Is there an optimal sampling time and number of samples for assessing exposure to fast elimination endocrine disruptors with urinary biomarkers?
- Competence Center for Methodology and Statistics
In studies investigating the effects of endocrine disruptors (ED) such as phthalates, bisphenols and some pesticides on human health, exposure is usually characterized with urinary metabolites. The variability of biomarkers concentration, due to rapid elimination from the body combined with frequent exposure is however pointed out as a major limitation to exposure assessment. This study was conducted to assess variability of urinary metabolites of ED, and to investigate how sampling time and number of samples analyzed impacts exposure assessment. Urine samples were collected over 6months from 16 volunteers according to a random sampling design, and analyzed for 16 phthalate metabolites, 9 pesticide metabolites and 4 bisphenols. The amount of biomarkers excreted in urine at different times of the day were compared. In parallel, 2 algorithms were developed to investigate the effect of the number of urine samples analyzed per subject on exposure assessment reliability. In the 805 urine samples collected from the participants, all the biomarkers tested were detected, and 18 were present in >90% of the samples. Biomarkers variability was highlighted by the low intraclass correlation coefficients (ICC) ranging from 0.09 to 0.51. Comparing the amount of biomarkers excreted in urine at different time did not allow to identify a preferred moment for urine collection between first day urine, morning, afternoon and evening. Algorithms demonstrated that between 10 (for monobenzyl (MBzP) phthalate) and 31 (for bisphenol S) samples were necessary to correctly classify 87.5% of the subjects into quartiles according to their level of exposure. The results illustrate the high variability of urinary biomarkers of ED over time and the impossibility to reliably classify subjects based on a single urine sample (or a limited number). Results showed that classifying individuals based on urinary biomarkers requires several samples per subject, and this number is highly different for different biomarkers.