Abstract P1098: Serum Metabolites and Heart Failure Risk: the Multi-Ethnic Study of Atherosclerosis and the Rotterdam Study

OBJECTIVE To examine the association between serum metabolites and incident heart failure (HF) and to assess the added value of identified metabolites for 10-year HF risk prediction. Methods: A total of 23,571 serum metabolomic spectral variables were measured using untargeted proton nuclear magnetic resonance ( 1 H NMR) spectroscopy-based method. Participants from Multi-Ethnic Study of Atherosclerosis (MESA) and Rotterdam Study (RS) cohorts with metabolomic profiling data and without prevalent HF at baseline were included. Multivariable Cox proportional hazards models were used to estimate the associations between metabolomic features and incident HF, and we further assessed how adjustment for cardiovascular risk factors attenuated the associations. P values were adjusted for multiple comparisons. Correlations among metabolites were tested. Utility of the metabolites in improving discriminative performance HF risk prediction beyond PREVENT-HF score was assessed using Harrell's c- statistic. Results: A total of 3942 MESA and 1515 RS participants were included, with a median of 15 years follow-up. In MESA, 15 metabolites were significantly associated with incident HF after adjusting for age, sex, race, estimated glomerular filtration rate and lifestyle behaviors, with 5 replicated in RS cohort. However, after further adjusting for diabetes or hypertension, these observed associations lost significance. Several metabolites showed moderate to high correlations with fasting glucose levels. Adding 5 replicated metabolites into the PREVENT-HF score reduced c-statistics in both cohorts, and adding 15 metabolites identified from MESA showed no statistically significant improvements. CONCLUSIONS In 2 cohorts of middle-aged to older adults, 15 1 H NMR measured metabolites were associated with incident HF, but the observed associations were not independent of diabetes or hypertension. Incorporating these metabolites into the PREVENT-HF score did not improve HF risk prediction.