Reproducibility and sources of variation of urinary biomarkers of food intake of fruits, vegetables and chocolate in European children and adolescents.

Background: Biomarkers of food intake may improve dietary assessment. Thereby, a key concern is their reproducibility over time. In epidemiological studies, it is important to accurately estimate habitual food intake and consequent disease risk associations. Objectives: We aimed to assess the reproducibility of 12 urinary metabolites linked to food intake and to investigate potential sources of their variation. Methods: The analyses are based on previously identified urinary metabolites associated with dietary intake of fruits, vegetables, and chocolate in the large-scale European Identification and Prevention of Dietary- and Lifestyle-Induced Health Effects in Children and Infants/ I.Family study. Metabolites were measured in 1788 urine samples from 599 children at study baseline (2007/2008, n = 597), at the first follow-up (2009/2010, n = 596), and at the third follow-up (2013/2014, n = 595) using high-resolution liquid chromatography–MS. Unadjusted and adjusted intraclass correlation coefficients (ICCs) were calculated for 2- and 4-y intervals. To identify sources of biomarker variability, various factors, including dietary intake, were analyzed. The amount of variance explained by each factor was quantified using the partial coefficient of determination (R 2 ). Results: The median ICCs were 0.27 (range: 0.11–0.54) and 0.28 (range: 0.15–0.51) over 2- and 4-y intervals, respectively. Individual factors explained a median of 17% (range: 9.8–42.4) of the variance for the 2-y interval and 14.6% (range: 8.3–43.8) for the 4-y interval. Country of residence explained the largest proportion of variance (median: 5% for the 2-y interval, 4.5% for the 4-y interval). Dietary intake explained only a variation of 0.7% (0.0–1.5) and 0.6% (0.0–1.1) for the 2- and 4-y interval, respectively. Conclusions: The reproducibility of urinary metabolites was poor to moderate over the 2- to 4-y periods, and only part of the variability could be explained by the studied factors. Future studies should explore shorter time intervals and other sources of variation, e.g., the influence of the gut microbiome and genetic factors.