Modelling the intake of polychlorinated dibenzo-p-dioxins and dibenzofurans: impact of energy under-reporting and number of reporting days in dietary surveys

Abstract

A probabilistic long-term intake estimation of dioxins was carried out using food consumption data obtained from the National FINDIET 2007 Survey (Paturi et al. 2008). The study population consisted of 606 participants who were first interviewed with a 48-h recall and then filled in a 3-day food record twice. The concentrations of dioxins were obtained from previously published studies. The intake was estimated using a semi-parametric Monte Carlo simulation. The analyses were done separately for the whole study population and for the population excluding energy under-reporters. To diminish the impact of intra-individual variation and nuisance effects, adjustment with software (C-SIDE®) was also done after Monte Carlo simulation. It was found that when C-SIDE® was used, the 95th percentile of intake and its confidence limit was higher with 2 reporting days than with a higher number of days. However, with a crude intake estimation (no adjustment), the confidence intervals of the 95th percentile were also smaller with a higher number of days, but the 95th percentiles were higher with a higher number of reporting days. When under-reporters were excluded the intakes increased, but the impact of energy under-reporting was smaller with 8 reporting days than with 2 days and smaller using C-SIDE® than with a crude estimation. To conclude, adjustment for intra-individual variation and taking energy under-reporting into account are essential for intake estimation of dioxins with food consumption data of a limited number of reporting days.

Keywords:

• exposure, probability modelling
• risk assessment
• survey
• dioxins
• fish and fish products

Acknowledgement

This study was funded by the EU-funded project BENERIS.