Proteomic markers enhance mortality prediction in heart failure

Clinical models incompletely capture the molecular pathways driving heart failure (HF) progression. This study evaluated whether molecular risk stratification provides incremental prognostic information beyond established clinical predictors in patients with HF.

A total of 2432 patients from the Global Congestive Heart Failure (G-CHF) registry with available genotyping, DNA methylation, and proteomic profiling were analysed. Three molecular scores were assessed: a composite cardiovascular polygenic risk score (PRS) from DNA sequence polymorphisms, a methylation risk score (MRS) derived from epigenome-wide associations, and a 23-protein-based score (ProteomicDeath23). Each score was tested individually and in combination with the clinical Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC) risk score and N-terminal pro-B-type natriuretic peptide (NT-proBNP) levels for mortality prediction. Validation was performed in an HF subset of the UK Biobank (UKB).

Over a median follow-up of 3.0 years in G-CHF, 523 patients died from any cause (7.64 per 100 person-years [PY]). In multivariable analyses, ProteomicDeath23 was the strongest independent predictor of all-cause mortality (hazard ratio [HR] per 1 standard deviation, 2.23), outperforming NT-proBNP (HR 2.00), MRSMortality (HR 1.66), PRSmetaCVD (HR 1.10), and the MAGGIC score (HR 1.70). A model combining ProteomicDeath23 with MAGGIC and NT-proBNP achieved the highest discrimination for mortality (C-index, 0.77). Addition of MRSMortality to this proteomic-clinical model resulted in only small improvements in discrimination (ΔC-index, +0.004, P = .0039), while the PRSmetaCVD provided no incremental benefit. Among patients with low NT-proBNP/MAGGIC score, mortality rates increased from 1.71 to 8.12 per 100 PY across ProteomicDeath23 tertiles. Consistent results were observed in the UKB-HF validation cohort.

A proteomic score was the strongest molecular predictor of mortality in HF. Integrating proteomic signatures with clinical risk factors significantly improved risk prediction.