Bayesian modelling of long-term dietary intakes from multiple sources

Area of Science:

• Environmental Chemistry
• Risk Assessment
• Statistical Modeling

Background:

• Human exposure to pesticides and chemicals occurs through combined food and drink consumption.
• Probabilistic risk assessments quantify population-level daily exposures using residue and consumption data.

Purpose of the Study:

• To present a novel statistical method for estimating the distribution of total daily chemical exposures.
• To model correlations in consumption frequency and amounts across multiple food types.

Main Methods:

• Developed a new statistical method using dietary survey and residue monitoring data.
• Employed Bayesian approaches to quantify uncertainty, comparing three models: multivariate, independent parametric, and aggregated parametric.
• Applied the method to UK children's consumption data for pesticides like pyrimethanil, captan, and chlorpyrifos.

Main Results:

• The importance of accounting for between-food correlations in consumption was demonstrated.
• Model (i) and (iii) highlighted the significance of correlations, while model (iii) showed potential discrepancies with bimodal aggregated intake distributions.
• The influence of residue uncertainty on exposure estimations was also evident.

Conclusions:

• A new statistical method improves the estimation of population exposure distributions to chemicals from diet.
• Accounting for correlations between food consumption patterns is essential for accurate risk assessment.
• Further research should consider the impact of residue uncertainty and complex intake distributions.