Differential metabolomic prediction of outcomes in HFpEF and HFrEF: a prospective international multi-ethnic cohort study

Abstract

Aims: Metabolic dysregulation accompanies heart failure (HF), yet coordinated biochemical shifts are rarely quantified for prognostication. We investigated whether early metabolic profiles predict morbidity and mortality in HF with preserved (HFpEF) or reduced ejection fraction (HFrEF). Methods and results: In the PEOPLE and SHOP cohorts (n = 1520), 288 plasma metabolites were quantified by tandem mass spectrometry. Weighted co-expression network analysis produced metabolic indices that were regressed against all-cause mortality and composite outcomes (death or readmission) using multivariable proportional hazards and win-ratio models. In HFpEF, metabolic indices of nitric oxide signalling and glycaemic control (netNO-pEF) and purine metabolism (netPurine-pEF) predicted outcomes. In HFrEF, indices of acylcarnitine (netACa-rEF) and amino acids (netAA-rEF) predicted outcomes. Every 20% higher netNO-pEF was associated with adjusted hazard ratios (HR20%higher) of 0.64 [95% CI 0.49–0.84] in PEOPLE-HFpEF and 0.40 [0.20–0.83] in SHOP-HFpEF; HR20%higher for netPurine-pEF were 1.71 [1.30–2.24] in PEOPLE-HFpEF and 1.96 [1.08–3.55] in SHOP-HFpEF; HR20%higher for netACa-rEF were 1.48 [1.23–1.79] in PEOPLE-HFrEF and 1.39 [1.15–1.68] in SHOP-HFrEF; HR20%higher for netAA-rEF were 0.75 [0.62–0.91] in PEOPLE-HFrEF and 0.51 [0.39–0.67] in SHOP-HFrEF. Adjusted inverse win-ratios were concordant: 1/WR20%higher for netACa-rEF were 1.40 [1.19–1.64] in PEOPLE-HFrEF and 1.17 [1.02–1.34] in SHOP-HFrEF; 1/WR20%higher for netNO-pEF were 0.74 [0.54–1.00] in PEOPLE-HFrEF and 0.55 [0.33–0.89] in SHOP-HFrEF. netNO-pEF was negatively associated with body surface area among HFpEF with concomitant hypertension and diabetes, but only in SHOP-HFpEF. Adding orotic acid, a constituent metabolite of netNO-pEF, to a base model containing five risk indices (MAGGIC score, NT-proBNP, hsTnT, GDF15, and E/e′ ratio) incrementally improved 2-year mortality prediction (AUROC: 0.82→0.83 in PEOPLE; 0.74→0.79 in SHOP). Conclusion: Differential metabolic signatures tied to metabolic inflammation in HFpEF and impaired energy metabolism in HFrEF enhance risk stratification and point to therapeutic targets.