Metabolomic profiling in heart failure as a new tool for diagnosis and phenotyping.

Classifying heart failure (HF) by stages and ejection fraction (EF) remains a debated topic in cardiology. Metabolomic profiling (MP) offers a means to identify unique pathophysiological changes across different phenotypes, presenting a promising approach for the diagnosis and prognosis of HF, as well as for the development of targeted therapies. In our study, MP was performed on 408 HF patients (54.9% male). The mean ages of patients were 62 [53;68], 67 [65;74], 68 [61;72], and 69 [65;73] years for stages A, B, C, and D, respectively. This study demonstrates high accuracy in HF stage classification, distinguishing Stage A from Stage B with an AUC ROC of 0.91 and Stage B from Stage C with an AUC ROC of 0.97, by integrating chromatography-mass spectrometry data through multiparametric machine learning models. The observed metabolic similarities between HF with mildly reduced EF and HF with reduced EF phenotypes (AUC ROC 0.96) once again highlight the fundamental differences at the cellular and molecular levels between HF with preserved EF and HF with EF < 50%. Hierarchical clustering based on MP identified four distinct HF phenotypes and 26 key metabolites, including metabolites of tryptophan catabolism, glutamine, riboflavin, norepinephrine, serine, and long- and medium-chain acylcarnitines. The average follow-up period was 542.37 [16;1271] days. A downward change in the trajectory of EF [HR 3,008, 95% CI 1,035 to 8,743, p = 0,043] and metabolomic cluster 3 [HR 2,880; 95% CI 1,062 to 7,810, p = 0,0376] were associated with increased risk of all-cause mortality. MP can refine HF phenotyping and deepen the understanding of its underlying mechanisms. Metabolomic analysis illuminates the biochemical landscape of HF, aiding in its classification and suggesting new therapeutic pathways.