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

Abstract

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.

Keywords: Advanced lipoprotein profiling; FINRISK; Heart failure; Metabolomics; PROSPER; Phenylalanine.

Figure 1

Flow diagram of the PROSPER study. Of the 5432 samples with 6‐month demographic records, including N‐terminal pro‐B‐type natriuretic peptide (NT‐proBNP), 5341 had 6‐month sample analysed by nuclear magnetic resonance (NMR) spectroscopy. Of these samples, 182 were from individuals who were later hospitalised for heart failure (HF). 160 samples were excluded due to missing metabolic measures. Of these 5181, 178 were later hospitalised for HF. HFH, heart failure hospitalisation.

Figure 2

Forest plot of hazard ratios and 95% confidence intervals (CI) of association for metabolites with incident heart failure hospitalisation (HFH) in PROSPER during 2.7 years of follow‐up. Associations were adjusted for treatment group, age, sex, smoking status, country, body mass index (BMI), myocardial infarction, systolic (SBP) and diastolic blood pressure (DBP), coronary artery bypass graft, percutaneous transluminal coronary angioplasty, transient ischaemic attack, stroke, angina, claudication, peripheral vascular disease, diabetes, estimated glomerular filtration rate (eGFR), N‐terminal pro‐B‐type natriuretic peptide (NT‐proBNP) concentration (6‐month) and treatment with angiotensin‐converting enzyme inhibitors, beta‐blockers, calcium channel blockers, anti‐arrhythmics and diuretics (note all medications are as recorded at baseline, 0 month). Nuclear magnetic resonance measures with P < 0.05 in Model A are shown. HDL, high‐density lipoprotein; LDL, low‐density lipoprotein. Model A: adjusted for sex, BMI, SBP, DBP, current smoking, diabetes, pravastatin/placebo, blood pressure‐lowering therapy, major coronary events/baseline cardiovascular disease. Model B: adjusted as for Model A plus eGFR. Model C: adjusted as for Model B plus NT‐proBNP.

Figure 3

Comparison of hazard ratios for incident heart failure hospitalisation or heart failure events in PROSPER and the FINRISK 1997 cohort. Hazard ratios are adjusted for age, sex, body mass index, systolic and diastolic blood pressure, current smoking, diabetes, pravastatin/placebo (in PROSPER); blood pressure‐lowering therapy, major coronary events/baseline cardiovascular disease; estimated glomerular filtration rate and N‐terminal pro‐B‐type natriuretic peptide. Only metabolic measures with P < 0.01 in at least one cohort are displayed (with the exception of acetate). CI, confidence interval; FA, fatty acid; HDL, high‐density lipoprotein; LA, linoleic acid; MUFA, monounsaturated fatty acid; PUFA, polyunsaturated fatty acid.