Pharmakokinetic Modeling to Support the Statistical Analysis of Blood Dioxin Concentration

ISEE-1696

Methods:

We develop a method to convert historical intake of dioxin-like compounds of a given individual into a 2005 actualized intake value. Three successive correction factors need to be applied. 1) Present food intakes are corrected to account for the change of intake with age, as a child eats less than an adult. 2) The second factor accounts for historical changes in dioxin concentration in the food chain, with a peak around 1968. 3) The half-life of each congener in the body is used to decay past intake over time. This approach is used as a pharmacokinetic (PK) pre-treatment of food intake variables-expressed in a 2005 actualized number of meals over lifetime-before performing the statistical regression analysis. It also provides insights on an adequate statistical model for predicting blood as a function of food intake and age. The method is applied to the University of Michigan Dioxin Exposure Study in Midland and Saginaw, involving more than 950 participants whose serum Dioxin concentration was sampled. Results of various linear and non linear models for the statistical analysis are systematically compared with a physiologically based pharmacokinetic model (PBPK).

Results:

In a traditional multiplicative analysis, age is the dominant parameter explaining variability in Dioxin concentration in blood. Using the actualized number of meals over lifetime enables us to explain more than 60% of the variability in blood concentration, with neither age nor BMI having a significant influence. This PK corrected model is compared with a non-linear hybrid model, composed of an exponential term for factor affecting decay rate (age, BMI, breastfeeding, smoking status) that multiplies a multi-linear sum of diet intakes, leading to comparable R² higher, close to 60% or higher. This model can be applied for the 29 WHO congeners and demonstrate a strong correlation between the age coefficient and the congener half-life in blood connected in the literature.

Conclusion:

The PBPK model enables us to better discuss the respective contributions of each food items considered in the food questionnaire to the population blood concentration.