Novel multimetabolite prediction of walnut consumption by a urinary biomarker model in a free-living population: the PREDIMED Study

Abstract

The beneficial impact of walnuts on human health has been attributed to their unique chemical composition. In order to characterize the dietary walnut fingerprinting, spot urine samples from two sets of 195 (training) and 186 (validation) individuals were analyzed by an HPLC-q-ToF-MS untargeted metabolomics approach, selecting the most discriminating metabolites by multivariate data analysis (VIP ≥ 1.5). Stepwise logistic regression analysis was used to design a multimetabolite prediction biomarker model. The global performance of the model and each included metabolite in it was evaluated by receiver operating characteristic curves, using the area under the curve (AUC) values. Dietary exposure to walnuts was characterized by 18 metabolites, including markers of fatty acid metabolism, ellagitannin-derived microbial compounds, and intermediate metabolites of the tryptophan/serotonin pathway. The predictive model of walnut exposure included at least one compound of each class. The AUC (95% CI) for the combined biomarker model was 93.4% (90.1-96.8%) in the training set and 90.2% (85.9-94.6%) in the validation set. The AUCs for individual metabolites were ≤85%. As far as we know, this is the first study proposing a combination of biomarkers of walnut exposure in a population under free-living conditions, as considered in epidemiological studies examining associations between diet and health outcomes.