A metabolomics-driven approach to predict cocoa product consumption by designing a multimetabolite biomarker model in free-living subjects from the PREDIMED study

Abstract

SCOPE: The aim of the current study was to apply an untargeted metabolomics strategy to characterize a model of cocoa intake biomarkers in a free-living population. METHODS AND RESULTS: An untargeted HPLC-q-ToF-MS based metabolomics approach was applied to human urine from 32 consumers of cocoa or derived products (CC) and 32 matched control subjects with no consumption of cocoa products (NC). The multivariate statistical analysis (OSC-PLS-DA) showed clear differences between CC and NC groups. The discriminant biomarkers identified were mainly related to the metabolic pathways of theobromine and polyphenols, as well as to cocoa processing. Consumption of cocoa products was also associated with reduced urinary excretions of methylglutarylcarnitine, which could be related to effects of cocoa exposure on insulin resistance. To improve the prediction of cocoa consumption, a combined urinary metabolite model was constructed. ROC curves were constructed to evaluate the model and individual metabolites. The AUC values (95% CI) for the model were 95.7% (89.8-100%) and 92.6% (81.9-100%) in training and validation sets, respectively, whereas the AUCs for individual metabolites were <90%. CONCLUSIONS: The metabolic signature of cocoa consumption in free-living subjects reveals that combining different metabolites as biomarker models improves prediction of dietary exposure to cocoa.