CIIT Centers for Health Research, Center for Computational Biology and Extrapolation Modeling, 6 Davis Drive, Research Triangle Park, NC 27709
ABSTRACT
The article highlighted in this issue is "A PBPK Modeling-Based Approach to Account for Interactions in the Health Risk Assessment of Chemical Mixtures" by Sami Haddad, Martin Béliveau, Robert Tardif, and Kannan Krishnan (pp. 125131).
The mixture toxicity problem has long challenged toxicologists and regulators (e.g., Jacobson et al., 1958; Jacobziner, 1962). The main public health concerns with mixtures are (1) that interactions of mixture components may lead to toxicities not seen with individual components and (2) that a mixture may exhibit synergism, i.e., that its potency may be much greater than expected based on knowledge of the components. In these situations the mechanism underlying the toxicity of concern includes one or more interactions of mixture components. As Haddad and colleagues point out in the featured article, the mechanism that generates the toxic response and is a prime determinant of the putative human health risk thus cannot be examined adequately by only studying the individual components of the mixture.
Biologically motivated models, particularly physiologically based pharmacokinetic (PBPK) models, are playing increasingly important roles in toxicology and risk assessment (e.g., Bogdanffy et al., 2001; Delic et al., 2000; Frederick et al., 2001; Jonsson et al., 2001). Perhaps the greatest strength of these models is their capability for extrapolation of target tissue dosimetry across routes of exposure, among species, and from high to low doses. This capability derives from the quantitative description of the relevant mechanisms, albeit the mechanisms abstracted to the level of the modes of action, and the ability to quickly and cheaply exercise the model on a desktop computer.
Standard default approaches to mixture risk assessment consider doses and responses of the mixture components to be additive (U.S. EPA, 1986). This assumption is reasonable given the common problem of conducting risk assessments in the absence of data needed for a science-based assessment. Significant uncertainties are inherent in this default approach, however. We now know very well that nonlinearities in the relationship between applied dose and tissue dose affect the overall dose-response relationship, as was shown years ago by Gehring et al. (1978) for vinyl chloride. While good progress has been made in the study of pharmacokinetic nonlinearities, we remain largely ignorant of the relationship between effective tissue dose and ultimate toxic effect at the level of the mechanism or mode of action. There are, no doubt, nonlinearities in this domain. Thus, even at the level of single compounds, we are uninformed of much of the biology that determines the shape of the dose-response curve. When the possibilities of pharmacokinetic and pharmacodynamic interactions of the components in a complex mixture are also considered, we can clearly see that the uncertainties in exposure-response characterization are large and that much basic biological and toxicological research remains to be done. Fortunately, the emerging technologies of genomics and proteomics hold great promise for providing an entrée into this difficult problem.
When uncertainties about mechanisms of toxicity for single compounds and their mixtures are large, then quantitative models of these mechanisms and the extrapolations obtained with them can be no less uncertain. Perhaps the greatest danger in our attempts to describe mechanisms quantitatively is not lack of knowledge per se but rather the use of speculative assumptions about the mechanism. As Mel Andersen often says, "It's what you know that's not so that gets you in trouble." Computer-generated graphics and impressive flow chart diagrams of the model structure can lend credibility where none is deserved. Every effort should be made during model development and in the equally important documentation of the model to clearly distinguish between what is known with some reasonable level of confidence and what is speculative. The basic model structure should reflect what is known with confidence. Extensions of the model that include speculative elements should be described as such. It is worth noting however that while models of this latter type can be misleading when used for risk assessment, their use for hypothesis generation can be useful in a research setting.
Recognition of the uncertainties in our understanding of mechanisms and of the fact that quantitative models of these mechanisms inherit these uncertainties can provide a reality check on our expectations of the models. Extrapolations are accurate to the degree that mechanisms and interactions among mixture components are described accurately. The model, when based on a solid biological foundation, can provide a valuable leveraging of our knowledge, as is nicely demonstrated in the Haddad article. One of the great challenges for both modelers and regulators in realizing the full value of these models to risk assessment is to identify, as accurately as possible, the point at which the solid biological foundation ends and speculative, hypothesis-generating modeling begins.
In the feature article, Haddad et al. describe competitive interactions at the level of metabolism for a five-component mixture. The pharmacokinetics of each of the mixture components has been well studied, and relatively mature PBPK models are available for each component. The most significant assumption used by Haddad and colleagues is that competitive interactions at metabolic enzymes are the only significant pharmacokinetic interactions for the five-component mixture. This assumption is consistent with the data and has been carefully evaluated by Haddad and coworkers in earlier studies (Haddad et al., 1999, 2000
). The model provides interesting predictions with significant implications for risk assessment, particularly the increased flux of methylene chloride down the GSH conjugation pathway when other mixture components compete with methylene chloride for metabolism at CYP450 and the decreased generation of benzene metabolites due to the same competitive interactions. These predictions illustrate another significant benefit of biologically motivated quantitative models: predictions of behaviors that are not seen experimentally but are testable in the laboratory. Confirming or rejecting at least some of the model-generated predictions in the laboratory is always desirable. Confidence in the model increases when its predictions are shown to be consistent with experimental data, while identification of inconsistencies illuminates our ignorance. Modeling is thus most useful for risk assessment when closely tied to experimental work, with the model serving as a natural bridge between the laboratory and risk assessment. The work by Haddad et al. is a good example of quantitative modeling based on an adequate foundation of experimental work. The modeling leverages the data to provide value, in the form of quantitative predictions of tissue dosimetry for the mixture components, that would not otherwise be available.
Finally, the competitive pharmacokinetics interactions studied by Haddad et al. are often dismissed as being a high-dose phenomenon with little import for low-level, real-world exposures. The model described by Haddad et al. can predict very small changes in flux through toxicologically important pathways, such as the glutathione-dependent metabolism of methylene chloride. When lifetime cancer risks of 106 are considered significant for regulatory purposes, then small changes in flux through important pathways may have regulatory significance, particularly if integrated over significant fractions of the human lifetime. The models can be used to evaluate these possibilities. The value of such exercises depends to a great extent on the availability of relevant exposure data. We should also remember that adequate characterization of pharmacokinetic behavior for single compounds and for mixtures is only the first step in the science-based approach to exposure-response assessment. The great black box of pharmacodynamics is waiting. Interactions significantly more complex than competition for metabolism must exist and may confound default-based risk assessments, and even risk assessments based on the much more sophisticated approach described by Haddad et al. The confluence of modeling approaches like that described by Haddad et al. and the new high throughput technologies in the laboratory make this a most exciting time to study mechanisms of toxic action and develop new quantitative methodologies in support of risk assessment.
NOTES
1 For correspondence via fax: (919) 558-1300. E-mail: rconolly{at}citt.org.
REFERENCES
Bogdanffy, M. S., Plowchalk, D. R., Sarangapani, R., Starr, T. B., and Andersen, M. E. (2001). Mode-of-action-based dosimeters for interspecies extrapolation of vinyl acetate inhalation risk. Inhal. Toxicol. 13, 377396.[ISI][Medline]
Delic, J. I., Lilly, P. D., MacDonald, A. J., and Loizou, G. D. (2000). The utility of PBPK in the safety assessment of chloroform and carbon tetrachloride. Regul. Toxicol. Pharmacol. 32, 144155.[ISI][Medline]
Frederick, C. B., Gentry, P. R., Bush, M. L., Lomax, L. G., Black, K. A., Finch, L., Kimbell, J. S, Morgan, K. T., Subramaniam, R. P., Morris, J. B., and Ultman, J. S. (2001). A hybrid computational fluid dynamics and physiologically based pharmacokinetic model for comparison of predicted tissue concentrations of acrylic acid and other vapors in the rat and human nasal cavities following inhalation exposure. Inhal. Toxicol. 13, 359376.[ISI][Medline]
Gehring, P. J., Watanabe, P. G., and Park, C.N. (1978). Resolution of dose-response toxicity data for chemicals requiring metabolic activation: examplevinyl chloride. Toxicol. Appl. Pharmacol. 44, 581591.[ISI][Medline]
Haddad, S., Tardif, R., Charest-Tardif, G., and Krishnan, K. (1999). Physiological modeling of the pharmacokinetics interactions in a quaternary mixture of aromatic hydrocarbons. Toxicol. Appl. Pharmacol. 161, 249257.[ISI][Medline]
Haddad, S., Charest-Tardif, G., Tardif, R., and Krishnan, K. (2000). Validation of a physiological modeling framework for simulating the toxicokinetics of chemicals in mixtures. Toxicol. Appl. Pharmacol. 167, 199209.[ISI][Medline]
Jacobson, K. H., Rinehart, W. E., Wheelwright, H. J., Jr., Ross, M. A., Papin, J. L., Daly, R. C., Greene, E. A., and Groff, W. A. (1958). The toxicology of an aniline-furfuryl alcohol-hydrazine vapor mixture. Am. Ind. Hyg. Assoc. J. 19, 91100.[Medline]
Jacobziner, H., (1962). Mixture of tranquilizers, lighter fluid, paint thinner, and iodine poisonings. N.Y. J. Med. 62, 862864.
Jonsson, F., Bois, F., and Johanson, G. (2001). Physiologically based pharmacokinetic modeling of inhalation exposure of humans to dichloromethane during moderate to heavy exercise. Toxicol. Sci. 59, 209218.
U.S. EPA (1986). Guidelines for the Health Risk Assessment of Chemical Mixtures. U. S. Environmental Protection Agency, EPA/630/R-98/002. Fed. Regist. 51, 3401434025.
|