Metabolomic clusters of new-onset heart failure in general population

Abstract Background Heart failure (HF) is a heterogeneous disorder with sex-specific disease pathophysiology. Identifying metabolomic clusters underlying HF risk may provide insight into HF pathophysiology and prevention for men and women. Purpose We aimed to determine the potential of metabolomic profiling in predicting incident HF among community-dwelling men and women. Methods 1,383 individuals (59% women) without a history of HF from a prospective population-based cohort had available data on 1,087 metabolites. After excluding unknown and highly-correlated (Spearman correlation 0.90) metabolites, analysis was conducted using 765 metabolites. Initially, SHAP values were computed through repeated cross-validation of the XGBoost model to select the, sex-specific, metabolites for survival analyses. The selected metabolites were incorporated into Self-Organizing Maps, along with Kmeans methods, to perform sex-specific clustering, allowing for a nuanced exploration of distinct profiles of HF risk. Extracted distinct clusters were then characterized according to cardiovascular risk factors, echocardiographic, and electrographic variables. Results During a median follow-up of 10 years, 306 incident HF occurred. 56 metabolites in men and 46 metabolites in women were selected. The clustering algorithm resulted in three distinct clusters of low-, moderate-, and high-risk based on incident of HF. In men, the incidence rates per 1000 person-years were 15.9 for low-risk, 31.6 for moderate-risk, and 41.2 for high-risk clusters. For women, the corresponding rates were 13.3 for low-risk, 20.0 for moderate-risk, and 29.4 for high-risk clusters. The low-risk cluster in both men and women was characterized by a favorable cardiovascular profile. Men in the moderate-risk cluster exhibited poor cardiovascular profile, based on lower estimated glomerular filtration rate (eGFR), increased left ventricular diastolic diameter, and extended R-to-R interval. The high-risk cluster in men was characterized by larger body mass index, more prevalence of hypertension, and larger left atrial diameter. In women, the high-risk profile was characterized by larger left ventricular diastolic diameter and lower eGFR. Specific metabolites played a role in distinguishing the risk clusters. Salicylate, 4-hydroxyphenylacetylglutamine, and 2,3-dihydroxyisovalerate in men and salicylate, sulfate of piperine metabolite C18H21NO3 (1), and piperine in women were the top 3 metabolites associated with distinct risk clusters. Conclusion We identified sex-specific cluster of metabolites characterizing varying HF risk in general population. Each cluster exhibited a unique cardiovascular profile, carrying the potential to improve HF risk management. Survival plot