Predictive models built upon annotated and validated intake biomarkers in urine using paired or unpaired analysis helped to classify cranberry juice consumers in a randomized, double-blinded, placebo-controlled, and crossover study

Abstract

Intake biomarkers of cranberry juice in women can assess consumption in clinical trials. Discriminant biomarkers in urine may explain urinary tract infection (UTI) preventive activities. We hypothesized that validated and annotated discriminant metabolites in human urine could be used as intake biomarkers in building predictive multivariate models to classify cranberry consumers. Urine samples were collected from 16 healthy women aged 18 to 29 years at baseline and after 3- and 21-day consumption of cranberry or placebo juice in a double-blind, crossover study. Urine metabolomes were analyzed using ultra high-performance liquid chromatography coupled with Orbitrap mass spectrometry. Paired and unpaired multivariate analyses were used to annotate or identify discriminant metabolic features after cranberry consumption. Twenty-six discriminant metabolic features (paired analysis) and 27 (unpaired analysis) after cranberry consumption in an open-label intervention were rediscovered in the blinded study. These metabolites included exogenous (quinic acid) and endogenous ones (hippuric acid). The paired analysis showed better model fitting with partial least-square discriminant analysis models built on all metabolites than the unpaired analysis. Predictive models built on shared metabolites by the unpaired analysis were able to classify cranberry juice consumers with 84.4% to 100% correction rates, overall better than the paired analysis (50%-100%). The double-blind study validated discriminant metabolites from a previous open-label study. These urinary metabolites may be associated with the ability of cranberries to prevent UTIs and serve as potential cranberry intake biomarkers. It reveals the importance of selecting the right predictive models to classify cranberry consumers with higher than 95% correction rates.