Untargeted metabolomics-based approach to discover biomarkers of apple intake: A randomized controlled trial.

BACKGROUND

Biomarkers of food intake (BFIs) provide an objective strategy for dietary assessment and may help to reduce the biases of traditional self-reported methods. So far, existing biomarkers of apple intake remain limited in terms of practical applicability.

OBJECTIVE

This study aimed to identify urinary biomarkers of apple intake with extended detection windows, compare their postprandial kinetics, and develop a multi-biomarker panel for identifying recent apple consumers.

METHODS

A randomized, controlled, crossover dietary intervention was conducted in 18 healthy adults. After a 2-day standardized run-in period, participants completed single-exposure, repeat-exposure and follow-up phases. Urine samples collected at multiple time points, along with original fruit and in vitro digested samples, were analyzed using untargeted metabolomics via dual-column ultra-high-performance liquid chromatography-mass spectrometry (UHPLC-MS). A systematic statistical pipeline based on postprandial kinetic profiles was applied to identify candidate biomarkers with prolonged postprandial responses in 24-h urine. Additionally, a complementary screening strategy focusing on the 0-6 h postprandial window was implemented to characterize early-responsive metabolites. XGBoost machine learning models were applied to establish an optimal multi-biomarker panel for assessing apple intake under free-living conditions.

RESULTS

Five distinct metabolites were identified as candidate urinary biomarkers for apple intake. These metabolites remained significantly elevated at 24 h. In parallel, 28 early-responsive metabolites were discovered within the 0-6 h postprandial phase, which included several previously reported apple biomarkers, such as phloretin-2'-O-glucuronide. Individual metabolites showed limited predictive performance under free-living conditions, whereas a four-metabolite panel achieved the best discriminative performance (AUC = 0.80) for predicting apple consumption within 1-2 days compared to other panels.

CONCLUSIONS

This study identified urinary biomarkers of apple intake with a detection window extending to at least 24 h and established a multi-biomarker panel for identifying recent apple intake.