Epigenetic scores and aging clocks capture mortality risk in heart failure

Abstract

Current prognostic tools for heart failure (HF) rely mainly on clinical parameters and often fail to reflect the biological heterogeneity underlying patient outcomes. DNA methylation (DNAm)-based aging clocks and trait-specific epigenetic scores capture biological age and cumulative lifestyle and environmental exposures, potentially identifying high-risk patients not detected by clinical risk tools. To determine whether DNAm-based aging clocks and trait-specific epigenetic scores improve mortality risk prediction beyond clinical assessment in a multinational HF population. We analyzed patients with established HF from the Global Congestive Heart Failure (G-CHF) study with available baseline whole-blood DNAm data. Eleven trait-specific DNAm scores (e.g., Smoking, Alcohol, CRP, BMI, diet) and five epigenetic aging clocks (GrimAge, HannumAge, PhenoAge, PaceOfAging, HorvathAge) were derived using published epigenome-wide association study weights. Mortality associations were assessed using multivariable Cox regression adjusted for age, sex, ancestry, and the Meta-Analysis Global Group in Chronic Heart Failure (MAGGIC) risk score. Patients were stratified according to number of DNAm score and aging clock extremes (≥1.5 standard deviation [SD]). LASSO regression was used to identify the most predictive DNAm scores and clocks, which were combined into a composite Epigenetic model. Model performance for mortality prediction was compared between the MAGGIC and the Epigenetic model using Harrell’s C-index. A total of 2,594 HF patients (mean±SD age, 63±14; 34% female) were included. During a median follow-up of 3 years, 565 (22%) patients died from any cause. In multivariable Cox models, several DNAm scores (CRP, Frailty, Mortality) and aging clocks (GrimAge, PhenoAge, PaceOfAging) were independently associated with mortality (Figure 1a). Having 3, 4 and ≥5 extreme DNAm scores/aging clocks conferred a 1.57-fold, 2.16-fold and 2.70-fold (P=2.96×10-3, P=4.36×10-5 and P=4.65×10-7) increased risk of death, respectively (Figure 1b). Using a LASSO-derived Epigenetic composite (including CRP, GrimAge, Mortality, PhenoAge, AHEI [Alternative healthy eating index]), patients with high Epigenetic but low MAGGIC had a 3.10-fold (95% CI 2.19-4.39) higher risk of death versus those with low Epigenetic and low MAGGIC (Figure 2a). The Epigenetic model improved mortality prediction beyond the MAGGIC score, particularly in patients from low-income countries (C-index 0.65 vs. 0.58, P=0.0115; Figure 2b). A cumulative epigenetic signature stratifies mortality risk in HF and outperforms clinical risk assessment. DNAm-based profiling may enable more precise identification of high-risk patients beyond conventional scores.Figure1Figure2