Evaluation of methodologies for risk assessment of combined toxic actions of chemical substances and establishment of PBTK/TD models for pesticides

Humans are simultaneously exposed to a number of chemicals via food and environment. These chemicals may have a combined action that causes a lower or higher toxic effect than would be expected from knowledge about the single compounds. Therefore, combined actions need to be addressed in the risk assessment process.

This Ph.D.-thesis provides an overview of the current knowledge on methods for risk assessment of combined actions of chemicals focussing on pesticides. Some of the methods are based on knowledge on the whole mixture and others are based on data on the single compounds in the mixture. The whole mixture approaches would be the ideal choice for assessment of e.g. pesticide residues in food.

However, they are normally not applicable since they require a large number of experimental data that are rarely available. This leaves the single compound approaches as the more realistic ones. The first step in the risk assessment of a mixture is to evaluate whether a group of compounds can be identified that induce a common toxic effect by a common mechanism of toxicity and therefore is suited for a cumulative risk assessment based on additivity.

Ideally the identification of a group of pesticides for cumulative risk assessment should be based on criteria providing the best and most robust grouping such as chemical structure, mechanism of action, common toxic mode of action or common toxic effect. Unfortunately, such data are seldom available for all of the compounds of concern.

Instead for pragmatic reasons, it is often more appropriate to consider the individual compounds as possible candidates for one (or more) cumulative assessment group(s). The cumulative risk assessment of this group will then be performed assuming simple similar action using one of the single compound approaches.

The hazard index based on a health based guidance value e.g. the acceptable daily intake (ADI) would normally be sufficient. However, the point of departure index is the most preferably method because it does not make use of a policy driven uncertainty factor and instead it is based on the most relevant toxicity data.

In case that more than one common mechanism group based on different simple similar actions are identified, they should be assessed separately. In addition, the potential for interactions between the groups (or single compounds) has to be considered. If no interactions are identified, simple dissimilar action can be anticipated and the response addition method should be used to assess the effect of the mixture.

In many cases the evaluators will probably tend to use very pragmatic approaches if lack of interaction between the compounds at the actual dose level can be assumed. This includes assuming all compounds in the mixture show dose additivity (simple similar actions). The hazard index or point of departure index would then be the preferred methods.

A crucial point in the assessment is whether there is interaction or no interaction between the compounds in the mixture. Although interactions among chemicals at high doses are wellknown, no single simple approach is currently available to judge upon potential interactions at the low dose levels of pesticide residues that humans are exposed to in food.

For this purpose, physiologically based toxicokinetic/toxicodynamic (PBTK/TD) modelling has been recommended as a tool to assess combined tissue doses and to help predict potential interactions including thresholds for such effects. Therefore, this thesis also focuses on such models and their applicability for use in risk assessment.

This type of model has been used for several years in the area of pharmacology but the use in the area of toxicology is relatively new. In a PBTK model the animal or man is described as a set of tissue compartments which is combined by mathematical descriptions of biological tissues and physiological processes in the body.

Thereby it is possible to quantitatively simulate the absorption, distribution, metabolism and excretion of chemicals and to predict the internal dose after exposure to the chemical (or metabolite) of concern. The PBTK models make it possible to extrapolate between species, from high-dose to lowdose, from route-to-route and between exposure scenarios.

In this way the risk assessor can simulate various scenarios including scenarios which cannot be studied experimentally. Models can be developed for subpopulations such as children and this may help the risk assessor determine whether special care should be taken for such groups. It is also possible to incorporate mechanistic information on interactions in the model and as mentioned above interaction threshold can be determined.

This would provide a helpful tool in the risk assessment of combined actions of chemicals. The PBTK model can be coupled with a toxicodynamic part in which the model attempts to estimate the effect resulting from the internal dose. The output of a PBTK model is linked to a toxicodynamic model by mathematical descriptions of the hypothesis of how compounds contribute to the initiation of cellular changes leading to the toxic responses.

In the present Ph.D. project a PBTK/TD model was established based on a previous published model. The model describes the organophosphorus pesticide chlorpyrifos and its metabolism to chlorpyrifos-oxon and 3,5,6-trichloro-2-pyridinol as well as the toxicodynamic of the chlorpyrifos-oxon i.e. inhibition of acetylcholinesterase activity in various tissues.

This paper was chosen because the model is on a relevant compound (a pesticide that is widely used) and one had the impression that the model work was described in details. The work in establishing this model clearly pointed out the importance that authors of such publications report their results with a high degree of transparency in order to enable colleagues to reproduce their work and e.g. evaluate it for further developing the model.

The model description should include model structure and equations as well as documentation of the choice of parameters and their origin. At present there is a lack of adequate data for use in the PBTK models and further studies in order to determine parameters for use in PBTK models are needed. The model developer is also forced to make assumptions and extrapolations.

It is of great importance that these are biologically based and explained. The PBTK/TD model on chlorpyrifos in the present project was used to illustrate how a no observed adverse effect level (NOAEL) can be established by the model and how to make extrapolations between species (rats and humans). The model underestimated the inhibition improvement before it can be used in risk assessment.

The PBTD modelling is still in its infancy and it will probably be better to put more effort into improving the toxicokinetic part especially including establishment of internationally acceptable reference values for various parameters before extending the model with a toxicodynamic part. PBTK models can be used to evaluate combined actions of a mixture of compounds.

In case of a mixture of compounds that do not interact (e.g. simple similar action) the PBTK modelling tool is useful to predict the combined doses in the target organ taking metabolism of the compounds into account. Such compounds should be dealt with in the PBTK models in the same way as single compounds.

In case of a mixture of interacting compounds mechanistic information on interactions can be incorporated in the PBTK/TD model and thereby it can be used e.g. to determine the interaction threshold. Such a model will consists of sets of identical equations, one set for each chemical as well as equations that specifically accounts for the interactions (e.g. competitive inhibition of metabolism in liver or induction of hepatic metabolism).

The development of PBTK models is complex and should only be used when it is considered essential. If adequate models are developed, they can provide better knowledge and understanding of the effects of mixtures in the organism and provide improved information on tissue dose levels and variations between species and within a population.

Moreover, scientifically supportable results about possible combined actions in humans after exposure to mixtures of pesticide residues in food would help making more reliable risk assessment. The PBTK models thus have a potential as an important tool in the risk assessment. Adequate documentation of the model is fundamental in order to increase the credibility of PBTK modelling.

Such credibility is crucial for a spreading of its use in risk assessment. This Ph.D.-project constitutes the initial work on implementing PBTK/TD models in the risk assessment of combined toxic action of chemical substances in food at the DTU National Food Institute. The work has revealed some major problems and pitfalls in the developing process.

However if reliable, these models will provide knowledge of the relationship between internal doses of the chemicals and the observed toxic effects and this knowledge will reduce the uncertainty in the risk assessment. Therefore, the work will continue implementing these models as a helpful tool in future risk assessment.