Title

Peripheral metabolite pattern in heart failure patients is an independent predictor of survival

Document Type

Conference Proceeding

Publication Date

8-2015

Publication Title

J Card Fail

Abstract

Background: Measurement of small molecules of intermediary metabolism is a method for characterizing biologic and disease states. Heart failure (HF) is an ideal candidate for this since evidence indicates that energetic derangements contribute to disease progression and metabolite correlations with survival have been described. We set out to assess whether a targeted metabolomics profile could predict survival in HF independent of established predictors.

Methods: We analyzed plasma samples from 400 patients with chronic HF. All participants met Framingham criteria for HF. Data on demographics, ejection fraction, and comorbid conditions were collected, a blood sample was obtained and aliquoted plasma stored at -70C. Eight-six amino acids (AA), organic acids (OA), and acylcarnitines (AC) were quantified using targeted metabolomic profiling. Analytes with a coefficient of variation<0.05 were considered non-variable and not analyzed further. Association of metabolites with survival time was assessed univariably using Cox proportional hazards regression. A lasso penalized version of multivariable Cox regression was used to construct a multi-metabolite survival score, and we performed 10-fold cross validation to obtain an unbiased estimate of the predictive value of the score (dichotomized at the median to high vs. low risk). We then retested the value of the metabolite risk category in a Cox model adjusted for conventional covariates including natriuretic peptide level (age, gender, race, ejection fraction, CAD, COPD, creatinine, Afib, stroke, and NTproBNP).

Results: The cohort was 50% African American, 50% female, 67% HFrEF, and had an average age of 70 years. Eleven metabolites had significant associations to survival time, with α-ketoglutarate being the strongest (p=1.24 x 10-9). Cross-validation of the lasso penalized model yielded a significant multi-metabolite predictor (p=3.73 x 10-5); the dichotomized risk score was associated with a 2.4 fold risk of death (Figure 1). When the multi-metabolite risk category was added to the conventional predictive model (which included NTproBNP, p=1.98 x 10-4), the multi-metabolite predictor remained independently associated with survival (HR 2.59, p=2.38 x 10-4).

Conclusion: Plasma metabolite profile was a strong predictor of survival among HF patients, independent of clinical factors and NTproBNP. The key drivers included citric acid cycle intermediates, arginine, and short-chain acylcarnitines. Validation in a larger independent data set is warranted.

Volume

21

Issue

8 Suppl

First Page

S14

This document is currently not available here.

Share

COinS