Introduction

The overall risk assessment process integrates hazard assessment data on chemical toxicity with exposure assessment information (Figure 1). Hazard assessment is the process by which the toxicity of a chemical is determined either by a series of bioassay experiments with intact test animals or by observing increased morbidity/mortality in exposed humans. Often there is no human epidemiology on particular chemicals, and risk managers have to rely solely on results of animal toxicity experiments for the hazard assessment. These animal experiments allow us to generate dose-response information on how much chemical is required to produce a specified degree of toxicity in test animals. The major challenges in the hazard assessment process are to generalize toxicity results in the test animal to (1) predict what will happen in test animals given much lower amounts of chemical; (2) predict what will happen in an entirely different species of animal, namely, humans; and (3) predict what will happen in a different species receiving a chemical by a route of administration different from that used in the animal studies. These are all problems of extrapolating beyond the conditions used in the actual toxicity studies to predict outcome under very different exposure conditions in a variety of species. What concepts tie these problems together and give us some confidence in the ultimate success of efforts to develop methods to conduct these extrapolations?

Figure 1

Elements of chemical risk assessment. The overall risk assessment process consists of both a hazard assessment and an exposure assessment. These provide information on which to make a risk analysis and give the risk manager detailed information on which (more...)

Basically, there seem to be two fundamental assumptions which toxicologists are forced to make in attempting quantitative extrapolations based on animal toxicity experiments. The first is that experimental animals are true surrogates for exposed humans. That is, chemicals would cause effects in the same tissues in humans as those tissues in which they cause effects in the exposed test animals, and the mechanisms of these effects would be qualitatively similar in the two different species. This assumption accepts that there is a qualitative similarity in effects in different species. There are instances where this assumption is suspect. For instance, vinyl chloride causes zymbal gland cancer in rats (Maltoni and Lefemine, 1974), but humans do not have this structure. However, vinyl chloride is obviously carcinogenic in a variety of animal species. In any case, the universal validity of this assumption of qualitative similarity in toxicity is really not within the purview of the papers in this volume. It does seem to be valid in the great majority of instances. Instead, this volume focuses our attention on the other basic assumption that we are forced to make to conduct our risk assessment calculations.

This second tenet is that there is a quantitative equivalence in the tissue chemical exposure required to produce an equivalent intensity of biological effect in various species. This is the concept of tissue dose equivalence. More simply stated, all species are regarded to have equal sensitivity to the toxic chemical. Again, there are notable exceptions, such as the extreme interspecies differences in toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin (Kociba and Schwetz, 1982). For more simple toxicities, related to reactive chemical moieties, this assumption seems entirely appropriate. In addition, the species and strain differences in dioxin toxicity might diminish substantially if the dose were expressed in relation to the concentration and affinity of the dioxin receptor(s) in these various animal species (Poland and Knutson, 1982). The catch, of course, is that tissue dose is not a simple concept and will be different for different chemicals. In fact, the real problem in hazard assessment is defining and measuring tissue dose under a variety of exposure conditions in several species.

A Dose of What?

This hazard assessment process sounds deceptively simple. Determine the toxic tissue dose in the test species and calculate the exposure conditions under which this dose is likely to be achieved in humans. All we need to do is to define tissue dose. As a working definition, we can say that an appropriate measure of tissue dose is some measure of the intensity of chemical exposure which is directly linked to the biological processes leading to toxicity or tumor formation. With this definition, it is clear that some presumption of the mechanism of interaction between the chemical and the tissue is required before we can define tissue dose for any particular chemical.

What then are the primary processes by which chemicals interact with tissue constituents to cause biological changes in the tissue? The first process is by direct chemical reaction in which the toxic chemical reacts with and consumes cellular constituents (Figure 2). With this type of interaction the expected degree of damage, as loss of cellular constituents or accumulation of bound reactive intermediate, should be related to the time integral of tissue exposure to the reactive chemical. This time integral of tissue exposure is also called the area under the tissue concentration curve for the reactive chemical. The equations for reactivity in Figure 2 are true only for acute-exposure situations. In chronic administration, the equation should be expanded to include terms for the synthesis and normal catabolism of the macromolecules.

Figure 2

Tissue dose metrics and their relation to toxicity. Toxic chemicals interact with tissues by two general processes. In one case, chemical reactivity, the toxic chemical, T, reacts with cellular macromolecules, MM, to cause covalent binding of the toxic (more...)

The second common process by which chemicals interact with tissue is by noncovalent binding to cellular receptor molecules. This is the mechanism by which dioxin is presumed to interact to initiate toxic changes in cells. This binding with concomitant changes in receptor occupancy causes some response on the part of the organism which is ultimately expressed as toxicity. The therapeutic action of most drugs is also related to specific receptor binding (Goldstein et al., 1974). With this type of interaction, the response of the cell is dependent on the occupancy of the receptor and occupancy is determined by the binding constant for the chemical and the free concentration of chemical in the cell. Thus, there are a variety of quantitative measures of tissue dose which may be regarded as the appropriate measures of the intensity of tissue exposure (Figure 3). In this paper these parameters which are proportional to the relevant measure of tissue exposure are referred to as tissue dose metrics. These metrics include estimates of time integrals of tissue exposure to a chemical or its metabolite(s), concentrations of these materials in tissues, or receptor occupancy caused by the presence of these chemicals in tissues. The choice of which of these metrics to use as the appropriate measure of tissue dose now depends on some knowledge of the mechanism by which the toxic effects are induced.

Figure 3

Some potential tissue dose metrics for toxic chemicals.

In this usage mechanism does not mean an exhaustive, complete description of the entire set of events associated with toxicity. It relates instead to certain general aspects of the nature and causes of a particular toxic event. For instance, is the effect related to chemical reactivity or to occupancy of cellular receptor molecules? Is the effect associated with parent chemical or with a metabolite? If it is a metabolite, does the metabolite have a sufficiently long half-time in the body to circulate freely throughout the body or is it so reactive that it never leaves the organ(s) in which it is formed? In terms of the effects themselves, are they essentially reversible cytotoxic phenomena or irreversible carcinogenic transformations? If it is cancer induction, is the process one of direct genotoxicity or is it epigenetic in origin and associated with either induced hyperplasia as a result of cytotoxicity or with tumor-promoting effects of the chemical?

These are some of the more important aspects of mechanism that must be considered in establishing the correct metric for expressing tissue dose.

Isn't This Volume about Pharmacokinetics?

The later portions of this chapter discuss tissue dose for several classes of chemical carcinogens. But right now, the reader may be wondering what tissue dose has to do with the theme of this volume—pharmacokinetics. Well, if determining and measuring the appropriate tissue dose is the problem in hazard assessment, then pharmacokinetic modeling is an indispensable tool in estimating tissue dose for a variety of exposure conditions. Pharmacokinetic modeling contributes to the process by which we translate from obvious measures of administered dose, such as amount of chemical instilled into the stomach or concentration in the inspired air, to estimate the more relevant measures of tissue dose which may not always be accessible to measurement by direct experimentation.

The development of a pharmacokinetic model for use in chemical risk assessment begins with identification of a toxic effect in a particular tissue. Based on some limited knowledge of the mechanism of toxicity and results from the literature, the appropriate measure of tissue dose is deduced, or alternatively, several potential measures of tissue dose might be proposed. Next, an analytical pharmacokinetic model should be developed to predict these relevant measures of tissue dose under a wide range of exposure conditions. What elements, if possible, need to be included in a useful pharmacokinetic model? It should contain structures to account for routes of administration, major storage tissues within the body, primary tissues involved with elimination, target tissues, and sufficient biochemical detail within target tissues to calculate the presumed measures of tissue dose. These models should be validated as much as possible with kinetic data or with ancillary experimentation to assess model parameters by experiments separate from the kinetic studies. The successful kinetic models can then be used to estimate tissue dose and correlate it with observed toxicity. Hazard assessment calculations for human exposures are subsequently conducted based on the expected human target tissue exposures under various exposure conditions.

In work in our laboratory in Dayton, Ohio, developing pharmacokinetic models for use in chemical risk assessments, we have relied heavily on use of so-called physiologically based pharmacokinetic models—PB-PK models (see H. J. Clewell III and M. E. Andersen, this volume). These models contain considerable physiological and biological information (Bischoff and Brown, 1966) and are amenable to interspecies extrapolation, a process which is essential for predicting human hazard based on results of animal toxicity studies (Dedrick, 1973). Gerlowski and Jain (1983) have provided a very good review of the status of PB-PK modeling of chemical disposition. These PB-PK models describe the body in terms of realistic tissue compartments with specified volumes, blood flows, partition coefficients, and tissue binding characteristics (Gargas et al., 1986; Ramsey and Andersen, 1984). Biochemical constants for metabolic pathways and for tissue binding can be included in the mass-balance equations for organs in which these interactions are important. For most of these metabolic pathways the important constants are the maximum velocity of the reaction (V_max; in milligrams per kilogram) and the binding affinity of the particular substrate for the metabolizing enzyme (in milligrams per liter). Complex metabolic pathways involving parallel or sequential reactions of the parent chemical or involving interactions between chemical metabolism and cofactor depletion can also be readily incorporated into these models, as necessary (H. J. Clewell III and M. E. Andersen, this volume).

The entire process of problem definition, tissue dose assignment, and pharmacokinetic model development can be captured in a simplistic flow diagram (Figure 4). In this representation the process of model formulation comes after evaluation of the nature of the problem and consideration of the impact of mechanism on the choice of tissue dose metric. It is followed by exercising the model, evaluating its success at predicting known kinetic and toxicity behavior, designing necessary experiments to collect crucial data for verifying or improving model performance, and refining the model when needed. A successful model can then be used as an integral part of the hazard assessment process. The take-home lesson here is that pharmacokinetic modeling is not some knee-jerk process where the investigator collects blood time course curves and draws limited inferences about the behavior of the chemical in the body by an abstract mathematical curve-fitting procedure. Instead, the pharmacokinetic modeling intended for risk assessment use is, an integrating process which should be done early on before major data collection efforts. It should provide a comprehensive description of chemical disposition in target organs and be designed to predict human kinetic behavior when the biochemical metabolic constants and the tissue-binding characteristics of the chemical have been determined in human tissues.

Figure 4

Simplified flow chart of the development of a pharmacokinetic model intended for use in risk assessment. Problem identification: the finding of a particular toxicity, in a particular organ(s), in a particular species. This effect is the benchmark on which (more...)

The remainder of this chapter discusses tissue dose for various mechanisms of carcinogenesis, identifies essential elements required in PK models for tracking these particular forms of tissue dose, and emphasizes that pharmacokinetic model development will often suggest a need to collect critical metabolic or kinetic data that might not be available from the literature. In fact, it would be completely wrong to believe that PK models should be developed on existing toxicity data bases. The existing literature can be helpful for model definition, for drawing conclusions about the nature of appropriate measures of tissue dose, and for providing limited PK information, but it is also replete with experiments which are virtually useless for hazard assessment. If a new approach, such as PB-PK modeling, is proposed as an adjunct to existing hazard assessment techniques, it will have data requirements of its own and require some independent experimentation not previously conducted on a routine basis for each chemical for which a risk assessment was planned. In general, this does not mean that there has to be major new data acquisition needs for each PK model that might be developed for risk assessment use. For many cases, this will be only limited, critical experiments that are required to provide important constants for use in the PK model (Figure 4) or to fill data gaps identified in the literature survey.

Genotoxic Carcinogens

In terms of the chemical carcinogens themselves, there are two broad mechanisms by which chemicals cause cancer: by some direct chemical interaction with the DNA structures of the cell or by indirect effects on the cellular environment which increase tumor yield without direct chemical alteration of DNA. The former are called genotoxic carcinogens and the latter, epigenetic carcinogens (Weisburger and Williams, 1980). As might be expected, the distinction between these two categories of carcinogens is not always clear-cut. Many substances appear to possess properties characteristic of both categories of carcinogens. However, this division can be profitably examined in terms of the importance of a proper definition of tissue dose for both types of chemical carcinogens.

Genotoxic chemical carcinogens themselves can be further subdivided on the basis of whether parent chemical or a metabolite is the moiety that reacts with DNA. The possibilities include cases where parent chemicals, such as ethylene oxide or dimethylsulfate, are genotoxic; cases where stable metabolites, such as ethylene oxide formed from ethylene or butadiene epoxides formed from butadiene, are genotoxic; and cases where reactive, nonisolatable metabolites, such as the epoxide formed from vinyl chloride or the chloromethylgluthathione formed from methylene chloride, are presumed to be responsible for genotoxicity (Figure 5). These three possibilities for the nature of the DNA-reactive chemical need to be considered independently.

Figure 5

Classification of genotoxic carcinogens: Genotoxic chemicals are defined as those materials which are sufficiently reactive in vivo to react with DNA bases to form adducts. The presence of these adducts increases the probability of a mutational event (more...)

Parent Chemical

For the simplest case there is a chemical reaction between DNA and parent chemical leading to chemical alteration of the DNA which can cause mutation during cell replication. As discussed for cases of chemical reactivity, the tissue burden of altered DNA is expected to be associated with integrated tissue exposure to the DNA-reactive chemical. The modeling problem is to identify the chemical-specific tissue solubilities or tissue-binding characteristics and the distribution and activity of chemical-specific detoxifying enzymes in various tissues. The goal of the PK model is to identify and understand the metabolic and physiological processes that limit the action of the parent chemical in the cells.

Interspecies scaling is the determination of how target tissue exposure is affected by animal size for a particular administered dose. An attempt was made in volume 6 of Drinking Water and Health (National Research Council, 1986) to predict the interspecies scaling of tissue dose, depending on the nature of the toxic moiety—for example, parent or metabolite, etc. This analysis assumed that both metabolic and physiological clearances scale as a partial power of body weight and that both tissue volumes and general volumes of distribution are directly proportional to body weight. For these idealized pharmacokinetic behaviors, a particular dose (in milligrams per kilogram) will give a larger area under the tissue curve for parent chemical in a larger species than in a smaller species. This occurs because the original dose is dispersed into a volume proportional to body weight, and animals, regardless of body weight, should attain the same initial internal concentration for a given dose (in milligrams per kilogram). Clearance, however, is expected to be proportionately larger in the smaller species because it increases as a fractional power of body weight, and therefore, the area under the blood and tissue curves will be smaller in the smaller animal species. This simplified PK analysis indicates that larger species are at greater risk (i.e., have larger area under the tissue curve) than are smaller animals for those instances where parent chemical is the DNA-reactive chemical. This behavior is consistent with the traditional surface area correction used for interspecies dose adjustments in the chemical risk assessments conducted by the Environmental Protection Agency (EPA). The use of this factor is usually justified by reference to studies on the interspecies differences in acute toxicity of a variety of chemicals used medically in cancer chemotherapy (Freireich et al., 1966).

Stable Metabolites

Ethylene oxide can also be produced in vivo by the oxidation of ethylene by microsomal metabolism. In developing a pharmacokinetic model for ethylene oxide as a DNA-reactive, stable metabolite, other PK modeling considerations become important. These include the rate of formation of the epoxide in various tissues, the stoichiometric yield of the epoxide from ethylene in vivo, and the distribution of the stable metabolite to target tissues. With butadiene, there are two epoxide metabolites that have genotoxic potential, and some provision for their differential DNA reactivities might have to be included in model development. A very elegant analysis of the relative risks of ethylene and ethylene oxide has been conducted by Bolt and Filser (in press). They developed pharmacokinetic models for both of these chemicals and attempted to predict ethylene exposure conditions that would yield carcinogenic tissue doses of the epoxide. They did not use a physiological model, and their results are limited to interpretation of the bioassay results in experimental animals. Nonetheless, their study is an excellent example of the use of sound pharmacokinetic principles in the analysis, interpretation, and design of toxicity experiments.

In instances in which the genotoxic chemical is a stable, freely circulating metabolite, the analysis of the effect of body size on tissue dose includes consideration of the metabolic formation of a DNA-reactive metabolite and its consumption by various clearance pathways. For the purposes of risk assessment, when there is no available information on the human population, it seems appropriate to assume that both the metabolic production and the clearance pathways are related to the same fractional power of body weight (National Research Council, 1986). Thus, equivalent doses on a body weight basis are expected to produce approximately equal tissue exposures expressed as area under the tissue curve of the genotoxic, stable metabolite. This suggests that larger animals should be at the same risk from equivalent doses of these chemicals as smaller animals. For this class of chemicals, the standard surface area correction used by EPA would overestimate the expected risk in humans based on extrapolation of toxicity results in small laboratory animals. The Food and Drug Administration approach which uses body weight for interspecies dose conversion is the more appropriate correction factor for this class of chemical carcinogens.

Reactive, Nonisolatable Metabolites

In a recent paper we attempted to develop a strategy for conducting a pharmacokinetically based risk assessment for methylene chloride (Andersen et al., 1987). On the basis of a variety of kinetic and chemical arguments, we suggested that the carcinogenicity of methylene chloride was related to the metabolites produced by conjugation of parent chemical with glutathione. If this proposed mechanism of toxicity is correct, the appropriate measure of tissue dose should be the time integral of the tissue concentration of the glutathione conjugate. This material is too reactive to measure directly, and a surrogate measure of tissue concentration of this chemical must be utilized in place of its concentration. The surrogate dose metric that was developed based on kinetic principles was a ratio of the integral of the amount of chemical metabolized by this pathway in the target tissue divided by target tissue volume. This same approach could be used with other chemicals, like vinyl chloride, where the presumed genotoxic metabolite is also too short-lived to measure directly.

When reactive metabolites are associated with carcinogenicity, the simplified pharmacokinetic analysis of the effect of animal size on integrated tissue exposure suggests that larger species will be at proportionately less risk than smaller species (National Research Council, 1986). The reason for this dependence is that metabolic production (the numerator) is proportional to a fractional power of body weight, while tissue volume (the denominator) is directly proportional to body weight. The ratio of the two then decreases with increasing body weight. This approach to interspecies scaling for vinyl chloride was previously suggested by Gehring et al. (1978) based on somewhat different arguments.

The above examples point out that it is very difficult to depend on a single approach to interspecies scaling. When scaling strategies are developed based on generalized pharmacokinetic principles, there are several very different types of interspecies scaling behaviors depending on the nature of the DNA-reactive chemical—whether it is parent chemical, stable metabolite, or a highly reactive, nonisolatable metabolite. These differences should in some way be reflected in the process of hazard assessment when the mechanism of carcinogenicity of a chemical is fairly well-established. Universal reliance on the surface area correction, or any one particular adjustment factor, should be avoided; however, in the absence of information on the mechanism of toxicity, the surface area correction would at least err on the conservative side.

Intercalating Agents

Another group of direct-acting, DNA-interactive chemicals are the intercalating agents, represented by acridine-type dyes and related chemicals (Rogers and Back, 1982). With these materials there is noncovalent bonding between the dye and DNA, and the interactions are probably best described by the mass action law with critical receptor site occupancy by the intercalated dye. For this type of tissue interaction we would need to know the dissociation constant(s) and binding capacity for the agent-DNA binding processes, and the time course of intercalator concentration in the target tissues. Tissue dose in this case is probably best represented as a time-weighted average receptor occupancy by the intercalating ligand. Thus, even for genotoxic chemicals there are possibilities that interactions can occur either by chemical reactivity or by mass action effects of particular chemical ligands. These two mechanisms lead to two very different expressions for tissue dose.

These estimates of tissue exposure with chemically reactive or intercalating agents can be used as the dose inputs to drive increased mutational frequency in biologically based cancer models such as that proposed by Moolgavkar and Knudson (1981). Combining pharmacokinetic and pharmacodynamic modeling of the cancer process (Figure 6) promises to greatly improve our ability to conduct interspecies scaling and support risk assessment extrapolations. It is important to remember, however, that tissue dose will often be nonlinear with respect to administered dose, and it is clearly wrong to use administered dose uncritically in developing realistic cancer models. To a very great extent, it is only the availability of accurate pharmacokinetic descriptions of tissue exposure which permits validation of biologically motivated models of chemical carcinogenesis. In fact, it is essential to have an adequate understanding of the pharmacokinetic characteristics of target tissue exposure before pharmacodynamic models are developed for any kind of toxic response.

Figure 6

A schematic of a two-stage cancer model: Normal cells (1) have a basal birth rate (B1) and death rate (D1). There is also a background mutational rate (P1). The probability of undergoing a mutational event is the product of the birth rate and the mutational (more...)

The mechanisms of carcinogenicity of directly acting genotoxic chemicals are still under active investigation in terms of fundamental questions about the nature of DNA adducts formed, rates of repair of damaged DNA, the presence of critical mutational sites on DNA, etc. Eventually, as more information is developed, it may even be possible to use the formation of particular adducts as the measure of tissue dose instead of the use of integrated tissue exposure. This would give metrics for tissue doses of carcinogens which were closer to the biological process of tumor induction.

Epigenetic Carcinogens

Despite the many outstanding questions with regard to the mode of action of genotoxic carcinogens and to the relative importance of particular DNA adducts, it is clear that a great deal more is known about the mechanisms of tumor initiation with these chemicals than about the detailed mechanisms by which epigenetic carcinogens cause tumor development. In general, there seem to be two very different groups of epigenetic carcinogens. The first group consists of those chemicals that cause overt cytotoxicity and cancer appears secondary to chronic tissue damage. Chloroform and carbon tetrachloride are examples from this group (Reitz et al., 1982). The second group consists of the tumor promoters which interact with the cell in such a way to induce expression of new, characteristic sets of enzyme activities. The altered cellular environment caused by these promoters then leads to enhanced tumor yield under appropriate exposure conditions. Examples here include phenobarbital and dioxin. Tissue dosimetry for these epigenetic carcinogens will be more complex than it is for genotoxic carcinogens.

For epigenetic mechanisms, it will be necessary to include some kind of pharmacodynamic modeling along with the pharmacokinetic description. For cytotoxicity (Figure 6) it will be necessary to model the linkage between tissue reactivity of reactive chemical with depletion of critical cellular macromolecules, cell death, and attendant hyperplasia. This new birth rate function can then be used in a biologically motivated cancer model (see R. B. Conolly, R. H. Reitz, and M. E. Andersen, this volume). With promoters the proper tissue dose metric will be related to receptor occupancy and a resulting induction of new protein synthesis. The pharmacokinetic modeling strategy ultimately devised for these promoters will require physiologically accurate models for the processes involved in enzyme induction. In the two-stage cancer model, the action of these promoters is considered to be on the birth and death rates of the stage 2 cell, providing cells with a single mutation with a selective growth advantage. While dosimetry with these epigenetic carcinogens is more demanding than with genetic carcinogens, there still appears to be two dose metrics that emerge—integrated exposure to reactive chemical for cytotoxicants and time-weighted average receptor occupancy for chemicals such as dioxin and phenobarbital. These seem to be the primary expressions required for understanding our problem here: that is, just what is tissue dose? On the other hand, what can be said of pharmacokinetics in these cases. The modeling strategy is still the same—produce an integrated description of chemical disposition and tissue exposure which is readily amenable to interspecies extrapolation and in which all biochemical, physical chemical, biological, and physiological processes are as clearly defined as possible.

Summary

Mechanistic information on the processes involved in cancer causation by a particular chemical is essential for defining the appropriate measure of target tissue dose. Tissue dose will usually either be a function of integrated tissue exposure or a function of the extent of receptor binding in a tissue. This latter metric of tissue dose is dependent on mass action principles and not simply on integrated tissue exposure. Measures of dose related simply to administered dose or total amount metabolized should be viewed with great caution unless there are compelling reasons for believing there is a direct correlation of these very coarse measures of dose with actual tissue exposure. Once the proper measure of dose is defined, a pharmacokinetic model should be developed to predict this dose metric for various exposure scenarios in a variety of species. The biological realism of physiologically based models confers on them certain advantages for use in the risk assessment arena. Finally, much better cancer risk assessments will be possible when validated pharmacokinetic models for tissue dose are used in conjunction with more biologically, realistic pharmacodynamic descriptions of the biological processes involved in chemical carcinogenesis.

References

• Andersen, M. E., H. J. Clewell III, M. L. Gargas, F. A. Smith, and R. H. Reitz. 1987. Physiologically-based pharmacokinetics and the risk assessment process for methylene chloride. Toxicol. Appl. Pharmacol. 87:185-205. [PubMed: 3824380]
• Bischoff, K. B., and R. G. Brown. 1966. Drug distribution in mammals. Chem. Eng. Prog. Symp. Ser. 62:33-45.
• Bolt, H. M., and J. G. Filser. In press. Kinetics and disposition in toxicology. Example: Carcinogenic risk assessment for ethylene. Proceedings of the International Conference on Contributions of Toxicology to Risk Assessment and Health Regulation. Arch. Toxicol.
• Dedrick, R. L. 1973. Animal scale-up. J. Pharmacokinet. Biopharm. 1:435-461. [PubMed: 4787619]
• Freireich, E. J., E. A. Gehan, D. P. Rall, L. H. Schmidt, and H. E. Skipper. 1966. Quantitative comparison of toxicity of anticancer agents in mouse, rat, hamster, dog, monkey and man. Cancer Chemother. Res. 66:55-68. [PubMed: 4957125]
• Gargas, M. L., H. J. Clewell III, and M. E. Andersen. 1986. Metabolism of inhaled dihalomethanes in vivo: Differentiation of kinetic constants for two independent pathways. Toxicol. Appl. Pharmacol. 82:211-223. [PubMed: 3945949]
• Gehring, P. J., P. G. Watanabe, and C. N. Park. 1978. Resolution of dose-response toxicity data for chemicals requiring metabolic activation: Example—vinyl chloride. Toxicol. Appl. Pharmacol. 44:581-591. [PubMed: 567390]
• Gerlowski, L. E., and R. K. Jain. 1983. Physiologically-based pharmacokinetic modeling: Principles and applications. J. Pharm. Sci. 72:1103-1126. [PubMed: 6358460]
• Goldstein, A., L. Aronow, and S. M. Kalman. 1974. Molecular mechanisms of drug action. Chapter 1 in Principles of Drug Action: The Basis of Pharmacology, 2nd ed. New York: John Wiley & Sons.
• Kociba, R. J., and B. S. Schwetz. 1982. Toxicity of 2,3,7,8-tetrachlorodibenzo-p-dioxin (TCDD). Drug Metab. Rev. 13:387-406. [PubMed: 6213397]
• Maltoni, C., and G. Lefemine. 1974. Carcinogenicity bioassays of vinyl chloride. I. Research plan and early results. Environ. Res. 7:387-405.
• Moolgavkar, S. H., and A. G. Knudson, Jr. 1981. Mutation and cancer: A model for human carcinogenesis. J. Natl. Cancer Inst. 66:1037-1052. [PubMed: 6941039]
• National Research Council. 1986. Dose-route extrapolations: Using inhalation toxicity data to set drinking water limits. Pp. 168-225 in Drinking Water and Health, Vol. 6. Washington, D.C.: National Academy Press.
• Poland, A., and J. C. Knutson. 1982. 2,3,7,8-Tetrachlorodibenzodioxin and related halogenated aromatic hydrocarbons: Examination of the mechanism of toxicity. Annu. Rev. Pharmacol. Toxicol. 22:517-554. [PubMed: 6282188]
• Ramsey, J. C., and M. E. Andersen. 1984. A physiologically based description of the inhalation pharmacokinetics of styrene in rats and humans. Toxicol. Appl. Pharmacol. 73:159-175. [PubMed: 6710512]
• Reitz, R. H., T. R. Fox, and J. F. Quast. 1982. Mechanistic considerations for carcinogenic risk estimations: Chloroform. Environ. Health Perspect. 46:163-168. [PMC free article: PMC1569040] [PubMed: 7151758]
• Weisburger, J. H., and G. M. Williams. 1980. Chemical carcinogens. Pp. 84-138 in J. Doull, editor; , C. D. Klaasen, editor; , and M. O. Amdur, editor. , Eds. Casarett and Doull's Toxicology: The Basic Science of Poisons, 2nd ed. New York: Macmillan.