Abstract MP07: Metabolic Signature Improves Heart Failure Risk Prediction In Older Adults

Abstract

Introduction: Circulating metabolome profiling holds promise in predicting HF risk, but its prediction performance among older adults is not well established. Hypothesis: We hypothesize that metabolic signatures are associated with the risk of HF and its subtypes (HF with preserved ejection fraction (HFpEF) and HF with reduced ejection fraction (HFrEF)), and they can improve HF risk prediction beyond established risk factors. Methods: We measured 828 serum metabolites among 4,030 African and European Americans free of HF from the Atherosclerosis Risk in Communities (ARIC) study visit 5 (2011-2013). We regressed incident HF on each metabolite using Cox proportional hazards models. A metabolite risk score (MRS) was derived by summing individual metabolite levels weighted by beta coefficients estimated from least absolute shrinkage and selection operator (LASSO) regularized regressions. We regressed incident HF, HFpEF and HFrEF on the MRS. Harrell’s C-statistics were calculated to evaluate risk discrimination. We replicated the association between MRS and HF in 3,697 independent ARIC participants with metabolite measured at visit 1 (1987-1989). Results: Among 4,030 participants, the mean (SD) age was 76 (5) years. Adjusting for HF risk factors, 302 metabolites were associated with incident HF (false discovery rate < 0.05). One SD increase of the MRS, constructed from 51 metabolites selected by LASSO, was associated with two to three-fold high risk of HF, HFpEF and HFrEF in the fully adjusted models ( Table ). Five-year risk prediction analysis showed that C statistics improved from 0.850 to 0.884 by adding MRS over ARIC HF risk factors, kidney function and NT-proBNP (ΔC (95%CI) = 0.034 (0.017,0.052)). In the replication analysis, a more parsimonious MRS constructed using 15 metabolites, was associated with incident HF ( Table ). Conclusions: We identified a metabolic signature that was associated with the risk of HF and improved HF risk prediction. Our findings may shed light on pathways in HF development and at-risk populations.