Nuclear magnetic resonance-based metabolomics identifies phenylalanine as a novel predictor of incident heart failure hospitalisation: results from PROSPER and FINRISK 1997.

AIMS: We investigated the association between quantified metabolite, lipid and lipoprotein measures and incident heart failure hospitalisation (HFH) in the elderly, and examined whether circulating metabolic measures improve HFH prediction.

METHODS AND RESULTS: Overall, 80 metabolic measures from the PROspective Study of Pravastatin in the Elderly at Risk (PROSPER) trial were measured by proton nuclear magnetic resonance spectroscopy (n = 5341; 182 HFH events during 2.7-year follow-up). We repeated the work in FINRISK 1997 (n = 7330; 133 HFH events during 5-year follow-up). In PROSPER, the circulating concentrations of 13 metabolic measures were found to be significantly different in those who were later hospitalised for heart failure after correction for multiple comparisons. These included creatinine, phenylalanine, glycoprotein acetyls, 3-hydroxybutyrate, and various high-density lipoprotein measures. In Cox models, two metabolites were associated with risk of HFH after adjustment for clinical risk factors and N-terminal pro-B-type natriuretic peptide (NT-proBNP): phenylalanine [hazard ratio (HR) 1.29, 95% confidence interval (CI) 1.10-1.53; P = 0.002] and acetate (HR 0.81, 95% CI 0.68-0.98; P = 0.026). Both were retained in the final model after backward elimination. Compared to a model with established risk factors and NT-proBNP, this model did not improve the C-index but did improve the overall continuous net reclassification index (NRI 0.21; 95% CI 0.06-0.35; P = 0.007) due to improvement in classification of non-cases (NRI 0.14; 95% CI 0.12-0.17; P < 0.001). Phenylalanine was replicated as a predictor of HFH in FINRISK 1997 (HR 1.23, 95% CI 1.03-1.48; P = 0.023).

CONCLUSION: Our findings identify phenylalanine as a novel predictor of incident HFH, although prediction gains are low. Further mechanistic studies appear warranted.