. Outcomes in HF vary considerably based on etiology, clinical features, and biomarker profiles. This study aims to assess the utility of multi-biomarker models, particularly B-type natriuretic peptide (BNP) and specific proteomic signatures, in predicting mortality across different HF etiologies.
Methods: We will analyze HF patients and matched controls from the UK Biobank, stratified into etiological subgroups (Idiopathic, Hypertensive, Ischemic, Alcoholic, and Other). Baseline demographics, clinical characteristics, plasma protein levels (n=2,923 proteins across 49,327 individuals), and metabolomics data will be evaluated for associations with mortality over a minimum follow-up period of 24 months. Kaplan-Meier survival analysis and Cox proportional hazards (CPH) models will be used to compare mortality risks among groups. Feature selection methods, including XGBoost and stochastic search variable selection (SSVS), will be applied to identify key protein biomarkers.
Expected Results: We anticipate that the combination of BNP with selected plasma proteins and/or metabolomics data will significantly improve mortality predictions in HF, demonstrating the potential of multi-biomarker models to enhance risk stratification and optimize patient management.
