Abstract MP076: Role of the Human Metabolome in Incident Heart Failure among African Americans (AAs) from the Atherosclerosis Risk in Communities (ARIC) Study

Abstract

Introduction: Study of the human metabolome can improve our understanding of the etiology and biology of the failing heart. However, longitudinal data assessing the relations between the human metabolome and incident heart failure (HF) are scarce. We explored the metabolomic antecedents of HF in a sample of AAs in the ARIC study. Methods: We examined 1744 (1111 females, aged 45-64 years) ARIC AAs without prevalent HF, having metabolomic and complete covariate data. We focused our analyses on 204 serum metabolites measured by gas chromatography-mass spectrometry/ liquid chromatography-mass spectrometry and showed high 4-6 week intra-individual repeatability (reliability coefficients ≥ 0.60). We used Cox proportional hazards regression to assess associations between each metabolite and incident hospitalized HF in three hierarchical models: Model 1 (fixed factors): age, sex, and prevalent coronary heart disease (CHD); model 2 (model 1 + baseline lifestyle factors): physical activity, education and smoking status; model 3 (model 2 + baseline physiology factors): BMI, systolic blood pressure, anti-hypertensive medication use, fasting glucose, prevalent diabetes, total cholesterol : HDL ratio, lipid-lowering medication use, and prevalent left ventricular hypertrophy. We used a modified Bonferroni correction accounting for correlations among metabolites to determine statistical significance (p <4×10 -4 ). Results: Over 20 years, 16% of the participants developed HF (n =276). In model 1, 2 and 3, there were 56, 43 and 16 metabolites significantly related to HF, respectively. The 16 metabolites identified after inclusion of all covariates (model 3) included 6 with names and 10 with unknown structural identity. Metabolite X11308 had the largest effect size and was the only one inversely associated with HF (HR=0.75 per SD difference in standardized metabolite value, i.e., (measured or imputed value - mean value)/SD; p=8.34×10 -5 ). Of the six named metabolites, four (N-acetylalanine, p-cresol sulfate, phenylacetylglutamine, pyroglutamine) are involved in amino acid metabolism; one (pro-hydroxy-pro) is in the peptide pathway; and one (erythritol) is a xenobiotic sugar alcohol. These metabolites, except pyroglutamine, were significantly associated with HF in all three models. Pyroglutamine was associated with HF only in model 3 (i.e., after inclusion of the physiology factors) with an 11% change in its HR (from 1.17 in model 2 to 1.29 in model 3), a change attributable almost solely to adjustment to fasting glucose and prevalent diabetes. Conclusions: In our study of ARIC AAs, 16 of 204 metabolites were statistically significant predictors of incident hospitalized HF, independent of measured demographic, lifestyle, and physiology HF risk factors. Metabolomic analyses such as these may identify novel biomarkers or pathways in the etiology of HF and point to new prevention strategies.