National Academies Press: OpenBook

Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment (1987)

Chapter: VI. Applications of Mathematical Modeling

« Previous: V. Poster Session
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 351
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 352
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 353
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 354
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 355
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 356
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 357
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 358
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 359
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 360
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 361
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 362
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 363
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 364
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 365
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 366
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 367
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 368
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 369
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 370
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 371
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 372
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 373
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 374
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 375
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 376
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 377
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 378
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 379
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 380
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 381
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 382
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 383
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 384
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 385
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 386
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 387
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 388
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 389
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 390
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 391
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 392
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 393
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 394
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 395
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 396
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 397
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 398
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 399
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 400
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 401
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 402
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 403
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 404
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 405
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 406
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 407
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 408
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 409
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 410
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 411
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 412
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 413
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 414
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 415
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 416
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 417
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 418
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 419
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 420
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 421
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 422
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 423
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 424
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 425
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 426
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 427
Suggested Citation:"VI. Applications of Mathematical Modeling." National Research Council. 1987. Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment. Washington, DC: The National Academies Press. doi: 10.17226/1015.
×
Page 428

Below is the uncorrected machine-read text of this chapter, intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text of each book. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

PART Vl Applications of Mathematical Modeling

Hazard Assessment Using an Integrated Physiologically Based Dosimetry Modeling Appproach: Ozone Frederick ]. Miller, John H. Overton, Jr., Elaine D. Smolko, Richard C. Graham, and Daniel B. Menze! INTRODUCTION In examining the possible role of pharmacokinetics and pharmacody- namics in risk assessment, the underlying philosophy of the legislative mandate through which the risk assessment is applied must be kept in mind. For example, the ability of the Food and Drug Administration to invoke the Delaney Clause to regulate a substance is different from the risk assessments required for the Environmental Protection Agency's Na- tional Ambient Air Quality Standards (NAAQSs). The kind and level of information that is available can vary greatly. Reevaluation of the NAAQSs for ozone (03) focuses on whether or not the value of the standard should be changed by as little as 20%, while many carcinogenic risk assessments are trying to establish the level of risk to within one to two orders of magnitude. In any case, the uncertainties identified in risk assessments help to establish areas in which additional research would be useful. The intent of this paper is to broaden the awareness that pharmacoki- netics and mathematical dosimetry models are useful tools in risk assess- ments of noncarcinogenic as well as carcinogenic effects. While examples The research described in this paper has been reviewed by the Health Effects Research Lab- oratory, U.S. Environmental Protection Agency, and approved for publication. Approval does not signify that the contents necessarily reflect the views and policies of the Agency nor does mention of trade names or commercial products constitute endorsement or recommendation for use. 353

354 FREDERICK J. MILLER et al. Of modeling applications have been predominantly related to carcinogen- esis, health effects such as emphysema and fibrosis are not to be ignored. This is particularly true for oxidant gases. For example, several animal studies in different species show that long-term exposure to nitrogen diox- ide produces emphysema (renters et al., 1973; Hyde et al., 1978; Riddick et al., 1968), and O3 causes pulmonary fibrosis (Fujinaka et al., 19851. Moreover, subchronic exposures at relatively low levels of these gases have been shown to lead to cellular changes (Barry et al., 1985; Chang et al., 1986) that are indicative of a structural remodeling of the lung. The impetus is clear that man must be protected from such effects. But to use the animal toxicological data more quantitatively in setting appro- priate NAAQSs for these gases, extrapolation modeling is required. In this volume the need to incorporate pharmacodynamics in the mod- eling process has been discussed. Thus far, most risk assessments have assumed a priori that an equivalency of response exists between animals and man. As one proceeds from the molecular level or biochemical event toward injury at the tissue or organ level, that assumption becomes less tenable because of possible species differences in repair processes, levels of antioxidant enzymes, etc. The dose, if it is sufficiently high, can produce damage anywhere from the molecular to the organ level. On the other hand, various host defense systems can interact, and if the damage is not sufficiently severe, they can yield recovery. If the delivered dose over- whelms these defense systems, various disease states can result. Defense systems might conceivably play a role in the etiology of lung disease. All of this is a dynamic situation. Figure 1 (based on Figure 1-1 of NRC, 1983) summarizes an overview of the research, risk assessment, and risk management processes and their interrelationships that lead to the recommendation of a NAAQS. This paper focuses on the phases of the processes outlined by the dashed rectangle. Here, the area of pharmacokinetics offers great potential for improved risk assessments, particularly when these assessments are re- quired to be more quantitative in nature. The development of dosimetry models that can provide a description of the uptake and distribution of chemical compounds throughout the body and the availability of toxico- logical data that can be used to establish dose-response relationships are integral to these efforts. The incorporation of laboratory and field obser- vations of adverse health effects, hazard information, and extrapolation methods to yield dose-response assessments is critical to the risk assess- ment process. The combining of these phases can be facilitated by the integrated physiologically based dosimetry modeling approach, illustrated schematically in Figure 2. The major components of the approach are the critical toxicity reference (CTR) system (Smolko et al., in press) and the physiologically based

Establishing Dose-Response Relationships 355 RESEARCH RISK ASSESSMENT RISK MANAGEMENT l Laboratory and field observations of adverse health effects and exposures to particular agents Information on extrapolation methods for high to low dose and animal to human Haza rd Identifi. ation Dose-Response Assessment Field measurements, esti mated expose res, characterization of populations _ _ _ J Exposu re Assessment _ Risk Characterization \ + \ Development of regulatory options Evaluation of public health, economic, social, political consequences of regulatory options T | Agency l decisions and actions FIGURE 1 Major elements of risk assessment and risk management. The phases of the process that can be addressed by using an integrated physiologically based dosimetry modeling approach are contained within the dashed rectangle. Overall figure schematic is based on Figure 1-1 of NRC (1983). dosimetry (PBD) model. The CTR system is comprised of four elements: (1) searching the literature for references reporting toxicity data relevant to the toxicant being studied; (2) abstracting information from the selected references in a form appropriate for both quantitative and descriptive extraction; (3) constructing a data base that consists of bibliographic and abstracted information; and (4) compiling concentration-response data by a series of searches of the developed data base. The primary factors considered in developing PBD models are mammalian physiological pro- cesses, anatomical characteristics, physicochemical properties of the gas and of relevant biochemical constituents, and mass transport processes. PBD model predictions of dose, appropriate for species with the specific characteristics described in the literature, are combined with the concen-

356 FREDERICK J. MILLER et al. ll L _ _ _ _ _ _ _ Critical Toxicity Reference (CTR) System Search Literature Abstract Information Construct Data Base Compile Concentration- Response Data Exposure Concentration Time of Exposure - parameters specific to 9 iven I animal, weight, and sex Data Base animal, for weight, ~ anatomical, physiological, sex biochemical, and physical parameters 1 1 LO Physiolog ical Iy-Based Dosi metry (PBD) Model T predicted dose response | Dose-Response Relationship | Species Sensitivity Host Defense Repair Processes Genetics Identifv Research Gaps . _ .. .., .. _ . and Design Experiments | Estimate Hu j an Toxicity ~ r | Risk Characterization 1 FIGURE 2 A schematic of the elements of an integrated physiologically based dosimetry modeling approach to estimating human toxicity that leads to one of the components of risk characterization. The procedure diagrammed in the dashed rectangle is the proposed approach for implementing the dashed rectangle process shown in Figure 1. tration-response data that have been collated by species and endpoint. The resulting dose-response relationships can use various expressions of dose to obtain the most appropriate quantitative representation of toxicological effects. The analyses can incorporate species sensitivity information for con- sideration of such factors as host defense, repair processes, and genetics. One outcome of this approach is an estimation of human toxicity that can be used as input into the risk assessment and risk management processes.

Establishing Dose-Response Relationships 357 A second outcome is information to aid in the identification of data gaps and design of experiments. This latter outcome is particularly important because it can eventually yield new data that will be used in the iterative application of the integrated PBD model approach which will strengthen the overall risk assessment and risk management processes. The remaining sections of this paper illustrate the integrated PBD mod- eling approach (Figure 2) by applying data from the CTR system and predictions from a PBD model for O3 to construct dose-response rela- tionships. This will be done first by discussing the formulation, as well as sample simulation results, of a PBD model that predicts the uptake and distribution of absorbed O3 in the lungs of mammals. Although the focus will be on 03, the intent is primarily to illustrate the methodology and to provide an understanding of various aspects of delivered dose. The meth- ods discussed will be particularly important when some of these procedures are extended to volatile organic compounds, for which first-pass metabolic effects in the lung are a concern. Also, a brief discussion is provided of a data base management system, CTR, that has been developed to store and retrieve quantitative data in a manner useful for mathematical dosi- metry models. An application to ozone toxicological data is presented to illustrate the methodology of using mathematical dosimetry models to examine quantitative dose-response relationships. LOWER RESPIRATORY TRACT MATHEMATICAL DOSIMETRY MODELI NG In the papers presented in this volume, much attention is placed on the use of physiologically based pharmacokinetic models to provide a de- scription of dose distribution following inhalation of a chemical com- pound. To date, these models predict average levels of the chemical throughout an entire organ (or body compartment). As illustrated below, however, additional complexities are involved when the distribution of inhaled gases is evaluated within the lower respiratory tract (LRT), where dose in various lung regions can vary greatly and consequently have different health effect outcomes. Mode! Conceptualization The major factors affecting the regional uptake of O3 are morphology of the respiratory tract, the route of breathing, the depth and rate of breathing, gaseous physiocochemical properties, the physical processes governing gas transport, and the physiocochemical properties of the trach- eobronchial (TB) liquid lining and of the air-blood barrier in the pulmonary

35~3 FREDERICK J. MILLER et al. region. All these factors interact in a complex way to determine dose and must be considered in developing a simulation model. The mathematical model formulation will focus on the lower respiratory tract for several reasons. Upper respiratory tract removal of inhaled gases and particles is amenable to experimental determination of deposition (Corn et al., 1976; Miller et al., 1979; Yokoyama, 1968; Yokoyama and Frank, 19721. The morphology of the upper respiratory tract is quite complex, difficult to measure, and variable between species (Schreider, 1986; Schreider and Raabe, 1981), so mathematical descriptions are dif- ficult to obtain. Further, airflow patterns are also complex (Patra et al., 1986), so proper treatment of gas transport processes is not apparent. Experimental values for upper respiratory tract uptake of the inhaled gas, however, can provide appropriate boundary conditions for mathematically modeling delivered doses to LRT regions. Aspects of LRT structure and their relationship to model compartments are illustrated in Figure 3. Briefly, anatomical descriptions of the lung are needed such as those available for man (e.g., Weibel, 1963; Yeh and Schum, 1980) and for various animals (e.g., Kliment, 1973; Schreider and Hutchens, 1980; Yeh et al. 19791. The top portion of Figure 3 shows Weibel's (1963) representation of the TB and pulmonary regions of the human lung. For mathematical modeling purposes, we conceptualize this representation into a series of right circular cylinders; and information is needed on the lengths, diameters, and radii of the TB airways. In the pulmonary region, data on the structure of the pulmonary ducts and sacs and on the number of alveoli, their surface areas, and volumes are nec- essary to apply the model description of gas transport processes. In the TB or conducting airways, a mucociliary layer protects the un- derlying ciliated, goblet, brush, and basal cells, etc., from direct insult by the inhaled gas. As can be seen in Figure 3, this layer consists of an epiphase and a hypophase. Lacking definitive data on the thickness and chemical composition of these phases in various portions of the TB region, the current model formulation combines them into one compartment. For ambient exposures to the highly reactive gases, such as 03, penetration to the bloodstream in the TB region can be ignored (see Miller et al., 1985, for details). Thus, in the TB region, model compartments correspond to the air, liquid lining layer, and underlying tissue. For highly soluble and nonreactive gases, however, a blood compartment and a description of the fate of the gas in other organs and compartments of the body probably would be needed. In the pulmonary region, the epithelium is chiefly comprised of type I and type II cells over which a very thin layer of a surfactant fluid can be found. Underlying the epithelium are the interstitium and the endothelial cells of the capillary network. Because the alveoli are arranged back to

Establishing Dose-Response Relationships 359 - TB ~ 14131 2 T BL BR B L ~ " PU LM ONARY t22|21|20|19|18|17 AS AD RBL ~ W~ s~ ~ W/~: O TRA- CHEA l- , a . . b ~ ,~ c rat I · _ 1 . A C A`' _ ,, A d AIR TISSUE\ a ~ r ~ / ~ TISSUE ~ ;/ l l LIQUID LINING, l l AIR , , _____________________ ________ ~ , ______________ A,. AIR _ _ _' . HY~ An-- ' "'AS' UQUlD uea~c Use. LIQUID LINING FIGURE 3 The relationships between morphologies and their model representations. Panel a is a schematic of the branching airways of the TB and pulmonary region of man (based on Weibel, 1963). The generations are labeled from the right, beginning with the trachea. BR, BL, and TBL indicate bronchi, bronchioles, and terminal bronchioles, respectively; the respiratory bronchioles, alveolar ducts, and alveolar sacs are indicated by RBL, AD, and AS, respectively. Below the TB portion of the lung schematic are three cylindrical figures that indicate the model representation of the airways. Panel c shows a diagram of the structure of the liquid lining and tissue of the TB region (diagram based on Jeffrey and Reid, 1979). The different cells represented are basal (BC), ciliated (CC), brush (BrC), goblet (GC), and conciliated serous (NCC). Panel e illustrates the model representation of TB liquid lining and tissue compartments. Panel b is an electron micrograph of the interalveolar septa (based on Gehr et al., 1978). The air spaces (A), capillaries (C), type I cells and their nuclei (EP1 and NEP1), endothelial cells and their nuclei (EN and NEN), and interstitial space (IN) are indicated. Panel d illustrates the model represen- tation of the liquid lining, tissue, and capillaries of the pulmonary region.

360 FREDERICK J. MILLER et al. back, with capillaries between them, the model formulation for the air, liquid lining, tissue, and blood compartments is assumed to be symmetrical (Figure 31. A one-dimensional equation of mass transport is used to describe the dynamics of the cross-sectional average concentration. Experimental ev- idence for this approach is discussed by Overton (19871. The processes of axial convection and axial dispersion, the loss of O3 to the liquid lining, and lung expansion and contraction are taken into account. Lung expansion and contraction during the breathing cycle become important when pol- lutant uptake for minute ventilations corresponding to heavy exertion is modeled. When mass transfer in the liquid lining, tissue, and blood com- partments is modeled, the transport of ozone is related to molecular dif- fusion and chemical reaction terms. The reader is referred to Figure 2 of Overton et al. (J. H. Overton, R. C. Graham, and F. J. Miller, this volume) for explicit forms of the partial differential equations used in the model. For a detailed discussion of the model formulation and assumptions, as well as the basis for these assumptions, see Miller et al. (1985) and Overton et al. (19871. Note that when other gases are modeled, modifications of the expression for loss of the gas to airway walls may be necessary. In addition, high solubility or nonreactivity may necessitate changes in some aspects of treating gas transport in the airways. The information required to model adequately transport in the mucus or surfactant lining liquid, tissue, and blood compartments includes quan- titative biochemical data on the constituents as well as on the chemical reactions of O3 with these constituents. Currently, a pseudo-first-order reaction scheme is assumed when ozone uptake is modeled (see Miller et al., 19851. Estimates of liquid lining, tissue, and blood compartment thickness are also required. At present, this is an area in which most of the available information is from animal studies. Data for man are very limited, and assumptions about the distribution of compartment thickness must be made. Compartmental diffusion coefficients and partitition coef- ficients, such as those used in physiologically based pharmacokinetic models, are also needed. Much of the above topic is discussed in detail elsewhere (Miller et al., 1985; Overton et al., 19871. Examples of O3 Dosimetry Modeling Results Utilizing the model formulation concepts discussed above, Miller et al. (1985) examined the uptake of O3 in the LRT of man and performed various sensitivity analyses to illustrate the importance of LRT secretions on delivered dose. Their work shows that the net amount of O3 removed in the trachea can be several orders of magnitude different from the amount of O3 that penetrates to the underlying tissue. Proceeding distally from

Establishing Dose-Response Relationships 361 the trachea, this difference diminishes mainly because of the decline in the thickness of the liquid lining. Furthermore, because the gas-exchange region is lined with a very thin fluid that does not contain many constituents that react with 03, the net amount of O3 removed and the tissue dose of O3 are essentially the same. According to Miller et al. (1985), the net O3 dose curves are much less dependent on the thickness of the TB liquid lining than are the O3 tissue dose curves. Another example of sensitivity analysis demonstrating the importance of model parameters is concerned with the values of the liquid lining rate constants. Simulations show that the effect of increased chemical reactivity in the TB liquid lining is to increase the net airway doses in the TB region and to decrease the tissue dose throughout the LRT. Furthermore, tissue dose is found to be much more sensitive to the TB liquid lining reactions than is the net dose. For example, the tracheal tissue dose is predicted to decrease by about 3 orders of magnitude, and the tracheal net dose is predicted to increase by about 1 order of magnitude because of a rate constant change from 0 to twice the reference value. However, pulmonary net and tissue dose values are predicted to change by less than a factor of 2 as a result of the same change in the TB liquid lining rate constant. By contrast, changes in the pulmonary liquid lining rate constant lead to simulation results for LRT net and tissue doses that are essentially un- changed from those obtained with the reference values. These types of analyses show that to determine dose-effect relationships for various cell types or components (e.g., cilia) in the conducting airways, better quan- titative data than are presently available on TB liquid lining rate constants and thickness are desirable. On the other hand, the lower sensitivity of predicted pulmonary doses to parameter uncertainties gives a greater con- fidence in making interspecies comparisons of dose-effect relationships in this region compared with making such comparisons in the TB region. An example of interspecies dosimetric comparisons is illustrated in Figure 4, in which the solid and dashed lines are O3 dosimetry simulation data for the rats and man, respectively. Predictions for the rat lung use the Yeh et al. (1979) anatomical model in which the morphometric data are represented in a generational-type model analogous to the Weibel (1963, generational morphometric data for the human lung. Both the net and tissue doses of O3 are shown for these species. Dose is expressed in Figure 4 as micrograms of O3 per square centimeter of airway surface area per minute, standardized to a tracheal O3 value of 1 ~g/m3. In the mathematical dosimetry model the processes are linear. Thus, dose is proportional to the tracheal concentration. For example, for a tracheal concentration of 1,000 ~g/m3, the predicted airway doses are 1,000 times the plotted values; for a tracheal concentration of 1 ppm, multiply the plotted values by 1,960 (~g/m3 per ppm).

362 FREDERICK J. MILLER et al. 10-5 -6_ o ~5 10-7 l - aJ . _ 1 ~ 10-8_ v _' o o 10-9 1 o NET I\ i / 2 4 6 2 4 6 TISSU E it\ VT(mt) - RAT 1 .84 - HUMAN 500. l 8 10 12 14 16 8 10 11 13 15 GENERATIONS \ f(BPM) \ 105 \ 15 .,,, _ 1 8 20 22 24 H U MAN 17 19 21 23 RAT FIGURE 4 An example of a way of comparing species doses by plots of airway dose versus generation for two species, man (dashed curves) and rats (solid curves). The first generations of the pulmonary regions (17 and 16 in the anatomical models for man and rat, respectively) have been matched at the dotted vertical line to allow a better comparison of equivalent regions; this results in the break in the rat generations along the abscissa. The tidal volume (VT) and breathing frequency (f) are given. Tissue dose is the predicted dose of the epithelial layer in the tracheobronchial region and of the air-blood barrier in the pulmonary region. Net dose is the total absorbed by the liquid lining, tissue, and blood compartments.

Establishing Dose-Response Relationships 363 As can be seen from Figure 4, the net removal of O3 iS very comparable in the rat and human lungs. To help examine tissue dose patterns, we have matched the first generation of respiratory bronchioles in man with the first alveolated airway in the rat lung. This region is where the his- topathological data show the first and the most pronounced lesions in O3 toxicological studies with rats (Barry et al., 1985; Boorman et al., 1980; Flopper at al., 1979; Stephens et al., 19731. The maximal dose in both man and rat is predicted to occur at this particular region. Almost a 10- fold increase in deposition or removal of O3 iS predicted in this region in the rat compared with that in man according to the dose curves shown in Figure 4, which are standardized to a unit tracheal O3 value. To make a more definitive interspecies dosimetric comparison at this point, however, nasopharyngeal removal of O3 in man versus rat must be considered. Nasopharyngeal removal studies are currently being conducted, and when available, they will facilitate translation of the net and tissue dose curves of Figure 4 into delivered dose curves associated with exposure to O3. CRITICAL TOXICITY REFERENCE SYSTEM Data collection for the O3 analyses was accomplished by the CTR system (Smolko et al., in press). The four basic elements of this system have been briefly described. Sources of references for inclusion in the CTR data base include searches of nationally available data bases (e.g., National Library of Medicine), comprehensive documents, personal communication with colleagues working with the toxicant being studied, and personal review of the relevant literature. Upon establishing an initial base of information from a thorough literature search, the same sources are used to update periodically the CTR data base. A coding strategy was developed to provide a mechanism for retrieving the specific data that are required for dose-response analyses. All information is abstracted onto a coding sheet according to a predetermined format. Abstractors are intensively trained and assist in all aspects of this phase upon completion of their training. The information abstracted onto the coding sheet is entered into the computerized data base. Each reference comprises a unique file and is assigned a unique identification number. A critical element for dose-response analyses is a quantitative repre- sentation of the effect observed and reported in the literature. The last phase of the CTR system, then, compiles the concentration-response data in a form useful for combination with the POD model predictions. Sub- data bases are generated for easy access to subsets of information or to facilitate future searches for very specific information. For example, a sub-data base could be established for all references reporting concentra- tion-response data. This could subsequently be searched to sort by animal

364 FREDERICK J. MILLER et al. species. Further searches of these sub-data bases could identify studies of particular biological endpoints at specific pollutant concentrations. After the concentration-response data are compiled, individual studies are examined for inclusion in the dose-response analyses. The basic criteria applied include a quantitative representation of effect, the use of a species appropriate for combination with the PBD model for 03, an adequate presentation of data, a statistical analysis of data, exposures to O3 only, a lack of variables that might confound results, and continuous exposures to a single concentration of O3. Other items to be considered when selecting studies concern both interpretive and technical quality. Interpretive quality criteria address items such as the use of an appropriate experiment to test the hypothesis, while technical quality cr~tena address items such as pro- viding precise descriptions of experiments and data analysis. When specific studies appropriate for dose-response analyses are iden- tified, information regarding, for example, weight of the species used is extracted. These data are used with the PBD model to predict the dose of O3 that reaches specific regions of the lung. Figure 5 shows PBD model 3.5 3 a' ~ 2.5 - 2 Q 15 a' o 1 red o 5 a, . _ a, O ~ 1 1 1 1 1 1 1 1 1 ~ 1 1 1 1 1 1 1 1 1 _< . a - ~ .dS' . ~ 5 ~ 1 1 1 1 1 1 1 1 1 1 1 1 1 00 150 200 250 300 3 50 400 450 500 550 Rat Weight (grams) FIGURE 5 Predicted quantity of absorbed O3 per breath versus rat weight for a 1-ppm exposure concentration with 25% upper respiratory tract removal. The three curves are for total lower respiratory tract (x), tracheobronchial region tissue (O), and pulmonary region tissue (/\) absorption. Predictions were made using the PBD model described in the text. Lower respiratory tract anatomical dimensions were based on Yeh et al. ( 1979) for a 330-g rat. Allometric equations were used to extrapolate the lung dimensions to those corresponding to other weights.

Establishing Dose-Response Relationships 365 predictions of dose to the mucus-lined tissue, the surfactant-lined tissue, and the total lung for rats (using the anatomical model of Yeh et al., 1979) exposed to 1,960 ~g/m3 (1 ppm) O3. Combining dosimetry data, such as those contained in Figure 5, with biological effects data to yield dose-response curves is illustrated using the formation of edema as an endpoint. Experimental studies (Ichikawa and Yokoyama, 1982, Maitani and Suzuki, 1981; Nambu and Yokoyama, 1981; Suzuki, 1976) were selected in which edema formation was indicated by an increase in lung wet weight. Because nasopharyngeal removal of O3 in the rat has not been determined experimentally, one analysis was conducted in which 10% removal of O3 in the head was assumed and another in which 25% removal was assumed. Unconfirmed preliminary experiments (Drs. Jean Wiester and Gary Hatch, personal communication) indicate that the rat may remove 10-25% of the inhaled O3 in the naso- pharyngeal region. Because edema is generally considered to reflect injury to gas-exchange regions, pulmonary tissue dose was used in the dose- response analysis (Figure 61. Data points represent various combinations of exposures to O3 concentrations, ranging from 1 to 5 ppm for durations ranging from 3 to 12 h. Although a linear fit to the dose-response data is depicted in Figure 6, other curves may be equally plausible for describing the dependence of increased lung wet weight on the dose of O3 delivered to the pulmonary region of the rat. The next step in the analysis would be to determine, using the PBD model, which O3 exposure profiles would result in comparable predicted doses of O3 in the human lung. This would allow better judgments on the reasonableness of extrapolating to man the toxicological data on edema in rats following exposure to O3. SUMMARY Determining tissue dose is a fundamental starting point when making interspecies dose-response comparisons. Because of the complexity and cellular diversity of the lung, toxicologically relevant lung dosimetry needs to be related to specific sites within the respiratory tract to achieve a better understanding of the tissue dose dependence on toxicity of inhaled pol- lutants. Direct measurement of tissue or cellular dose by radiometric, physical, or chemical means within specific segments of the lung is one approach to determining equivalent doses between species. Such experi- ments are inherently limited, however, because providing experimental data on all exposure situations of interest for hazard assessment is prac- tically impossible, and obtaining such a broad data base in man is ethically impossible. One approach to this problem is to develop mathematical dosimetry models of the lung that take into account not only the physical and chemical properties of the inhaled toxicant but also the anatomical

366 FREDERICK J. MILLER et al. 100 90 ~ 80 a' 70 a, 60 50 ~ 40 A 30 a) is, 20 ~ 10 TO OF O -10 ~ / 0 ~ ~ ~ / ~ / ° '.? lo. ''o ~ A': 1 1 1 1 1 1 1 1 1 ~ O . 0 .1 .2 .3 .4 .5 .6 .7 .8 .9 1.0 Predicted Pulmonary O3 Tissue Dose per Body Weight (119 O3/9) FIGURE 6 Estimated dose-response curves. Percent increased lung weight caused by O3 ex- posure versus predicted pulmonary O3 tissue dose per rat body weight resulting from 25% (o, dashed line) and 10% (., solid line) upper respiratory tract removal is plotted. Dose is the total quantity of O3 absorbed per body weight during exposure. Curves are fit by linear regression. Data from the CTR data management system, such as breathing frequency, time of exposure, animal weight, increased lung weight, and exposure concentration, combined with the PBD model predictions, as plotted for 25% head removal in Figure 5 and in other similar figures, were used to construct this figure. and physical properties of the lung of the exposed animal species or man. This approach, with dose predicted by respiratory tract region or airway generation, provides (1) a guide for future biological experimentation, (2) a framework for further refinement and validation of the assumptions of mathematical dosimetry models, and (3) a basis for quantitative inter- and intraspecies comparisons of toxicological results. A data base management system is available to store and retrive quan- titative data in a manner that is usable as input files in dosimetric models. The effects of O3 on all studied organ systems have been encoded, and citations reporting health effects are retrievable. Scaling for age, weight, sex, and species for critical characteristics are provided to yield appropriate data to use in conjunction with the PBD model for O3. Predicted dose and toxicological response are combined to establish dose-response re- lationships. The methodology was illustrated for pulmonary edema, in which results from several laboratories using different exposure times, concentrations, animal sizes, and experimental methods were consolidated to obtain a single dose-response relationship. Ultimately, when the O3

Establishing Dose-Response Relationships 367 dosimetry modeling approach is sufficiently developed, it can provide an improved basis for estimating dose-response relationships for toxic effects observed in experimental animal studies for the purpose of estimating human risk. These assessments, in turn, can serve as an important input to risk management decisions concerning the regulation of O3. ACKNOWLEDGMENTS We are most appreciative of the efforts of Mr. John Boger in the generation of model predictions for use in the analyses. We also wish to express their appreciation to Ms. Carolyn Wheeler for her excellent typing of the manuscript. This work was supported in part by EPA Contracts 68-02-3809 and 68- 02-3869 and by grants RR01693 and CA14236 from the National Institutes of health. REFERENCES Barry, B. E., F. J. Miller, and J. D. Crapo. 1985. Effects of inhalation of 0.12 and 0.25 ppm ozone on the proximal alveolar region of juvenile and adult rats. Lab. Invest. 53:692-704. Boorman, G. A., L. W. Schwartz, and D. L. Dungworth. 1980. Pulmonary effects of prolonged ozone insult in rats. Lab. Invest. 43(2):108-115. Chang, L., J. A. Graham, F. J. Miller, J. J. Ospital, and J. D. Crapo. 1986. Effects of subchronic inhalation of low concentrations of nitrogen dioxide. Toxicol. Appl. Phar- macol. 83:46-61. Corn, M., N. Kotsko, and D. Stanton. 1976. Mass-transfer coefficient for sulphur dioxide and nitrogen dioxide removal in cat upper respiratory tract. Ann Occup. Hyg. 19:1. Fenters, J. D., J. P. Findlay, C. D. Port, R. Ehrlich, and D. L. Coffin. 1973. Chronic exposure to nitrogen dioxide. Arch. Environ. Health 27:85-89. Fujinaka, L. E., D. M. Hyde, C. G. Flopper, W. S. Tyler, D. L. Dungworth, and L. O. Lollini. 1985. Respiratory bronchiolitis following long-term ozone exposure in bonnet monkeys: A morphometric study. Exp. Lung Res. 8:167-190. Gehr, P., M. Bachofen, and E. R. Weibel. 1978. The normal human lung: Ultrastructure and morphometric estimation of diffusion capacity. Respir. Physiol. 32:121-140. Hyde, D., J. Orthoefer, D. Dungworth, W. Tyler, R. Carter, and H. Lum. 1978. Mor- phometric and morphologic evaluation of pulmonary lesions in beagle dogs chronically exposed to high ambient levels of air pollutants. Lab. Invest. 38:455-469. Ichikawa, I., and E. Yokoyama. 1982. Effect of short-term exposure of ozone on the lecithin metabolism of rat lung. J. Toxicol. Environ. Health 10:1005-1015. Jeffrey, P. K., and L. E. Reid. 1979. The respiratory mucous membrane. Pp. 193-245 in Respiratory Defense Mechanism, Part 1, J. D. Brain, D. F. Proctor, and L. M. Reid, eds. New York: Marcel Dekker. Kliment, V. 1973. Similarity and dimensional analysis, evaluation of aerosol deposition in the lungs of laboratory animals and man. Folia Morphol. 21:59-64. Maitani, T., and K. T. Suzuki. 1981. Changes to essential metal levels in lungs of rats acutely exposed to ozone. Toxicol. Lett. 8:99-104.

3663 FREDER ICK J. M ~ LLER et al. Miller, F. J., C. A. McNeal, J. M. Kirtz, D. E. Gardner, D. L. Coffin, and D. B. Menzel. 1979. Nasopharyngeal removal of ozone in rabbits and guinea pigs. Toxicology 14:273- 281. Miller, F. J., J. H. Overton, Jr., R. H. Jaskot, and D. B. Menzel. 1985. A model of the regional uptake of gaseous pollutants in the lung. I. The sensitivity of the uptake of ozone in the human lung to lower respiratory tract secretions and exercise. Toxicol. Appl. Pharmacol. 79:11-27. Nambu, Z., and E. Yokoyama. 1981. The effect of age on the ozone-induced pumonary edema and tolerance in rats. Jpn. J. Ind. Health 23:146-150. NRC (National Research Council, Committee on the Institutional Means for Assessment of Risks to Public Health, Commission on Life Sciences). 1983. P. 91 in Risk Assessment in the Federal Government: Managing the Process. Washington, D.C.: National Academy Press. Overton, J. H. 1984. Physicochemical processes and the formulation of dosimetry models. J. Toxicol. Environ. Health 13:273-294. Overton, J. H., R. C. Graham, and F. J. Miller. 1987. A model of the regional uptake of gaseous pollutants in the lung. II. The sensitivity of ozone uptake in the laboratory animal lungs to anatomical and ventilatory parameters. Accepted for publication by Toxicol. Appl. Pharmacol. 88:418-432. Patra, A. L., A. Gooya, and K. T. Morgan. 1986. Airflow characteristics in a baboon nasal passage. J. Appl. Physiol. 61:1959-1966. Plopper, C. G., C. K. Chow, D. L. Dungworth, and W. S. Tyler. 1979. Pulmonary alterations in rats exposed to 0.2 and 0.1 ppm ozone: A correlated morphological and biochemical study. Arch. Environ. Health 34:390-395. ~~r----Y; )~ Riddick, J. A., K. I. Campbell, and D. L. Coffin. 1968. Histopathologic changes secondary to nitrogen dioxide exposure in dog lungs. Am. J. Clin. Pathol. 49:239. (Abstract) Schreider, J. 1986. Comparative anatomy and function of the nasal passages. Pp. 1-25 in Toxicology of the Nasal Passages, C. S. Barrow, ed. Washington, D.C.: Hemisphere. Schreider, J., and O. Raabe. 1981. Anatomy of the nasal pharyngeal airways of experimental animals. Anat. Rec. 200:195-205. Schreider, J. P., and J. O. Hutchens. 1980. Morphology of the guinea pig respiratory tract. Anat. Rec. 196:313-321. Smolko, E. D., D. J. McKee, and D. B. Menzel. In press. Critical toxicity reference system. I. An approach for managing quantitative toxicity data. J. Am. College Toxicol. Stephens, R. J., M. F. Sloan, M. J. Evans, and G. Freeman. 1973. Early response of lung to low levels of ozone. Am. J. Pathol. 74:31-58. Stephens, R. J., M. F. Sloan, M. J. Evans, and G. Freeman. 1974. Alveolar type 1 cell response to exposure to 0.5 ppm O3 for short periods. Exp. Mol. Pathol. 20:11-23. Suzuki, T. 1976. Studies on the acute toxicity of 5-hydroxytryptamine in the rats pre- exposed to ozone. Bull. Tokyo Med. Dent. Univ. 23:217-226. Weibel, E. R. 1963. Morphometry of the human lung. New York: Academic Press. Yeh, H. C., and G. M. Schum. 1980. Models of human lung airways and their application to inhaled particle deposition. Bull. Math. Biol. 42:461-480. Yeh, H. C., G. M. Schum, and M. T. Duggan. 1979. Anatomic models of the tracheo- bronchial and pulmonary region of the rat. Anat. Rec. 195:483-492. Yokoyama, E. 1968. Uptake of SO2 and NO2 by the isolated airways. Bull. Inst. Public Health 17:302-306. Yokoyama, E., and R. Frank. 1972. Respiratory uptake of ozone in dogs. Arch. Environ. Health 25: 132- 138.

Role of Pharmacokinetic Modeling in Risk Assessment: PerchIoroethylene as an Example Chao W. Chen and Jerry N. Blancato INTRODUCTION The assessment of potential risks to humans from xenobiotic chemicals is a major component of the environmental decision-making process. To ensure high-quality assessments, the risk assessor must consider all avail- able data and perform a wide range of interpretations. All such interpre- tations should be based on sound and logical scientific thought and should use state-of-the-art methodologies. Ideally, prediction of human risk because of exposure to environmental agents should be made on the basis of human experience and sound laboratory findings. The available human data are, however, rarely suf- ficient for this purpose. A common alternative approach is to predict human risk on the basis of experimental animal data. The estimation of human risk from these types of data involves several steps and assump- tions. These steps include extrapolation from risk in animals exposed at high doses to risk in animals that might be exposed at low doses, because the available experimental data frequently are the result of high-dose experiments. Thus, even at this first level of extrapolation, there are seldom enough data to permit verification of the assumptions made with regard to the shape of the dose-response relationship at low doses. Other steps in the human risk assessment process include extrapolating from The views expressed in this paper are those of the authors and do not necessarily reflect the views or policies of the U.S. Environmental Protection Agency. 369

370 CHAD W. CHEN AND JERRY N. BLANCATO animal low-dose risk to human low-dose risk (species conversion) and from one route of exposure to another (route-to-route extrapolation). Each of these steps involves certain assumptions; for high- to low-dose extrap- olation, usually a parametric dose-response relationship is assumed; for species conversion, the assumption is usually made that animals and hu- mans have equal lifetime risks from exposure when the dose rate is ex- pressed in a particular unit (equivalent dose). The dose units that have been used as equivalent doses include milligrams/body surface area/day, milligram/kilogram of body weight/day, milligram/target tissue volume (or mass)/day, parts per million in air, parts per million in diet, and milligrams/kilogram/lifetime. These conversion factors are presumed to account for all of the differences between animals and humans, including longevity, body size, and metabolic processes and rates. The basis for these conversions and extrapolations is not well documented or proven. For route-to-route extrapolation, the Stokinger and Woodward (1958) ap- proach is usually employed, with the assumption of 100% absorption from air and oral exposure. This assumption can result in significant error in a number of cases. For example, there is little evidence that many substances are fully absorbed after they are inhaled, even at low concentrations. In the case of ingestion by the oral route, less than 100% absorption may occur, depending on the physicochemical nature of the substance, diet contents, species, and other factors. The use and analysis of pharmacokinetic and metabolic information in risk assessments may reduce some of the ambiguities associated with these three major areas of uncertainty, namely, high- to low-dose extrapolation, species conversion, and route-to-route extrapolation. This is particularly true of those uncertainties associated with route-to-route and, to a lesser extent, high- to low-dose extrapolation. Consideration of metabolism and pharmacokinetics alone cannot completely eliminate uncertainties, how- ever, especially in the case of interspecies extrapolation. It is essential that the carcinogenic processes that are affected by environmental agents be better understood if the uncertainties involved in risk assessment are to be minimized. Pharmacokinetic analyses can range from very simple arithmetic op- erations, which are used to calculate the amount of the dose that is con- verted to some suspected or proven toxin, to complicated physiologically based pharmacokinetic (PB-PK) models. PB-PK models can be used to study the kinetics of absorption, distribution, metabolic processing, and elimination of the parent compound and its metabolites. They may also be used to predict the disposition of a compound at different dosage regimens. In principle, such predictions may be made across species lines, thus greatly reducing the number of different species of animals that must be used to test a given compound. When properly formulated, PB-PK

PK Modeling: Perchloroethylene 371 models may give information regarding the concentration of a putative toxin at a target tissue or cell. Most importantly, they can provide a description of the disposition of a compound over time rather than just at some predetermined endpoint. As these models become better tested, a variety of questions pertaining to particular circumstances can be an- swered. For example, the effects of less-than-lifetime exposure, inter- mittent exposure, or bursts of exposure can be examined; and differences with regard to various parameters across and within species can be mea- sured and accounted for systematically. As more and more parameters are considered and accurately determined, the uncertainties involved in risk assessment will be correspondingly reduced. This paper discusses the use of metabolic and pharmacokinetic data on perchloroethylene (PCE) in the formulation of PB-PK models. The results of these models are applied in a cancer risk assessment of PCE. The emphasis here is on methodologies for improving risk assessment, rather than on the risk assessment of PCE per se. The basis for the discussion contained in this paper was work done in preparation of an Addendum to the Health Assessment Document for Tetrachloroethylene (EPA, 19861. PCE offers an attractive opportunity to evaluate the usefulness of PB-PK modeling in cancer risk assessment for the following reasons: 1. The same tumors (hepatocellular carcinomas) have been observed in both male and female mice exposed to PCE by Savage and inhalation. This information permits the evaluation of uncertainty with regard to route- to-route extrapolation. 2. The metabolism and pharrnacokinetics of PCE have been studied extensively in both laboratory animals and humans. The available data are sufficient to enable a PB-PK model to be independently constructed for each of the three species: humans, rats, and mice. These results serve as a useful basis for evaluating certain assumptions about parameters in human models when not enough data exist to permit construction of a PB-PK model for humans. PCE is biotransformed by microsomal monooxygenases (cytochrome P-450 system) to reactive metabolites, including the putative carcinogen trichloroacetic acid (TCA) and a short-lived, highly reactive epoxide intermediate that can bind covalently to cellular macromolecules. It is generally considered that the hepatic toxicity and carcinogenicity potential of PCE resides in its biologically reactive intermediate metabolites rather than in the parent compound itself. This factor further strengthens the argument for using metabolites, rather than the administered dose, in risk assessment.

372 CHAD W. CHEN AND JERRY N. BLANCATO METABOLIC DATA PERTINENT TO THE CONDITIONS OF PCE CARCINOGEN BIOASSAYS Oral Studies Pegg et al. (1979) and Schumann et al. (1980) administered t~4C]PCE in corn oil vehicle as single intragastric doses to rats and mice. ~4C- radioactivity was determined for exhaled breath, urine, feces, and carcass for 72 h following dosing. Pulmonary excretion of unchanged PCE was measured. The data from these two studies were collected and collated and then re-expressed in milligram equivalent units (Table 11. Inhalation Studies ANIMAL DATA Pegg et al. (1979) and Schumann et al. (1980) also determined body burdens and metabolism of rats and mice after inhalation exposure to PCE for 6 h. The animals were exposed to ti4C]PCE; the radioactivity was determined in urine, feces, expired air, etc.; and unchanged t~4C]PCE was determined in expired air for 72 h postexposure. These data are TABLE 1 Disposition of [~4C]PCE Radioactivity for 72 h After Single Oral Doses to Sprague-Dawley Rats and B6C3F1 Mice mg Eq./animal Rats a Drug 1 mg/kg 500 mg/kg Micea(500 mg/kg, disposition (0.25 mg/animal) (125 mg/animal) 12.25 mg/animal) Expired unchanged 0.174 (71%) 110.67 (90%) 8.90 (83%) Metabolized '4CO2 0.007 0.57 0.14 Urine 0.040 5.72 1.53 Feces 0.015 4.82 0.13 Carcass 0.008 1.41 0.05 0.070 (29%) 12.52(10%) 1.85 (17%) Total recovered 0.244 123.19 10.75 aThere was an average of three rats and three mice. Data are based on average experimental animal weights: 250 g for rats and 24.5 g for mice. SOURCE: Adapted from Pegg et al. (1979) and Schumann et al. (1980).

PK Modeling: Perchloroethylene 373 TABLE 2 Disposition of [~4C]PCE Radioactivity for 72 h After Inhalation Exposure for 6 h by Sprague-Dawley Rats and B6C3F1 Mice mg Eq./animal Drug Ratsa Micea disposition 10 ppm 600 ppm (10 ppm) Expired unchanged Metabolized 14cO2 Urine Feces Carcass Total recovered 1.008 (demo) 0.053 0.275 0.076 0.063 0.467 (32%) 1.475 68.39 (88%) 0.54 4.54 2.36 1.67 9.11 (demo) 77.50 0.048 (12%) 0.032 0.285 0.027 0.012 0.356 (88%) 0.404 aThere was an average of three rats and three mice. Data are based on average experimental animal weights: 250 g for rats and 24.5 g for mice. SOURCE: Adapted from Pegg et al. (1979) and Schumann et al. (1980). collated and re-expressed for comparison of rat and mouse metabolism in Table 2. Metabolic data from oral and inhalation studies have been used to estimate metabolic constants in the PB-PK models for mice and rats. Further details on the use of these data are provided in Appendix A. HUMAN DATA Estimates of the extent of PCE metabolism in man by measuring me- tabolites excreted in urine have been made from several human volunteer studies. A review by Monster (1984) provides useful data for constructing TABLE 3 Mean Amounts of the PCE Metabolite TCA Excreted in Human Unne over a 72-h Penod After a Single Inhalation Exposure to PCE Exposure Duration Observed concentration of exposure TCA No. of (mg/liter) (h) (mg) subjects Reference 0.480 4 5.9 6 Monster et al., 1979 0.960 4 11.2 6 Monsteretal., 1979 1.029 8 23.5 2 Fernandez et al., 1976

374 CHAD W. CHEN AND JERRY N. B~NCATO TABLE 4 Mean Concentrations of PCE in Human Alveolar Air After a Single Inhalation Exposure to PCE Exposure Duration concentration of exposure Postexposure alveolar concentration (mg/liter) at: (mg/liter) (h) 0.5 h 1 h 2 h 16 h 64 h 0.686 8 0.137 0.103 0.082 0.027 0.014 1.029 8 0.278 0.216 0.165 0.042 0.020 1.371 8 0.343 0.260 0.192 0.055 0.027 SOURCE: Fernandez et al. (1976). a PB-PK model for humans. The data presented in Tables 3 and 4 have been used to estimate the metabolic constants in the PB-PK model. These data were selected because they appeared to be obtained from well- conducted studies and were from similarly designed experiments. Addi- tional details on the utilization of these data to estimate metabolic constants in the PB-PK model are given in Appendix A. The metabolic data presented in Tables 1 through 4 enable us to in- dependently construct a PB-PK model for each of the three species: hu- mans, rats, and mice. These data have been used to optimize the metabolic constants Vm and Km. The following sections describe the use of these models to assess the cancer risk of PCE. TABLE 5 Tumor Incidence and the Corresponding Metabolized Dose for B6C3F1 Mice in NCI (1977) Gavage Study, Calculated by PB-PK Model Daily metabolized Hepatocellular Administered dose (mg/kg/ carcinoma do se ( mg/kg/day ) day )-a inc idenceb Males (0.03 kg) 0 0 2/20 536 56.32 32/48 1,072 75.25 27/45 Females (0.025 kg) 0 0 0/20 386 40.58 19/48 772 57.52 19/45 aDaily metabolized dose = Average daily metabolized dose x 78 week/90 week, where the factor 78/90 reflects the fact that mice were exposed to PCE for 78 weeks out of a 90-week lifetime. The average metabolized doses are given in Table 7. bOnly animals surviving beyond the first liver tumor death (27 weeks for males and 41 weeks for females) are included in the denominators.

PK Modeling: Perchloroethylene 375 TABLE 6 Tumor Incidence and the Corresponding Metabolized Dose for B6C3F1 Mice in NTP (1985) Inhalation Bioassay, Calculated by PB-PK Model Daily metabolized dose (mglkgl day)a Administered dose (ppm) . Males (0.03 kg) Hepatocellular carcinoma incidenceb 0 0 7/49 100 47.23 25/47 200 66.29 26/50 Females (0.025 kg) 0 0 1/46 100 54.97 13/42 200 76.76 36/47 aTaken from Table 8. bOnly animals surviving beyond the first death from liver tumor (60 weeks) are included in the denominators. USE OF PB-PK MODEL TO ESTIMATE METABOLIZED DOSE FOR NCI (GAVAGE) AND NTP (INHALATION) BIOASSAY MICE As mentioned earlier, PCE experiments in mice offer an attractive opportunity to evaluate the usefulness of PB-PK models in cancer risk assessment because tumors have been induced in the same site by both gavage and inhalation routes of exposure. The dose-response data from gavage (NCI, 1977) and inhalation (NTP, 1985) are presented, respec- tively, in Tables 5 and 6. Under NCI gavage or NTP inhalation dosing patterns, the amount of metabolized dose reaches the steady state during weekdays and is almost totally eliminated during the weekend (nonexposed period). Tables 7 and 8 present daily metabolites (in milligrams/day) for mice in the NCI and NTP bioassays. CALCULATION OF THE DOSE-RESPONSE RELATIONSHIP The two-stage model, P(d) = 1 - expL—qO—qid—q2d2), (1) is used to calculate the dose-response relationship by using hepatocellular carcinoma incidence rates and the corresponding metabolized dose (in milligrams/kilogram/day) presented in Tables 5 and 6. The incremental cancer risk (for mice) is defined as F(d) ~ EP(d)—P(O)~/~1 —P(O)] = 1 —expE—qid—q2d21. (2) The maximum likelihood estimates of parameters q~ and q2, along with

376 CHAD W. CHEN AND JERRY N. B~NCATO the 95% upper limit of A, denoted qu, are presented in Table 9. At low doses, the upper-bound risk is predicted by qU x d at dose rate d (mil- ligram of metabolites/kilogram/day). It is seen from Table 9 that qu, cal- culated on the basis of different data bases, Savage or inhalation, male or female mice, are all comparable, ranging from 6.27 x 1O-3 to 1.76 x 10-2 per mg of metabolites/kg/day. the geometric means of qU are 1.6 x 10-2 and 9.4 x 1O-3, respectively, calculated from gavage and inhalation bioassays. All risk estimates presented in this paper are 95% upper-limit estimates. To predict human cancer risk caused by exposure to PCE, the assump- tion must be made as to what constitutes an equivalent dose among species, that is, the dose unit that induces the same magnitude of tumor response in animals and humans. Table 10 presents risk for humans continuously exposed to PCE by inhalation under various equivalent dose assumptions. Table 11 presents parallel risk estimates as in Table 10 when exposure is 8 in/day, 5 days/week. In these calculations, the geometric means of the two upper-bound risks calculated from male and female mice dose-re- sponse curves are used. TABLE 7 Daily Amounts (mg) of Total Metabolites in Male and Female Mice Exposed to PCE by Gavage According to the NIP Bioassay Pattern In males (0.035 kg bw) In females (0.025 kg bw) At low dose At high dose At low dose At high dose (536 (1,072 (386 (772 Day mg/kg/day) mg/kg/day) mg/kg/day) mg/kg/day) Week 1 1 2.675 3.545 1.621 2.291 2 2.707 3.583 1.634 2.308 3 2.708 3.584 1.634 2.308 4 2.708 3.584 1.634 2.308 5 2.708 3.584 1.634 2.308 6 0.138 0.349 0.036 0.090 7 0.002 0.006 0.0004 0.0009 Week 2 1 0.002 3.545 1.621 2.291 2 2.675 3.583 1.634 2.308 3 2.707 3.584 1.634 2.308 4 2.708 3.584 1.634 2.308 5 2.708 3.584 1.634 2.308 6 0.138 0.349 0.036 0.090 7 0.002 0.006 0.0004 0.0009 Average (week 2) 1.949 mg or 2.605 mg or 1.170 mg or 1.659 mg or 64.98 mgl 86.83 mg/kg/ 46.82 mgl 66.37 mg/kg/ kg/day day kg/day day

PK Modeling: Perchloroethylene 377 TABLE 8 Daily Amounts (mg) of Total Metabolites in Male and Female Mice Exposed to PCE by Inhalation According to the NIP Bioassay Pattern In males (0.035 kg bw) In females (0.03 kg bw) At low dose At high dose At low dose At high dose Day (100 ppm) (200 ppm) (100 ppm) (200 ppm) Week 1 1 2.273 3.176 2.285 3.150 2 2.302 3.214 2.312 3.185 3 2.303 3.215 2.312 3.186 -4 2.303 3.215 2.312 3.187 5 2.303 3.215 2.312 3.187 6 0.085 0.202 0.087 0.222 7 0.0016 0.004 0.0013 0.003 Week 2 1 2.274 3.176 2.285 3.150 2 2.303 3.214 2.312 3.185 3 2.303 3.215 2.312 3.186 4 2.303 3.215 2.312 3.187 5 2.303 3.215 2.312 3.187 6 0.085 0.202 0.087 0.222 7 0.0016 0.004 0.001 0.003 Average 1.653 mg or 2.320 mg or 1.649 mg or 2.303 mg or 47.23 mgl 66.29 mglkgl 54.97 ma/ 76.76 mglkgl kg/day day kg/day day COMPARISON OF RISK ESTIMATES FOR DRINKING WATER AND INHALATION EXPOSURE, WITH AND WITHOUT THE USE OF PB-PK MODELING The conventional approach to risk estimation, without the use of PB-PK modeling, would be to assume 100% absorption by either the oral or the inhalation route of exposure. In the following calculations, it is assumed that a person weighs 70 kg, drinks 2 liters of water per day, and inhales 15 m3 of air per day. The value of IS m3/day is used because it is approximately the value used in PB-PK modeling. For convenience, only body surface equivalent dose is used in these calculations, except where parts per million is assumed to be equivalent. When the PB-PK model is used, water intake is assumed to occur uniformly over a 16-h period in a day. Table 12 summarizes cancer risk estimates associated with 1 ~g/liter (1 ~g/m3) of PCE in water (air), calculated with and without PB-PK modeling.

378 CHAD W. CHEN AND JERRY N. B~NCATO TABLE 9 Parameter Estimates of Two-Stage Model Based on Data from Gavage and Inhalation Bioassays - Geometric at q2 qu mean Data base Gavage (NCI, 1977) Male mice 1.35 x 10-2 0 1.76 x 10-2 Female mice 1.08 x 10-2 0 1.40 x 10-2 1.6 x 10- Inhalation (NTP, 1985) Male mice 1.0 x 10-2 0 1.40 x 10-2 Female mice 0 1.87 x 10-4 6.27 x 10-3 9.4 x 10-3 arose unit = milligrams of metabolite/kilogram/day. TABLE 10 Predicted Human Cancer Risk Caused by Continuous 24- Hour Exposure to PCE by Inhalation at Different Assumed Equivalent Doses Assumed equivalent dosesa Dose Concentration metabolized mg/liver tissue (ppm) (mg/day) mglW2'3lday mg/kg/day volume/day ppmb _ 0.01 0.0215 3.8 x 10-5 2.9 x 10-6 5.6 x 10-6 2.6 x 10-4 0.10 0.215 3.8 x 10-4 2.9 x 10-5 5.6 x 10-5 2.6 x 10-3 1.00 2.15 3.8 x 10-3 2.9 x 10-4 5.6 x 10-4 2.6 x 10-2 50 97.56 1.6 x 10-' 1.3 x 10-2 2.5 x 10-2 1.0 x 10° 100 178.47 2.7 x 10-i 2.3 x 10-2 4.6 x 10-2 1.0 x 10° aHuman body weight (W) is assumed to be 70 kg. bppm in air is assumed to be equivalent between mice and humans. TABLE 11 Predicted Human Cancer Risk Caused by Inhalation Exposure (8 in/day, 5 days/week) at Different Assumed Equivalent Doses Dose Assumed equivalent doses Concentration metabolized mg/liver tissue (ppm) (mg/day) mglW2'3lday mg/kg/day volume/day ppm 0.01 5.12 x 10-3 9.0 x 10-6 6.9 x 10-7 1.3 x 10-6 6.0 x 10-5 0.10 5.12 x 10-2 9.0 x 10-5 6.9 x 10-6 1.3 x 10-5 6.0 x 10-4 1.00 5.12 x 10-' 9.0 x 10-4 6.9 x 10-5 1.3 x 10-4 6.0 x 10-3 50 24.55 4.2 x 10-2 3.3 x 10-3 6.4 x 10-3 2.9 x 10-' 100 47.23 8.0 x 10-2 6.3 x 10-3 1.2 x 10-2 5.6 x 10-'

PK Modeling: Perchloroethylene 379 TABLE 12 Risk at a Unit Dose of 1 ~g/liter in Water or 1 ~g/m3 in Air With and Without the Use of PB-PK Modeling Without PB-PK modeling Based on inhalation data Medium Based on gavage data With PB-PK modeling Based on Based on inhalation Method 2a Method 2b gavage data data NEa NE 1.5 x 10-7 NE Water (1 1lg/liter) Air (1 ~g/m3) 7.4 x 10-7 6.1 x 10-6 3.8 x 10-6 8.4 x 10-6 9.4 x 10-7 5.5 X 10-7 aNE = Not estimated. There are several ways to calculate air risk without using a PB-PK model: 1. Indirect estimate from gavage data. On the basis of gavage data, the potency slope (i.e., 95% upper bound of the linear parameter in the two- stage model, under the body surface equivalent dose assumption) is cal- culated to be 2.8 x 10-2/mg/kg/day. This is equivalent to 6.1 x 10-6/ ~g/m3, using the Stokinger and Woodward (1958) procedure and the assumption of 100% absorption. 2. Direct estimate from inhalation data. a. When parts per million is assumed to be equivalent, the potency slope is 2.6 x 10~2/ppm or, equivalently, 3.8 x 10-6/~g/m3. b. When animal dose in parts per million is converted to milligrams/ kilogram/day, assuming 100% absorption, the potency slope is 3.9 x 10-2/mg/kg/day or, equivalently, 8.4 x 10-6/~g/m3. These calculations indicate that estimation of inhalation risk without adjusting for metabolism overestimates risk by about an order of magnitude when compared with that estimated by PB-PK modeling. For drinking water, the conventional approach overestimates risk by about 5 times when compared with that estimated by PB-PK modeling. It should be noted that the variability observed in these estimates is mainly due to metabolism adjustment and not to the use of different dose-response models and/or different assumptions about equivalent dose. DISCUSSION In calculating the dose-response relationship for PCE, the amount me- tabolized per day is considered to be an effective dose. The use of this

380 CHAD W. CHEN AND JERRY N. B LAN CATO surrogate effective dose may not eliminate the uncertainty associated with the low-dose extrapolation because the dose actually reaching the receptor sites may not be linearly proportional to the total amount of metabolites, and the shape of the dose-response relationship is still unknown. It seems reasonable to expect, however, that the uncertainty with regard to the low- dose extrapolation would be somewhat reduced by the use of the metab- olized dose because the metabolized dose better reflects the dose-response relationship, particularly in the high-dose region. Ideally, a dose-response function should be derived according to the mechanism of carcinogenic action if the available scientific information is sufficient. A two-event model proposed by Moolgavkar and Venzon (1979) and Greenfield et al. (1984) provides a promising framework for building a biologically based dose-response function. Under this model, the probability of cancer is a function of mitotic rate (M), transition rate (P), birth rate (B), and death rate (D) for cells at different stages: normal, initiated, or transformed. The ultimate goal is to construct a biologically based dose-response function and a physiologically based pharmacokinetic model that estimates target dose under any exposure pattern. For a given target dose d, however, one would not expect mitotic rate M(d), transition rate P(d), birth rate B(d), or death rate D(d) to be identical between animals and humans. If they were identical, one would see much more cancer incidence in humans than in animals, because humans have more cells and a longer life span than animals. Therefore, even if target dose is known, one still faces the problem of dose scaling unless one can measure M(d), P(d), B(d), and D(d) in both humans and animals. PCE provides a rare opportunity to compare the relationship of the metabolic constants Vmax and Km among species. The allometric equations calculated from data from mice, rats, and humans are as follows: Vmax = 0.016W°89, and Km = S.OW0 44, (3) (4) where W is body weight, in kilograms. These equations imply: Vmaxthumans)/Vmaxfanimals) = (Wh/Wa)089, and (5) Km~humans)/Km~animalS) = (Wh/wa)° 44' (6) where Wh and Wa are body weights for humans and animals, respectively. These observations are not consistent with assumptions that are usually made in the absence of human data; i.e., the assumptions that Vmax is related to body weight by the 0.67th power and that K,71 is identical between animals and humans (Ramsey and Andersen, 19841. However, one should not make any judgment on the basis of a single observation.

PK Modeling: Perchloroethylene 381 SUMMARY The overall goal of this paper has been to demonstrate the use of a physiologically based pharmacokinetic model to estimate the risk to hu- mans from various concentrations of PCE. A relatively simple physiologically based pharmacokinetic model has been used to calculate the amount of PCE metabolized by animals that showed tumor responses in bioassay testing. From the amount metabo- lized, a multistage model was used to calculate the dose-response rela- tionship for mice. Subsequently, the physiologically based pharmacokinetic model, with appropriate changes in parameters to account for human physiological and biochemical processes, was used to calculate the amounts of PCE that would be metabolized in humans under various ambient exposure concentrations. The amounts, as determined from the model, were then used, together with the dose-response function from the mouse bioassay studies, to calculate human risk. The metabolized amounts calculated for humans were expressed in a variety of equivalent forms, because some expression of dosimetry is needed to equilibrate the tumor response between species. Because there is no agreed upon standard factor for species conversion, several forms of equivalent dose have been chosen to express the risk calculated from the metabolized dose. The cancer potency per unit of metabolized dose appears to be comparable regardless of whether gavage or inhalation data are used. Comparisons were also made on the calculation of risk for inhalation based on potencies calculated from gavage, without consideration for route-to-route differences. The results of calculating the human risk by the pharmacokinetic model method show that by using the gavage potency, without accounting for pharmacokinetics, calculating risk for inhalation results in error. Thus, pharmacokinetic considerations must be accounted for when risk is to be calculated from animal studies using a different route of exposure from the route of potential human exposure. From these PB-PK model analyses, it is observed that the daily for- mation of metabolites reaches the steady state faster in rodents than in humans. One possible explanation for this is that the percentage of body fat is greater in humans than in rodents. FUTURE DIRECTIONS We believe that we have demonstrated here the utility and advantages of using physiologically based pharmacokinetic modeling in exposure and risk assessments. We anticipate that in the future this approach will be used more extensively. As further data become available on the charac-

382 CHAD W. CHEN AND JERRY N. BLANCATO teristics of the disposition, metabolism, and mechanism of action of other compounds, PB-PK models will become correspondingly more sophisti- cated. Already, models have been formulated to describe the disposition of foreign chemicals in far greater detail than the model presented here (Bischoff et al., 1971; King et al., 1986; Lutz et al., 19761. The concen- trations of the metabolites can be described and predicted not only in the organ of formation but also in other body regions after hemodynamic transport. The sequestering of parent compound and metabolite can also be described (Angelo et al., 19841. This situation can be of particular importance for compounds exhibiting a long biological life, as, for ex- ample, in the adipose tissues. It is widely accepted that differences in exposure patterns could have an effect on the true human risk, which is not discernible by the traditional risk calculation approach. As demon- strated in this paper, with the use of these types of models, information can be quickly and easily obtained to help determine those exposure conditions under which risk is greatest, and what types of studies are needed to elucidate more information. As more becomes known about mechanisms of action, risk assessors will benefit from knowing the toxin concentrations at subcellular sites. For example, dose-response functions can be calculated by using the concentration at a particular subcellular site as the dose. Describing and modeling exposure at such sites requires a greater understanding of mechanisms of action but ultimately can result in reducing some uncertainties that are inherent in present risk assessments. Subcellular models have already been formulated for the purpose of ex- posure assessments (Blancato and Bischoff, 1985; Gehring and Young, 19781. These types of models, and the data that are generated in their formulation, are valuable in providing guidance in the design and imple- mentation of key laboratory studies that help to identify and discern the mechanisms of action of toxic substances. ACKNOWLEDGMENT The authors wish to thank members of the Carcinogen Assessment Group and the Exposure Assessment Group, Office of Health and Envi- ronmental Assessment, for their contributions to this presentation. REFERENCES Angelo, M. J., K. B. Bischoff, A. B. Pritchard, and M. A. Presser. 1984. A physiological model for the pharmacokinetics of methylene chloride in B6C3F1 mice following i.v. administrations. J. Pharmacokinet. Biopharm. 12(4):413-436.

PK Modeling: Perchloroethylene 383 Bischoff, K. B., R. L. Dedrick, D. S. Zaharko, and J. A. Longstreth. 1971. Methotrexate pharmacokinetics. J. Pharm. Sci. 60(8):1128-1133. Blancato, J. N., and K. B. Bischoff. 1985. Sub-cellular pharmacokinetics of 2,5-hexane dione. Paper presented at North American Symposium on Risk Assessment and the Biological Fate of Xenobiotics, Key Biscayne, Fla., November 17-22, 1985. EPA (U.S. Environmental Protection Agency). 1986. Addendum to the Health Assessment Document for Tetrachloroethylene (Perchloroethylene): Updated Carcinogenicity As- sessment for Tetrachloroethylene (Perchloroethylene, PERC, PCE). EPA/600/8-82/00SFA. External Review Draft. (Available from National Technical Information Service, Spring- field, Va., as publication no. PB86-174489). Fernandez, T., E. Guberan, and J. Caperos. 1976. Experimental human exposures to tetra- chloroethylene vapor and elimination in breath after inhalation. Am. Ind. Hyg. Assoc. ~ J. 37(March): 143- 150. Gehring, P. J., and J. P. Young. 1978. Application of pharmacokinetic principles in practice. Pp. 119-141 in Toxicology: Proceedings of First International Congress on Toxicology, G. L. Plaa and W. H. M. Duncan, eds. New York: Academic Press. Gehring, P. J., L. Watanabe, and C. Park. 1979. Risk of angiosarcoma in workers exposed to vinyl chloride as predicted from studies from rats. Toxicol. Appl. Pharmacol. 49: 15- 21. Greenfield, R. E., L. B. Ellwen, and S. M. Cohen. 1984. A general probabilistic model of carcinogenesis: Analysis of experimental urinary bladder cancer. Carcinogenesis 5(4):437- 445. Hattis, D., S. Tuler, L. Finkelstein, and Z.-Q. Luo. 1986. A pharmacokinetic/mechanism- based analysis of the carcinogenic risk of perchloroethylene. Cambridge, Mass.: Center for Technology, Policy and Industrial [Development, Massachusetts Institute of Tech- nology. King, F. L., R. L. Dedrick, and F. F. Farris. 1986. Physiological pharmacokinetic modeling of cis-dichlorodiamineplatnium (II) (DDP) in several species. J. Pharmacokinet. Bio- pharm. 14(2):131-155. Lutz, R. J., R. L. Dedrick, H. B. Matthews, T. E. Eling, and M. W. Anderson. 1976. A preliminary pharmacokinetic model for several chlorinated biphenyls in the rat. Drug Metab. Dispos. 5(4):386-396. Monster, A. C. 1984. Tetrachloroethylene. Pp. 131-139 in Biological Monitoring and Surveillance of Workers Exposed to Chemicals, A. Aitio, V. Riihimaki, and H. Vainio, eds. New York: McGraw-Hill. Monster, A. C., G. Boersma, and H. Steenweg. 1979. Kinetics of tetrachloroethylene in volunteers; influence of exposure concentrations and work load. Int. Arch. Occup. En- viron. Health 42:303-309. Moolgavkar, S. H., and D. J. Venzon. 1979. Two-event models for carcinogenesis: In- cidence curves for childhood and adult tumors. Math. Biosci. 47:55-77. NCI (National Cancer Institute). 1977. Bioassay of tetrachloroethylene for possible car- cinogenicity. DHEW Pub. No. (NIH) 77-813. Bethesda, Md.: Public Health Service, National Institutes of Health, U.S. Department of Health, Education, and Welfare. NTP (National Toxicology Program). 1985. Toxicology and carcinogenesis of tetrachloro- ethylene in F344/N rats and B6C3F1 mice (inhalation studies). Draft technical report. Ohitsu, T., K. Sato, A. Koisumi, and M. Kumai. 1983. Limited capacity of humans to metabolize tetrachloroethylene. Int. Arch. Occup. Environ. Health 51:381-390. Pegg, D. G., J. A. Zempel, W. H. Brown, and P. G. Watanabe. 1979. Deposition of tetrachloro(~4C)ethylene following oral and inhalation exposure in rats. Toxicol. Appl. Pharmacol. 51 :465-474.

384 CHAD W. CHEN AND JERRY N. BLANCATO Ramsey, T. C., and M. E. Andersen. 1984. A physiologically-based description of the inhalation pharmacokineties of styrene in rats and humans. Toxicol. Appl. Pharmaeol. 79:389-400. Schumann, A. M., T. F. Quast, and P. G. Watanabe. 1980. The pharmaeokineties of perehloroethylene in mice and rats as related to oneogenieity. Toxieol. Appl. Pharmaeol. 55:207-219. Stokinger, M. E., and R. L. Woodward. 1958. Toxicological methods for establishing drinking water standards. J. Am. Water Works Assoe. 50:517.

PK Modeling: Perchloroethylene 385 APPENDIX: CONSTRUCTION OF PB-PK MODELS Description of the PB-PK Mode' for inhaled PCE PCE in vapor form in air is readily absorbed through the lungs into blood by first-order diffusion processes. Pulmonary uptake of PCE is largely determined by the ventilation rate, duration and concentration of exposure, solubility in blood and body tissues, and metabolism. PCE is eliminated by pulmonary excretion and metabolism (primarily in the liver). Figure A-1 depicts the PB-PK model used to simulate PCE distribution and metabolism over time. This model was used by Ramsey and Andersen (1984) to study the behavior of inhaled styrene. Abbreviations used in the figure are given in Table A-1. The model in Figure A-1 is described by a system of differential equa- tions that quantify the rate of change of PCE concentration in each tissue group over time. Pulmonary uptake and elimination are described by the following equation of mass balance of PCE entering and leaving the lungs: Qa(C! cay = Q~(Cart CvenJ (A-1) This equation assumes a concentration equilibrium between arterial blood and alveolar air. By substituting the relationship N = Car~/Ca into the equation above, we have: Cart = (QaC~ + Q~Cven)/(Q' + QalN), (A-2) where venous blood concentration is given by: Cven = (QiC`IPi + QfcflPf + Qrcrlpr + QpCplPp~lQ~. (A-3) The metabolism of PCE occurs mainly in the liver, with a rate of metabolism characterized by the Michaelis-Menten type equation: dt (VmCilPi~l (Km + CiIPi) . (A-4) The PCE concentrations in each of the four tissue groups are described by the following equations. Metabolism tissue group: ~ ~ — Ql(Cart Nonmetabolism tissue group: - CiIPiJ Vi aft' = Qi(Cart — CilPi), dAm dt (A-S) (A-6)

386 CHAD W. CHEN AND JERRY N. BLANCATO Qa Q.' Alveolar Space · Lung Blood Oven . Cvf I Fat risque Group _ Cvr Richly Perfused Tissue Group C)vp ~ Poorly Perfused | Tissue Group Cvl Liver (Metabolizing) Tissue Group 1 °a Qt Cart Qf .4 Cart Q can Cart Q. Cart I m ~ Metabolites Km FIGURE A- 1 Diagram of the physiologically based pharmacokinetic model. Abbreviations are defined in Table A-1.

PK Modeling: Perchloroethylene 387 Table A-1 Abbreviations Used in Figure A-1 Qa Alveolar ventilation rate (liters/min) A Cardiac blood output (liters/min) A Blood flow rate to liver (metabolizing) tissue group (liters/min) Of Blood flow rate to fat tissue group (liters/min) Qr Blood flow rate to richly perfused tissue group (liters/min) Qp Blood flow rate to poorly perfused tissue group (liters/min) Via, Vf, Vr, and Vp Volumes of tissue groups (liters) corresponding, respectively, to liver, fat, richly perfused, and poorly perfused tissue groups N Blood:air partition coefficient Pit Pf, Pr' and Pp Tissue:blood partition coefficient corresponding, respectively, to liver, fat, richly perfused, and poorly perfused tissue groups Vm Maximum velocity of metabolism (mg/min) Km Michaelis constant (mg/liter of blood) Am Amount metabolized (mg) Car, Concentration in arterial blood (mg/liter) CVen Concentration in venous blood (mg/liter) C/, Cf. Cr. and Cp Concentrations in tissue groups corresponding, respectively, to liver, fat, richly perfused, and poorly perfused tissue groups Car Concentration in inhaled air (mg/liter) C`, Concentration in alveolar air (mg/liter) where the subscript i represents fat, richly perfused tissue groups. Parameters Used in the Mode' , and poorly perfused Table A-2 presents the physiological and biochemical parameters used in the model. How these parameters were obtained is described below. PHYSIOLOGICAL PARAMETERS These values are taken from the U. S. Environmental Protection Agency (EPA) preliminary draft entitled "Reference Physiological Parameters in Pharmacokinetic Modeling," a contract project headed by Dr. Curtis Travis of the Oak Ridge National Laboratory, Oak Ridge, Tenn. The document has exhaustively reviewed over 100 articles on rodent and human phys- iological parameters. In our calculation, it is assumed that within a given species the ventilation and blood flow rates are proportional to the two- thirds power of body weight, and the volume of each tissue group is directly proportional to the body weight. For example, the alveolar ven- tilation rate for a 0.35-kg rat is calculated by the following equation: Ha = 0.075 x (o.35/0.3~2'3 = 0.083 liters/min, where 0.075 liter/min is the ventilation rate for 0.30-kg rat.

388 CHAD W. CHEN AND JERRY N. BLANCATO TABLE A-2 Physiological and Biochemical Parameters Used in the Model Parameters Rats Mice Humans Body weight (kg) 0.35 0.025 70 Alveolar ventilation rate Qa 0.083 0.028 7.500 Blood flow rate Qt 0. 104 0.019 6.200 Ql 0.0389 0.0048 1.550 Of 0.00920 0.0017 0.314 Qr 0. 0434 0.0097 2.765 Qp 0.0126 0.0029 1.570 Tissue volume V, 0.0140 0.0012 1.700 Vf 0.0315 0.0027 14.000 Vr 0.0150 0.0015 3.500 Vp 0.2550 0.022 43.400 Partition coefficient N 18.9 16.9 10.3 P. 3.719 4.159 3.719 Pf 108.994 121.893 108.994 Pr 3.719 4.159 3.719 Pp 1.058 1.183 1.058 Vm 0.00586 0.003 0.703 Km 2.9378 1.472 32.043 PARTITION COEFFICIENTS The parameters for rats and mice were provided by Curtis Travis of the Oak Ridge National Laboratory, which is the contractor for the EPA project "Interpretation of Metabolic Data in Exposure Analysis," which is sponsored by the Exposure Assessment Group, Office of Health and Environmental Assessment. It is our understanding that these parameters were obtained by Dr. Melvin E. Andersen of the Air Force Medical Research Laboratory at Wright-Patterson Air Force Base, Ohio, by using the equilibrium vial technique. The tissue:blood concentration ratio was calculated by taking the ratio of two concentration ratios, air/blood and air/homogenate tissue. The tissue:blood coefficients for humans are as- sumed to be identical to those of rats. METABOLISM RATE CONSTANTS, Vm AND Km Rats The constants Vm and Km were estimated by least-square optimization by using the system of differential equations and data from a metabolism

PK Modeling: Perchloroethylene 389 study by Pegg et al. (1979). As presented in Table 2 in the text, the amounts metabolized over 72 h for a 0.25-kg rat exposed to PCE at 10 and 600 ppm for 6 h were recorded to be 0.467 and 9.11 ma, respectively. These two data points (72 h, 0.467 mg and 72 h, 9.11 ma) were used to fit the system of differential equations, described previously, with Vm and Km as unknown parameters. The resultant estimates are Vm = 0.00468 mg/min and Km = 2.937 mg/liter. For the NTP (1985) bioassay rats, which had approximately 0.35 kg of body weight, the estimates are Vm = 0.00468 x (0.35/0.25~2'3 = 0.00586 mg/min, with Km being the same, independent of body weight. Mice As indicated in Tables 1 and 2 of the text, there are two metabolic data points: one from the Savage study and another from the inhalation study. The metabolic constants are estimated to be Vm = 0.003 mg/min and Km = 1.472 mg/liter by using two metabolic data points to optimize the system of differential equations. When PCE is given by Savage, the equa- tion involving liver tissue is given by: Vat d—t = Q~<Car' — CLIPS) — fir + Dke kit, (A-7) where D (in milligrams) is the administered dose by Savage, and k = 11 (mind) is estimated from Savage data from rats (Table 11. Varying the value of k within 1 order of magnitude does not appear to affect the predicted metabolism over 72 h. (Obviously, varying k would affect met- abolic time pattern over a very short period.) Humans The constants Vm and Km are estimated by using the urinary metabolites, as presented in Table 3 of the text, assuming that TCA in urine accounts for 30% of total metabolites. The value of 30% is selected because it fits the data in Table 4 better than other assumptions about the percentage of urinary TCA in total metabolites. The resultant estimates are Vm = 0.703 mg/min and Km = 32.043 mg/liter. The predicted (observed) alveolar concentrations for the three subjects (Table 4) are given in Table A-3. It is interesting to compare the metabolic constants calculated above and those calculated by other investigators. On the basis of the data of Ohitsu et al. (1983), Hattis et al. (1986) estimated 0.784 mmol/day of urinary metabolites after 8 in/day of occupation exposure. The Vma,` was estimated as follows:

390 CHAD W. CHEN AND JERRY N. BLANCATO 0.784 mmol/day man 24 in/day x 60 min/h = 5.00 X 10-4 mmol/min. (Am) Hattis et al. (1986) assumed that the urinary metabolites account for 60% of the total metabolites, and adjusted the V,~ accordingly to a value of 5.00 X 10-4 mmol/min/0.6 = 8.0 x 10-4 mmol/min, which can be converted to 0.13 mg/min by 8.0 x 10-4 mmol/min x 164 mg/mmol. In actual risk calculation, Hattis et al. (1986) adjusted the Vma,~, for better fit, to 9.07 x 10-4 mmol/min, which would be about one-fifth the Vim estimated in this paper. We, however, propose two changes from the method used by Hattis et al. (19861. 1. As stipulated by Gehring et al. (1979) for rapidly metabolized com- pounds, the metabolic period should be 8 h (the exposed penod) and not 24 h. 2. Our calculations indicate that human urinary metabolites account for only 30~o rather than 60% of the total metabolites. Taking the value of 0.784 mmol/day of urinary output and calculating in the same manner as Hattis et al. (1986), but with the two changes discussed, the Vm`= = 0.869 mg/min, a value very close to our V,,~ of 0.7 mg/min. As for the value of Km' Hattis et al. (1986) calculated, on the basis of the data of Ohitsu et al. (1983), a value of 3.4 x 10-4 mmol/liter of blood. Using the value of Vma,` = 0.131 mg/min, the ratio of V,t,~,` to Km is 0.0235. Assuming that this ratio is correct, and using the corrected Vma,` of 0.869 mg/min, the Km would be equal to 37 mg/liter of blood, a value nearly equal to our value of 32 mg/liter of blood. TABLE A-3 Predicted (Observed) Alveolar Concentrations for Three Subjects 8-Hour Predicted (Observed) Alveolar Concentration at the Following Postexposure Concentration Periods: (mg/liter) 0.5 h 1 h 2 h 16 h 64 h 0.686 0.24(0.14) 0.17 (0.10) 0.07 (0.08) 0.01 (0.03) 0.008 (0.010) 1.029 0.37 (0.28) 0.26 (0.22) 0.10 (0.17) 0.02 (0.04) 0.01 (0.02) 1.371 0.49 (0.34) 0.34 (0.26) 0.14 (0.19) 0.02 (0.05) 0.02 (0.03)

Development of Multispecies, Multiroute Pharmacokinetic Models for Methylene Chloride and I, I, I-TrichIoroethane (Methyl Chloroforms Richard H. Reitz, Richard ]. Nolan, and Alan M. Schumann INTRODUCTION Chronic toxicity tests in rodents are typically conducted with homo- geneous populations of young, healthy animals. For practical reasons, only small numbers of animals and limited numbers of doses can be evaluated. Consequently, doses are usually set as high as possible, in the hope of providing maximum sensitivity for the test. The animals used in these tests develop large numbers of certain types of tumors in the absence of any chemical treatment (e.g., liver tumors in B6C3F1 mice, leukemia in Fischer 344 rats), and have a normal life span of approximately 2 years. The test agent is usually administered by the most experimentally convenient route. When these studies are completed, they are used to estimate the risk that tumors will be produced in human populations exposed to the same agents. In most cases, large numbers of heterogeneous humans of varying ages and health status are involved. The human exposures are usually orders of magnitude lower than the doses employed in the animal studies. The average lifetime of the human populations is much greater than that of the rodent species, and the spontaneous incidence of specific tumor types is usually much lower. Furthermore, the primary routers) of exposure in human populations may differ from the route used in the animal study. Furthermore, there appear to be at least two general types of chemical agents that are capable of affecting the tumor incidences in chronic bioas- says: those interacting directly with genetic material (genotoxic agents), 391

392 RICHARD H. RElTZ ET AL. and those which do not interact directly with genetic material (nongeno- toxic agents). There are theoretical reasons for believing that the dose dependency of the two types of agents may be quite different. Considering all these factors, it is clear that risk estimations based upon animal data contain large areas of uncertainty. Physiologically based phar- macokinetic (PB-PK) analysis offers us a method for eliminating some of the uncertainty in these risk estimates, and may also provide a tool for testing our theories of chemical carcinogenesis. PB-PK analysis cannot, however, eliminate all sources of uncertainty from risk estimation, and for optimum use, PB-PK analysis must be integrated with other disciplines. In the examples that follow, it is shown how the techniques of PB-PK are used to aid in estimating the internal doses associated with exposures to methylene chloride (MeCl2), methylchloroform (MC), and perchloroe- thylene (tetrachloroethylene; PERC). Risk estimations based on internal dose at the target organ are considerably more reliable than risk estimations that fail to consider such factors as nonlinear metabolic processes and physiological differences between species. ROLE OF METABOLISM (MeCI2) Bioassay Results MeCl2 is a solvent with a variety of uses, including paint removal, metal decreasing, aerosol applications, and food processing. Several stud- ies of the chronic toxicity of MeCl2 have been conducted. Inhalation studies in the Syrian golden hamster, exposed to concentrations up to 3,500 ppm for 6 in/day, did not show a tumorigenic response at any site (Burek et al., 19841. Similarly, a drinking water study sponsored by the National Coffee Association failed to reveal a dose-related statistically significant increase in tumors in the B6C3F1 hybrid mouse strain (Serota et al., 1984a) or the Fischer 344 rat (Serota et al., 1984b). Two inhalation studies of MeCl2 in the Sprague-Dawley rat (at doses up to 3,500 and 500 ppm, respectively) revealed treatment-related effects on the number of spontaneously occurring benign mammary tumors per tumor-bearing rat. Low incidences of tumors in the ventral neck region in and around the salivary gland of male rats were noted in one study (Burek et al., 1984) but were not noted in the other two rat inhalation studies (Burek et al., 1984; NTP, 19851. Significant increases in spontaneously occurring lung and liver tumors were noted when B6C3F1 mice were exposed to 2,000 or 4,000 ppm of MeCl2. In contrast to the studies mentioned above, in which MeCl2 either failed to affect the tumor incidence or affected primarily benign tumors, the incidence of both benign and malignant tumors was elevated in this study (NTP, 19851.

Multispecies, Multiroute PK Models 393 Obviously, it would be useful to understand the reason for the different results obtained in the different studies. PB-PK analysis can be used in conjunction with biochemical studies to provide a single hypothesis that is consistent with all of these results. Electrophilic Intermediates Miller and Miller (1966) noted that most of the potent chemical car- cinogens either were directly electrophilic or could be converted to elec- trophilic species by metabolic activation. Miller and Miller hypothesized that these electrophilic species would be capable of covalently binding to DNA and inducing mutations and/or cancer. The generation of reactive inte~ediates also appears to play a role in the tumorigenicity of agents which do not react directly with DNA. For example, PERC is inactive in almost all of the short-term tests for gen- otoxicity (direct reaction with DNA), but the toxicity/carcinogenicity of PERC appears to be strongly correlated with the rate of metabolic acti- vation (Buben and O'Flaherty, 1985; Schumann, 1984; Schumann et al., 19801. The chemical reactivity of MeCl2 itself is very low, and it is unlikely that it undergoes any direct reaction with macromolecules. Therefore, it appears likely that metabolism would play a role in the toxicity/carcino- genicity of MeCl2. Metabolic Pathways MeCl2 is metabolized via two pathways: (1) an oxidative pathway (Ku- bic et al., 1974) that appears to yield CO as well as considerable amounts of CO2 (Gargas et al., 1986, Reitz et al., 1986) and (2) a glutathione- dependent pathway (Ahmed and Anders, 1978) that produces CO2 but no CO (Figure 11. Both pathways release 2 mol of halide ion per mol of dihalomethane consumed. The oxidative pathway (MFO) is readily sat- urated at concentrations of a few hundred parts per million, but the glu- tathione-S-transferase (GSH) pathway showed no indication of saturation at inhaled concentrations up to 10,000 ppm (Gargas et al., 19861. Reactive, potentially toxic intermediates are formed in both pathways (Figure 1~. The question is which of these pathway~s) is related to the in vivo tu- morigenic activity of MeCl2? There are several reasons for believing that metabolism by MFO is not involved in the tumorigenicity of MeCl2. First, it is clear that the incidence of most of the treatment-related tumors increases as the exposure con- centration is raised from 2,OOO to 4,000 ppm (Table 11. However, it has been demonstrated that HbCO, a major product of MFO activity, does

394 RICHARD H. RElTZ ET AL. CYT P 450 1 H 0- CH2 C~ ' ~ H-O-C-CL NADPH I | CYTOSOL ~ GSH | ~ _ HCL GS- CH~CL HCL 1 GS- CH2-OH NAD. GS-C ~ ~ OH HCOOH ~ GSH CL CHiO ~ GSH co2 o ll - C · :C = 0 | HCL POSH ~0 G-S-Cx H H GS C ~ OH co2 H. GSH co2 FIGURE 1 Proposed metabolic pathways for methylene chloride metabolism (based on Ahmed and Anders, 1978, and Kubic et al., 1974). Potentially reactive intermediates are formed in each pathway: forrnyl chloride in the cytochrome P-450 (MFO) pathway, and chloromethyl glutathione in the cytosolic (GSH) pathway. Either metabolic pathway can produce carbon dioxide in this scheme, but only the MFO pathway yields carbon monoxide and carboxyhemoglobin. not increase in rats or hamsters as exposure levels of MeCl2 are raised above 500 ppm (Burek et al., 1984; McKenna et al., 19821. In addition, Green (1983) has separated enzymes from the MFO and GSH pathways and studied their ability to metabolically activate MeCl2 in vitro. He found that cytosolic preparations from rat liver (containing GSH transferase) significantly increased the yield of bacterial mutations in Salmonella typhimurium exposed to DCM. However, rat liver micro- somes rich in MFO failed to show this effect. We have developed a PB-PK model (Andersen et al., 1987) that allowed us to investigate the question further by making direct comparisons be- tween a National Coffee Association drinking water study in B6C3F' mice and the NTP inhalation study in B6C3F1 mice; The model allowed

Multispecies, Multiroute PK Models 395 TABLE 1 Comparison of Average Daily Values of Dose Surrogatesa and Tumor Incidences in Lung and Liver Tissues of Female B6C3F1 Mice in Two Chronic Bioassays of Methylene Chloride Inhaled Inhaled Drink Tissue Control (2,000 ppm) (4,000 ppm) (250 mg/kg/day) Liver MFO Pathway 0.0 3,575.0 3,701.0 5,197.0 GSH Pathway 0.0 851.0 1,811.0 15.1 Tumors (%) 6.0 33.0 83.0 6.0 Lung MFO Pathway 0.0 1,531.0 1,583.0 1,227.0 GSH Pathway 0.0 123.0 256.0 1.0 Tumors (Jo) 6.0 63.0 85.0 8.0 arose surrogate is calculated as the average amount of metabolite (milligram equivalents of MeCl2) formed by each pathway per day per liter of organ. us to calculate the levels of reactive intermediates formed by each pathway in the various tissues. Because these intermediates were presumed to be too reactive to leave the site of their formation, the average amounts of metabolite formed per day per liter of tissue have been calculated as dose surrogates for each organ. The model predictions for these dose surrogates are summarized in Table 1. Reference to Table 1 indicates that the dose surrogates for the MFO pathway in the liver or lung tissue of B6C3F1 mice are virtually identical in each dose of the inhalation as well as the drinking water study. This is inconsistent with the observed tumor incidences. Tumors in the inhal- ation study increased with dose and were significantly different than the control, but tumor incidences in the drinking water study were not sig- nificantly different than those in controls (Table 11. Tumor incidences were consistent with the dose surrogates produced by the GSH pathway, however, supporting the hypothesis that the tu- morigenicity of MeCl2 is related to the production of GSH metabolites in the target organ. This conclusion has significant implications for the es- timation of human risk. For example, the model indicates that exposure to high concentrations of MeCl2 (which saturate the MFO system) produces disproportionately high levels of the GSH dose surrogates. In addition, the model suggests that when humans and rodents are exposed to equiv- alent concentrations of MeCl2 in the atmosphere or drinking water, the rodents produce higher levels of the GSH dose surrogates because of the relatively higher rates of metabolism in the rodent species.

396 RICHARD H. RElTZ ET AL. Future Research Another important outcome of the PB-PK analysis is that it suggests important directions for future research. For example, it is obvious that the values chosen for the human metabolic rate constants affect the levels of the dose surrogates calculated by the model. The rate constants for the MFO pathway in humans were estimated from existing in vivo human data as described elsewhere (Andersen et al., 19871. Because in vivo determinations of the activity of the GSH pathway in humans would involve exposing humans to high concentrations of MeCl2, however, the in vivo activity of the GSH pathway in humans was estimated by allometric scaling of the rodent data. Use of allometric scaling is supported by the observation that allometric scaling of the GSH pathway worked reasonably well in the three rodent species; intrinsic clearances (scaled to body weight to the 0.3 power) for this pathway were nearly identical in mice, hamsters, and rats (Andersen et al., 19871. Nevertheless, measurement of in vitro GSH activities (with MeCl2 as the substrate) would be very useful in verifying the predictions of this model. We have subsequently conducted such studies, and they are in reasonable agreement with the predictions of this model (Reitz et al., manuscript in preparation, 19871. EXTRAPOLATION B ETWEEN SPECI ES (MC) 1, 1,1-Trichloroethane (methylchlorofo~, MC) is a volatile chlorinated hydrocarbon. It is widely used as a metal-degreasing solvent, and is also present in commercial aerosols and adhesive products. Several studies of the absorption, distribution, and elimination of MC have been conducted in humans as well as different species of laboratory animals (Nolan et al., 1984; Monster et al., 1979, Schumann et al., 1982a,b). However, a comprehensive model capable of describing a variety of pharmacokinetic endpoints in multiple species after administration by different routes has not been reported. We have developed a PB-PK model for MC based on that of Ramsey and Andersen (1984~. Metabolic parameters for the rat were determined from the ~4C balance study of Schumann et al. (1982a), and metabolic parameters for the B6C3F1 mice and human volunteers were calculated by scaling body weight to the 0.7 power as described by Ramsey and Andersen (19841. Partition coefficients were determined by M. Gargas (Armstrong Aerospace Medical Research Laboratory, Wright-Patterson Air Force Base, Ohio) using a vial equilibration technique similar to that of Sato and Nakajima (19791. Physiological parameters were taken from

Multispecies, Multiroute PK Models 397 the scientific literature (Caster et al., 1956; Davis and Mapleson, 1981; ICRP, 19751. Physiological and biochemical parameters used in the model are summarized in Table 2. The objective of these studies was to see whether a PB-PK model developed from one species could be employed to predict data gathered in other species. The PB-PK model was used to predict time course data for MC in venous blood for all species. In addition, several other types of data were collected either from the }4C balance studies in animals or from clinical pharmacokinetic studies in human volunteers. The additional parameters evaluated included (1) time/course data for MC in expired air, (2) end- exposure body burdens, (3) MC concentrations in fat, (4) MC concentra- TABLE 2 Parameters Used in the Physiologically Based Pharrnacokinetic Model for Methylchloroform Parameter Human Rat 1 Rat 2 Mouse 1 Mouse 2 Body weight (kg) 83.0 0.215 0.468 0.029 0.038 Percentage of Body Weight Liver 2.6 4.0 4.0 4.0 4.0 Rapidly perfused 6.4 5.0 5.0 5.0 5.0 Slowly perfused 63.5 76.0 54.7 79.0 67.0 Fat 19.5 7.0 28.3 4.0 16.0 Flow (liters/in) Alveolar ventilation 330.7 5.11 8.81 1.26 Cardiac output 330.7 5.11 8.81 1.26 Percentage of Cardiac Output Liver 24 24 24 24 24 Rapidly perfused 53 53 53 53 53 Slowly perfused 14 18 18 21 21 Fat 9 5 5 2 2 Partition coefficients Blood/air 2.53a 5.76 5.76 l0.8a l0.8a Liver/blood 3.40 1.49 1.49 0.796 0.796 Rapidly perfused/blood 3.40 1.49 1.49 0.796 0.796 Slowly perfused/blood 1.25 0.547 0.547 0.292 0.292 Fat/blood 104.0 45.7 45.7 24.4 24.4 Metabolic constants Vmax (mg/h) 5.84 0.0904 0.156 0.0222 0.0269 K,n (mg/liter) 6.43 6.43 6.43 6.43 6.43 NOTE: Abbreviations: Rat 1, young rats (2-4 months); Rat 2, old rats (18.5 months); Mouse 1, young mice; Mouse 2, old mice. Parameters used for drinking water simulation, Savage, and intravenous administration in rats are very close to those for Rat 1 and are not shown. aValue of partition coefficient from M. L. Gargas and M. E. Andersen (personal communication, 1985).

3963 RICHARD H. RElTZ ET AL. lions in liver, and (5) total amounts of MC metabolized. The model predictions were tested by comparison with actual data sets for humans (Nolan et al., 1984), rats, and mice (Schumann et al., 1982a). Rats The first parameter was the time course of the venous blood concen- trations of MC in rats after exposure to 150 or 1,500 ppm of MC. Ex- perimental data collected by Schumann et al. (1982a) were used to check the predictions. The data showed good agreement with the model predic- tions over the period studied (Figure 21. Comparisons between the model predictions and actual data for several other parameters are summarized in Table 3. The model was reasonably successful in predicting values for these additional parameters. The ratio of predicted to actual data ranged from a low of 0.73 to a high of 1.89, end the mean ratio was 1.17 + 0.44for l50ppm end 1.28 + 0.38 for 1,500 ppm (Table 3~. loo 10 . - ~ - 1 0.1 o o · me\ em' O calm 1500 ppm \ o Cot o .^ ,. 150 ppm 3 4 5 6 8 9 10 TIME (h) FIGURE 2 Simulated and observed values of venous blood concentration in Fischer 344 male rats during and following a 6-h inhalation exposure to 150 or 1,500 ppm of methyl chloroform (MC). Computer predictions are shown as solid lines, while actual data are given as open circles (1,500 ppm) or closed circles (150 ppm). Data are milligram equivalents of MC/liter of blood.

Multispecies, Multiroute PK Models 399 TABLE 3 Comparison of Values for Selected Parameters in Young Rats and Young Mice as Predicted by the Models and Observed by Schumann et al. (1982a) Parameter Model Prediction Actual Data Ratioa Young rat (150-ppm inhalation exposure) End expt. blood level (mg/liter) Body burden at 6 h (Emil) Amount metabolized (wool) Conc. in fate (nrnol/g) Conc. in livers (nmol/g) Young rat (1,500-ppm inhalation exposure) End expt. blood level (mg/liter) Body burden at 6 h (1lmol) Amount metabolized (Emil) Conc. in fat (nmol/g) Conc. in liver (nmol/g) Young mouse (150-ppm inhalation exposure) End expt. blood level (mg/liter) Body burden at 6 h (Emil) Amount metabolized (wool) Conc. in fat (nmol/g) Conc. in liver (nmol/g) Young mouse (1,500-ppm inhalation exposure) End expt. blood level (mg/liter) Body burden at 6 h (1lmol) Amount metabolized (Emil) Conc. in fat (nmol/g) Conc. in liver (nmol/g) 4.43 25.10 1.90 1,304.00 49.60 44.70 241.30 5.62 13,126.00 502.00 8.70 2.96 0.62 1,575.00 51.70 87.90 25.20 1.23 15,915.00 525.00 2.81 33.03 1.90 724.00 68.20 23.7 264.28 5.53 8,403.00 504.00 9.27 4.97 0.65 1,329.00 76.00 111.60 39.90 1.19 16,198.00 631.00 1.58 0.76 1.00 1.80 0.73 1.89 0.91 1.02 1.56 1.00 0.94 0.60 0.95 1.19 0.68 0.79 0.63 1.03 0.98 0.83 aThe ratio is equal to the model prediction column divided by the actual data column. Concentrations in fat and liver were those observed at the end of the 6-h exposure. Mice The model was then used to describe the pharmacokinetic behavior of MC after inhalation exposure in male B6C3F1 mice. To convert from rat simulations to mouse simulations, the appropriate physiological and bio- chemical parameters were adjusted as outlined in Table 2. Metabolic constants for the mouse were obtained by setting the Michaelis constant (Km) for the mouse equal to that of the rat, and scaling the value of Vmax to the 0.7 power of body weight. The size of the body fat compartment

400 RICHARD H. REITZ ET AL. in the young mice was reduced from 7% of body weight to 4% of body weight, and the percentage of cardiac output directed to the fat tissue was reduced from 5% to 2%. The time course of MC in venous blood (collected at the orbital sinus) was accurately described by the model for two different exposure con- centrations: 150 and 1,500 ppm (Figure 31. The physiologically based model predicted that MC would be eliminated from the mouse much more rapidly than from the rat. This is consistent with the data of Schumann et al. (1982a), who reported that initial elim- ination half-lives for MC in mice were 5- to 10-fold less than the corre- sponding half-lives in rats. As outlined for rats, five additional parameters simulated by the model were checked against experimental data collected by Schumann et al. (1982a) in the mouse. The model was also reasonably successful in de- scribing these data (Table 3~. The ratio of predicted to actual data ranged from a low of 0.60 to a high of 1.19, and the mean ratio was 0.87 + 0.21 for 150 ppm and 0.85 + 0.14 for 1,500 ppm (Table 31. 1000 100 ~0 ._ e {53 E 1 0.1 F _ - _ ctcO 1500 ppm Hi o -of - r O ; 0 0 ~ 0 ~ 150 ppm . . . . . 4 5 6 7 8 9 10 TIME (h) FIGURE 3 Simulated and observed values of venous blood concentration in B6C3F1 male mice during and following a 6-h inhalation exposure to 150 or 1,500 ppm of methyl chloroform (MC). Computer predictions are shown as solid lines, while actual data are given as open circles (1,500 ppm) or closed circles (150 ppm). Data are milligram equivalents of MC/liter of blood.

Multispecies, Multiroute PK Models 401 from a low of 0.60 to a high of 1.19, and the mean ratio was 0.87 + 0.21 for 150 ppm and 0.85 + 0.14 for 1,500 ppm (Table 31. Human Data After adjustment of model parameters, as described above for mice, the time courses of MC in expired air (Figure 4a) and venous blood (Figure 4b) during the following human exposures to 350 or 35 ppm MC were simulated with the model. Elimination of MC was triphasic in nature, and the model clearly reflected this (Figures 4a and 4b). The model predicted that expired air and venous blood concentrations should be proportional to exposure concentration throughout the range of 35-350 ppm MC. These predictions were also in good agreement with the experimental data of Nolan et al. (1984) (Figures 4a and 4b). The model was then used to estimate the amount of MC metabolized by humans during the two exposures. The model predicted that 2.47 mg equivalents of MC would be metabolized during the 240 h after a 6-h exposure to 35 ppm of MC, and 4.32 mg equivalents of MC metabolites were actually recovered (Nolan et al., 19841. Similarly, the model pre- dicted that 18.6 mg equivalents of MC metabolites would be formed during the 240 h following a 6-h exposure to 350 ppm of MC, and 32.9 mg equivalents were actually recovered. Was the Simulation Successful? In this exercise, a single basic PB-PK model has been used to predict a variety of pharmacokinetic data from three different species. Only those parameters known to differ between species or measured in the laboratory were varied between species. In general, the model predictions and animal data were within a factor of 2, and often they were substantially closer. In view of the fact that apparently identical animal studies of toxicity and/ or carcinogenicity often differ between themselves by more than a factor of 2, the consistency of the PB-PK model seems remarkable. Furthermore, the fact that this consistency extends across a variety of endpoints, in- cluding end-exposure blood levels, total amounts metabolized, and tissue concentrations, suggests that this technique offers considerable promise in understanding and predicting interspecies differences in the delivery of internal dose to various organs. MODELING DRINKING WATER EXPOSURES (MC, PERC) Halogenated hydrocarbons such as MeCl2, MC, and PERC have limited solubilities in water. Consequently, the doses of these materals that can

402 RICHARD H. RElTZ ET AL. ~ (a} 1 0.1 1~° 0.00 0.000 0.0000 10 1 ~o~ o o 350 ppm 35 ppm 0 50 100 150 200 250 TIME (h) (b) . . . 0.0 0.00 0.1 - __- ~ ~ —^_ C; _ _ ; ~ 350 ppm 0 50 10 150 200 250 TIME (h) FIGURE 4 (a) Simulated and observed values of exhaled air concentration in human volunteers during and following a 6-h inhalation exposure to 35 or 350 ppm of methyl chloroform (MC). Computer predictions are shown as solid lines, while actual data are given as open circles (350 ppm) or closed circles (35 ppm). Data are milligram equivalents of MC/liter of air. (b) Simulated and observed values of venous blood concentration in human volunteers during and following a 6-h inhalation exposure to 35 or 350 ppm of methyl chloroform (MC). Computer predictions are shown as solid lines, while actual data are given as open circles (350 ppm) or closed circles (35 ppm). Data are milligram equivalents of MC/liter of air.

Multispecies, Multiroute PK Models 403 be administered to animals in their drinking water are limited. Because small quantities of these materials are sometimes found in human drinking water supplies, however, it would be useful to have a procedure for predicting their toxicity from studies conducted by another route. We have employed the PB-PK model described above to simulate drinking water exposure, and have collected }4C disposition data for MC to check the model predictions. In the first experiment, animals were given a bolus Savage with water containing dissolved MC. At various intervals, samples of venous blood were withdrawn and analyzed for MC, and the results are shown in Figure 5a. The model gave a reasonable simulation of the time course of MC absorbed from water in the gastrointestinal tract. In a second experiment, animals were presented with water containing radioactive MC (saturated solution, 4.4 mg/ml) and allowed to drink ad libitum. Radioactive drinking water was available for 8 h. Samples of urine, expired air, and selected tissues were analyzed for radioactivity at various times thereafter. The average water consumption of the rats (mean weight, 250 g) was 8.1 + 3.8 ml, corresponding to an average dose of 143 mg/kg. Overall recovery of radioactivity in these experiments (based on water consumption) was 95.2 + 4.33%. Most of the radioactivity was recovered on the charcoal traps within the first 12 h of the experiment, with small amounts of radioactivity also seen in the urine, CO2, carcass, and skin fractions. For the purpose of simulation, it was assumed that rats ingested MC in their drinking water at a constant rate and that absorption of MC occurred rapidly. Input of MC was set equal to the average rate of intake (4.47 mg of MC/h) for the entire period in which the animals had access to treated water. It was assumed that all of the MC ingested by the rats entered the liver compartment directly, and the differential equations were written to reflect these assumptions. These animals were slightly larger than the young rats used in the inhalation studies; and the body weight, cardiac output, alveolar ventilation rate, and metabolic velocity were scaled as outlined previously. Values used in the model are listed in Table 2. Other than this, the PB-PK model was not changed for the simulation of drinking water exposure. The average rate of elimination of MC (milligrams of MC/hour) during the interval was calculated and plotted at the midpoint of the excretion interval. Results are shown in Figure Sb. The time course of t~4C]MC in expired air was reasonably well described for the first 24 h of exposure. The model predicted that 3.58 wool of MC would be metabolized, and 8.19 Wool (3.0% of the administered radioactivity) was actually recovered in urine and CO2 (Table 41. The model also predicted that 97% of the total ingested radioactivity would be eliminated within 12 h after the treated

404 RICHARD H. RElTZ ET AL. 10.0 1.0 0.1 10.000 1.000 - 0.100 - 0.010 0.001 (a) 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 TIME (h) o 5 10 15 20 25 TIME (h) FIGURE 5 (a) Simulated and observed values of venous blood concentration in Fischer 344 male rats following gavage with a water solution of MC (14.2 mg/kg). Computer predictions are shown as a solid line, while actual data are given as closed circles. Data are milligram equivalents of MC/liter of blood. (b) Simulated and observed values of exhaled air concentration in Fischer 344 male rats during and following drinking water exposure to a saturated solution of MC (8-h exposure; dose, 143 mg/kg). Computer predictions are shown as a solid line, while actual data are given as closed circles. Data are milligram equivalents of MC/liter of air.

Multispecies, Multiroute PK Models 405 TABLE 4 Comparison of Values for Selected Parameters for Disposition of MC in Old (approximately 18.5 months) Rats and Old Mice as Predicted by the Models and Observed by Schumann et al. (1982b) Model Actual Parameter Prediction Data Ratioa Old rat (18.5 months; 1,500-ppm inhalation exposure) Body burden at 6 h (wool) 669.5 724.3 0.92 Amount metabolized Remold 13.8 18.8 0.73 Conc. in fatb (nmol/g) 4,604.0 5,685.0 0.81 Conc. in livers (nmol/g) 440.4 606.0 0.73 Charcoal (Amos) - 0-3 h 142.7 185.3 0.77 3-6 h 84.4 128.2 0.66 6-12 h 125.7 138.3 0.91 12-24 h 152.0 128.8 1.18 24-36 h 76.5 63.8 1.20 Old mouse (18.5 months; 1,500-ppm inhalation exposure) Body burden at 6 h (Errol) 68.6 Amount metabolized (Errol) 2.21 Conc. in fat (nmol/g) 10,217.0 Conc. in liver (nmol/g) 491.3 Charcoal (Errol) 0-3 h 3-6 h 6-12 h 12-24 h 24-36 h Young rat (drinking water exposure, 143 mg/kg) Amount metabolized (Errol) Percentage of dose eliminated 0-12 h 29.2 14.9 14.3 6.99 0.91 3.58 97.0 141.0 8.04 10,603.0 1,794.0 93.48 26.5 11.0 1.80 0.49 0.27 0.96 0.27 0.31 0.56 1.30 3.88 6.07 8.20 0.44 91.2 1.06 aThe ratio is equal to the model prediction column divided by the actual data column. Concentrations in fat and liver were those observed at the end of the 6-h exposure. water was withdrawn. Actual data show that 91.2 + 3.6% of the ingested radioactivity was recovered during this period. These data are in fair agreement with the model prediction, considering the difficulties associ- ated with administration of MC through this route. Frantz and Watanabe (1983) reported a detailed comparison of the fate of t~4C]PERC administered in drinking water and via inhalation. Although their data were not analyzed with a PB-PK model, they reported that the elimination kinetics of PERC were consistent with results generated by

406 RICHARD H. REITZ ET AL. both Savage and inhalation exposures, and that the fate of PERC was not substantially different from the disposition resulting from other routes of administration. These data offer encouragement that pharmacokinetic models may play an important role in understanding and predicting the toxicity of materials found in drinking water based on studies conducted by another route. MODELING INTRAVENOUS INJECTION EXPOSURES (MC) Another common route of exposure is intravenous injection. To see if the PB-PK model for MC was capable of describing this different route, groups of six rats were injected with 8.8 or 47.0 mg/kg of MC. Samples of venous blood were drawn at various times and analyzed for MC. For modeling purposes, injection of MC was described as a rapid in- fusion into pooled venous blood. No other changes were made to model parameters. The disposition of MC administered by this route was well described by the standard PB-PK model (Figure 61. Booboo 10.00 1.00 0.10 0.01 o~ ,~;— °~0—47 mg/kg \- 8.8 mg/kg ........ . 0 0.5 1 1.5 2 2.5 3 3.5 4 TIME (h) FIGURE 6 Simulated and observed values of venous blood concentrations in Fischer 344 male rats following intravenous injection of MC (dose rates, 8.8 and 47 mg/kg). Computer predictions are shown as a solid line, while actual data are given as closed circles. Data are milligram equivalents of MC/liter of venous blood.

Multispecies, Multiroute PK Models 407 PHARMACOKINETIC STUDIES IN OLD ANIMALS Most pharmacokinetic studies are conducted with young adult animals, while chronic toxicology studies are conducted with animals ranging in age from very young at the beginning of an experiment to very old. We have used the PB-PK model developed for MC to determine whether we could develop reasonable hypotheses to predict the effect of aging upon disposition of this material. Old Rats Schumann et al. (1982b) studied the pharmacokinetics of inhaled MC in old male rats from a chronic inhalation toxicology study. The age of these animals was approximately 18.5 months, and the average body weight was 468 g. They noted that the older animals appeared to have more fat tissue than young rats, and also observed several differences between the disposition of MC in young and old animals. Data from the old rats exposed to 1,500 ppm of MC were not particularly well simulated by the standard pharmacokinetic model for young rats (simulation not shown). Other investigators have noted this phenomenon, and have found it necessary to increase the relative percentage of fat when constructing pharmacokinetic models in older animals (Lutz et al., 19771. Conse- quently, the fat compartment was increased from 7.08% to 28.3% of total body weight, and the percentages of body weight assigned to the non- perfused and slowly perfused tissue compartments were reduced to com- pensate for this (Table 2~. Otherwise, the model was constructed exactly as outlined for young rats. When the simulation was rerun under these conditions, the model described the data reasonably well. Nine parameters were chosen for comparison of model simulations and actual data. These included the end-exposure body burden, the amount of MC metabolized, the concentration of MC in fat and liver tissue at the end of the exposure, and the amount of radioactivity (presumed to be MC) recovered on an activated charcoal trap at 0-3, 3-6, 6-12, 12-24, and 24-36 h postexposure. The results are summarized in Table 4. The ratio of predicted values to actual data varied from a low of 0.66 to a high of 1.20, and the mean ratio was 0.88 + 0.18. Old Mice Schumann et al. (1982b) also collected data in geriatric male mice of similar ages (approximately 18.5 months). The model for young mice was modified by increasing the relative fat content from 4% of body weight

408 RICHARD H. RElTZ ET AL. to 16% of body weight, and the disposition of MC in old animals was simulated after this change. The data were not particularly well simulated by the model. The ratio of predicted values to actual data varied from a low of 0.27 to a high of 6.07, and the mean ration was 1.57 + 1.92 (Table 41. It appears that age-related changes in the disposition of MC in geriatric mice may involve additional factors beyond the increase in body fat. Further research will be needed to elucidate these factors. Summary In summary, unified PB-PK models encompassing data from a variety of sources were developed for several of the organohalides. These models can be used to strengthen the scientific basis of risk assessment, improve experimental design of chronic studies, and give insight into mechanisms of toxicity and metabolism. Many uncertainties certainly remain in the risk assessment process, but risk assessments that properly consider the role of physiologically based pharrnacokinetics should be significantly more reliable than those that do not. REFERENCES Ahmed, A. E., and M. W. Anders. 1978. Metabolism of dihalomethanes to formaldehyde and inorganic halide. I. In vitro studies. Drug Metab. Dispos. 4:357-361. Andersen, M. E., H. J. Clewell, 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. Buben, J. A., and E. J. O'Flaherty. 1985. Delineation of the role of metabolism in the hepatotoxicity of trichloroethylene and perchloroethylene: A dose-effect study. Toxicol. Appl. Pharm. 78: 105- 122. Burek, J. E., K. D. Nitschke, T. J. Bell, D. L. Waskerle, R. C. Childs, J. D. Beyer, D. A. Dittenber, L. W. Rampy, and M. J. McKenna. 1984. Methylene chloride: A two- year inhalation toxicity and oncogenicity study in rats and hamsters. Fund Appl. Toxicol. 4:30-47. Caster, W. O., J. Poncelet, A. B. Simon, and W. D. Armstrong. 1956. Tissue weights of the rat. I. Normal values determined by dissection and chemical methods. Proc. Soc. Exp. Biol. Med. 91:122-126. Davis, N. R., and W. W. Mapleson. 1981. Structure and quantification of a physiological model of the distribution of injected agents and inhaled anaesthetics. Br. J. Anaesth. 53:399-404. Frantz, S. W., and P. G. Watanabe. 1983. Tetrachloroethylene: Balance and tissue dis- tribution in male Sprague-Dawley rats by drinking water administration. Toxicol. Appl. Pharmacol. 69:66-72. Gargas, M. L., H. J. Clewell, and M. E. Andersen. 1986. Metabolism of inhaled dihal- omethanes in vivo: Differentiation of kinetic constants for two independent pathways. Toxicol. Appl. Pharmacol. 82:211-223.

Multispecies, Multiroute PK Models 409 Green, T. 1983. The metabolic activation of dichloromethane and chloro-fluoromethane in a bacterial mutation assay using Salmonella typhimurium. Mutat. Res. 118:277-288. ICRP (International Commission on Radiation Protection). 1975. Report of the Task Group on Reference Man, ICRP Publication 23, W. S. Snyder, M. J. Cook, E. S. Nasset, L. R. Rauhausen, G. P. Howells, and I. H. Tiptom, eds. New York: Pergamon Press. Kubic, V. L., M. W. Anders, R. R. Engel, C. H. Barlow, and W. S. Caughey. 1974. Metabolism of dihalomethanes to carbon monoxide. I. In vivo studies. Drug. Metab. Dispos. 2:53-57. Lutz, R. J., R. L. Dedrick, H. B. Matthews, T. E. Eling, and M. W. Anderson. 1977. A preliminary pharmacokinetic model for several chlorinated biphenyls in the rat. Drug Metab. Dispos. 5:386-396. McKenna, M. J., J. A. Zempel, and W. H. Braun. 1982. The pharmacokinetics of inhaled methylene chloride in rats. Toxicol. Appl. Pharmacol. 65:1-10. Miller, E. C., and J. A. Miller. 1966. Mechanisms of chemical carcinogenesis-genesis: Nature of proximate carcinogens and interactions with macro-molecules. Pharmacol. Rev. 18-805. Monster, A. C., G. Boersma, and H. Steenweg. 1979. Kinetics of 1, 1,1,-trichloroethane in volunteers: Influence of exposure concentration and work load. Int. Arch. Occup. Environ. Health 42:293-302. NTP (National Toxicology Program). 1985. NTP Technical Report on the Toxicology- Toxicology and Carcinogenesis Studies of Dichloromethane in F-344/N Rats and B6C3F1 Mice (Inhalation Studies). NTP-TR-306 (Board Draft). Nolan, R. J., N. L. Freshour, D. L. Rick, L. P. McCarty, and J. H. Saunders. 1984. Kinetics and metabolism of inhaled methylchloroform (1, 1,1-trichloroethane) in male volunteers. Fund. Appl. Toxicol. 4:654-662. Ramsey, J. R., and M. E. Andersen. 1984. A physiologically based de-description of the inhalation pharmacokinetics of styrene in rats and humans. Toxicol. Appl. Pharmacol. 73:159-175. Reitz, R. H., F. A. Smith, and M. E. Andersen. 1986. In vivo metabolism of '4C-methylene chloride (MEC). The Toxicologist 6:260. Sato, A., and T. Nakajima. 1979. Partition coefficients of some aromatic hydrocarbon and ketones in water, blood, and oil. Br. J. Ind. Med. 36:231-234. Schumann, A. M. 1984. Inhalation kinetics. Food Solvents Workshop I: Methylene Chlo- ride, Bethesda, Md. Washington, D.C.: Nutrition Foundation. Schumann, A. M., J. F. Quast, and P. G. Watanabe. 1980. The pharmacokinetics and macromolecular interaction of perchloroethylene in mice and rats as related to oncogen- icity. Toxicol. Appl. Pharm. 55:207-219. Schumann, A. M., T. R. Fox, and P. G. Watanabe. 1982a. ['4C]methylchloroform (1,1,1- trichloroethane): Pharmacokinetics in rats and mice following inhalation exposure. Tox- icol. Appl. Pharmacol. 62:390-401. Schumann, A. M., T. R. Fox, and P. G. Watanabe. 1982b. A comparison of the fate of inhaled methylchloroform (1, 1,1-trichloroethane) following single or repeated exposure in rats and mice. Fund. Appl. Toxicol. 2:27-32. Serota, D., B. Ulland, and F. Carlborg. 1984a. Hazelton chronic oral study in mice. Food Solvents Workshop I: Methylene Chloride, Bethesda, Md. Washington, D.C.: Nutrition Foundation. Serota, D., B. Ulland, and F. Carlborg. 1984b. Hazelton chronic oral study in rats, Food Solvents Workshop I: Methylene Chloride, Bethesda, Md. Washington, D.C.: Nutrition Foundation.

Methotrexate: Pharmacokinetics and Assessment of Toxicity Paut F. Morrison, Robert L. Dedrick, and Robert ]. Lutz I NTRODUCTION Many of the physical and biological processes encountered in physio- logical pharmacokinetics play a role in determining the pharmacokinetic behavior of the antifolate compound methotrexate (MTX). Thus, results of studies with this compound, with its long and extensive history as a subject of both experimental and theoretical research, provide good ex- amples of the quantitation of such phenomena. In addition, the biochemical origins of toxicity of this compound are reasonably well understood so that connections can be made between pharmacokinetic variables and toxic endpoints. In this paper, we briefly summarize MTX pharmacokinetics and then focus on the dose scheduling aspects of these pharmacokinetics that affect target tissue concentrations and expected toxicity. Because of the highly nonlinear character of the pharmacokinetics and pharmacodynamics, we will see that dose-toxicity relationships are complex. MTX PHARMACOKINETICS Methotrexate is a folate analog that, following administration, distrib- utes primarily to the non-fatty tissues of the body. The principal organs containing the compound are shown in Figure 1. Transport of MTX across the capillary and cell membranes of the liver, kidney, and skin is rapid, so that equilibrium ratios of tissue to plasma concentrations (plasma con- 410

Methotrexate: Pharmacokinetics and Toxicity 41 1 1 I PLASMA, ASP 1 , SPLEEN QL~ C]! ~ ~ . __ _- LIVER Biliary Secre Lion tl|~t ^ ~~f-c-s Gut tum~ G.l. TRACT l C]G OK KIDNEY I Urine _ __ - MUSCLE L SKIN l _______ ~ MARROW FIGURE 1 Scheme of principal organs in the MIX model. am ASK QB

412 PAU ~ F. MORR ISON ET AL. centrations > 1 ~M) are established on a time scale consistent with plasma flow limitation. This ratio is also established quickly in muscle, although transport across muscle cells is absent. The tissues of the gastrointestinal (GI) tract, spleen, and bone marrow differ from the others in that transport of MTX across cell membranes is slow. As a consequence initial delivery of MTX to these organs is membrane transport limited rather than blood flow limited. MTX is cleared from the body through both biliary and urinary routes. Most of the drug excreted into the bile passes through the intestine and is excreted focally, but the drug is also subject to partial intestinal reab- sorption and to metabolism by enteric bacteria (Breithaupt and Kuenzlen, 1982; Zaharko and Dedrick, 19841. Excretion of MTX by the kidney is a result of both glomerular filtration and tubular secretion. The net result of plasma protein binding, filtration, and saturable secretion (and possibly reabsorption) is a kidney clearance that is of the order of, but not generally equal to, creatinine clearance. Concentration, pH, and competitive anion effects have been observed (Zaharko and Dedrick, 19841. In some organs, most notably the liver and kidney but also in intestine, marrow, and many tumors (Baugh et al., 1973; Jacobs et al., 1977; Whitehead et al., 1975), MTX undergoes metabolism to active polyglu- tamate derivatives. Because these polyglutamates are capable of being retained in some tissues far longer than the parent MTX, the total exposure time to MTX and its polyglutamate derivatives may, depending on sched- ule, greatly exceed that to unreacted drug alone (Balinska et al., 1982; Kennedy et al., 1983; Morrison and Allegra, 19871. It has been argued that this effect is of great importance in the killing of tumor cells by high doses of MTX but of relatively lesser importance in normal intestinal mucosa and bone marrow cells (Goldman and Matherly, 1986), which do not allow large levels of MTX polyglutamyl derivatives to accumulate. MTX also undergoes hydroxylation by liver aldehyde oxidase to form 7-hydroxymethotrexate, a metabolite with a long half-life of 24 h in hu- mans (Breithaupt and Kuenzlen, 1982~. Immediately following 6-h in- fusions, the concentration of this metabolite in human plasma reaches about 6% of the MTX concentration; 12 h after the infusion, the hydroxy metabolite and MTX have similar concentrations. A similar time to at- tainment of equal concentrations is estimated to occur following bolus injection. Hence, the antifolate activity of this compound, as well as that of MTX, must often be considered several hours after administration of the drug to humans. MECHANISM OF TOXICITY The overall mechanisms by which MTX induces cell toxicity are rea- sonably well understood, although many details of these mechanisms are

Methotrexate: Pharmacokinetics and Toxicity 413 still under active investigation. The primary event is the very strong in- tracellular binding of MTX to dihydrofolate reductase, an enzyme needed for the continued production of folate cofactors required for both thymi- dylate (Figure 2) and punne biosynthesis. Binding of drug prevents this enzyme from allowing continued production of DNA precursors, resulting in a cessation of DNA synthesis and, if continued long enough, in cell death. Polyglutamate derivatives of MTX, when formed, bind even more strongly to the enzyme than the parent drug and are thus potent mediators of cytotoxicity themselves (Jolivet and Chabner, 19831. Even the poly- glutamates of 7-hydroxymethotrexate may have some of this activity (Goldman and Matherly, 19861. For the drug to block reductase sufficiently for cell kill, binding of enzyme must be greater than 95% complete (Jackson and Harrap, 1973, 19791. This high percentage is derived from the presence in cells of a quantity of reductase far in excess of the amount required to maintain adequate folate cofactor production for survival. Furthermore, because of / / CH FH 24 dUMP FH - NADP \/ + ~ 2 ~ - TMP l ~ NADPH MTX(PG) FIGURE 2 Site of dihydrofolate reductase inhibition by methotrexate (MTX) and its polyglu- tamates (MTXPG). Other compounds are abbreviated as follows: CH2FH4, methylene tetrahy- drofolate; FH2, dihydrofolate; FH4, tetrahydrofolate; dUMP, deoxyuridine monophosphate; IMP, thymidine monophosphate; DNA, deoxyribonucleic acid; NADP+ and NADPH, oxidized and reduced forms of nicotinamide adenine dinucleotide phosphate.

414 PAUL F. MORRISON ET AL. the very large buildup of normal dihydrofolate substrate that occurs behind the inhibited enzyme (Figure 2), such a high percentage is obtained only if the intracellular concentration of MTX (and its polyglutamates) is main- tained several orders of magnitude in excess of the reductase inhibition constant (10-~i M). Hence, effective intracellular concentrations of drug are in the 10-7 to 10-8 M range rather than 10-~ M, and the most immediate measure of cytotoxic potential is a high free intracellular con- centration of MTX. A METHOTREXATE PHARMACOKINETIC MODEL A mathematical model that describes the physiological pharmacoki- netics of MTX in several species is summarized in Figure 1. This has been described at length by Bischoff et al. (1971) and Dedrick et al. (1973), and therefore, only a brief overview is presented here. Eight organ regions have been included in the model, although the overall pharma- cokinetics is only a weak function of inclusion of the spleen. The most sensitive sites of normal tissue toxicity are the intestinal mucosa cells (GI tract) and bone marrow. The liver excretes drug into the bile which, after being delayed by transport through the biliary system, enters the intestinal lumen where some of it is reabsorbed. The model consists of the set of differential mass-balance equations constructed for each organ region. The parameters of the model are numerous, as summarized for the rat in Table 1. These are the parameters required for the model to simulate drug behavior over 0- to 4-h periods following bolus administration. For much longer periods of time or for periods following long-term infusion of drug, more constants and differential equations accounting for hydrox- ylation and polyglutamation are required. If one ultimately wishes to describe toxic drug effects in several species, these parameters must be available for each species. In general, the pro- cedures by which these parameter values may be obtained fall into two classes. The class to which each parameter belongs is identified in Table 1. The first class consists of the extracellular volume (ECF), organ plasma flow rate (Qj, organ volume (V), kidney clearance (kK), fecal transit time (kF), bile residence time art, and MTX reductase dissociation constant Aft. These parameters are relatively invariant in passing from one species to another (ECF and e), are known for all species and do not generally depend on drug (kF and at, or can be scaled on the basis of body size. All remaining parameters belong to the second class and require that in viva or closely allied experiments be performed on each species because no adequate a priori rules or scaling laws are available. In the case of the tissue/plasma distribution ratio R. the value for muscle is the same for all species because the drug only equilibrates between plasma and ECF in

Methotrexate: Pharmacokinetics and Toxicity 415 TABLE 1 Parameters of MTX Physiologic Pha~Tnacokinetic Model Class 1 Organ volume Extracellular volume Blood flow rate Kidney clearance Fecal transit time Bile residence time MTX reductase dissociation constant Class 2 Tissue/plasma distribution ratio Dihydrofolate reductase concentration Biliary clearance rate Intestinal reabsorption rate (saturable) Membrane transport rate (saturable) ( + Metabolism constants) vi ECF Q. kK kF R; al kL kG, KG kj, K! this organ, but the value is quite different across species for the kidney and liver. For example, the mouse concentrates fivefold more drug in liver than does the dog (Bischoff et al., 19711. No general rule exists across species for the ability of hepatocytes to transport and bind MTX, and thus separate experiments are required for each species to determine R. The tissue-specific dihydrofolate reductase concentration (in MTX equivalents, a) has also been placed in this class of parameters. After the fact, we know that these concentrations do not vary much across species, but when MTX was first under investigation, there was no a priori reason for assuming this to be the case. Arguments based on the observation that reductase, because of its role in deoxynucleotide synthesis, is most nec- essary for proliferating cells, and that tissue concentrations are therefore determined by the fraction of proliferating cells, have some appeal as a guide to interspecies extrapolation. At the outset of such an extrapolation, however, the question would still have remained as to the species-specific concentration in a single proliferating cell. Presently, there is also no straightforward way to scale biliary clearance Okay and intestinal reabsorption parameters from one species to the next. As an example, there appear to be significant differences between species with respect to saturation of liver clearance. In the rat, Kates and Tozer (1976) reported a Michaelis constant of 70 EM for excretion of MTX into bile; however, saturation is not observed in the mouse when plasma con- centrations are in substantial excess of this value (D. S. Zaharko and R. L. Dedrick, unpublished data). Intestinal reabsorption parameters also seem to require a species-specific investigation. Intestinal absorption ap- pears to be saturable, as reviewed by Zaharko and Dedrick (19841. For

416 PAUL F. MORRISON ET AL. compounds that are not actively absorbed at the intestinal wall, general models may eventually become available for interspecies extrapolation based upon molecular and solution features such as molecular weight, charge, ionic strength, and species-specific mucous composition and thick- ness; however, as yet, these models are only in their early stages of development (Peppas et al., 19841. Remaining parameters such as membrane transport constants (k, K) and metabolism rate constants for hydroxylation and polyglutamation are still other quantities that require species-specific work. In vitro studies on cell lines derived from different animals indicated that the Michaelis transport constant K was only weakly species dependent. The other transport con- stant k, reflecting membrane carrier density and maximum rate of trans- membrane transport, is more variable across cell lines and, furthermore, is not easily obtained from in vitro studies of normal transport-limited tissues. In vitro experiments designed to measure MTX uptake rates in perfused tissue specimens might be attempted, but in viva testing raises fewer questions about the representativeness of the experimental model. Metabolism rate constants are particularly difficult to extrapolate across species. Metabolic enzymes are subject to major interspecies variability, such as that due to differences in regulation or to gross structural differ- ences that affect substrate binding. It is nearly impossible to know a priori if such differences exist between species without direct testing. In the case of MTX, in which polyglutamation and hydroxylation reactions occur in the liver, in vitro assessment of enzyme parameters from cultured hepa- tocytes derived from various species may provide an initial look at whether a single set of tissue parameters has interspecies applicability. It is well known, however, that hepatocytes in culture rapidly diverge from their behavior in intact liver (Balinska et al., 1982), and standardization of assays across species could prove difficult. DOSE SCALING In strict form, dose scaling refers to the ability to estimate a tissue concentration at an arbitrary dose level by scaling the known concentration at some other level by the dose ratio. As long as the pharmacokinetics of the tissue region is governed by linear differential equations describing drug distribution and metabolism, this is an allowable procedure. MTX plasma concentrations are, in fact, scalable over a wide dose range after bolus administration (Dedrick et al., 19701. Figure 3 shows plasma concentrations in the rat over a dose range of 0.05 to 25 mg/kg administered intravenously (i.v.) (Dedrick et al., 1973), in which the last three doses (b through d) differ by factors of 10. It is apparent in these log-linear plots that the plasma concentration curves are virtually identical

Methotrexate: Pharmacokinetics and Toxicity 4 ~ 7 ~ 1.0 O 0.1 z o 0.01 U] ~ 0.001 I o 10 3 C' z o UJ cC 0.1 0.01 _ 1 0 n ~ Marrow Z 0~1 o 0.01 , , , , , 1 0.001 _ Am_ 1 Plan 0 20 40 60 80 100 ~ 20 140 3 MINUTES (a) 10 Marrow W \asma to 103 - z 100 o At ~ 10 o at. IJJ 1 I . ~ 20 ' ' ' ~ 'A i 01 MINUTES (c) Marrow \; \ \ 0 20 40 60 80 100 120 140 MINUTES (b) \asma Marrow . 0 20 40 60 80 100 120 140 MINUTES (d) FIGURE 3 Plasma and bone marrow concentrations of MIX in rat. The bone marrow con- centrations are total tissue concentrations and are thus an average over the extracellular and intracellular regions. These concentrations are nonzero and dose scalable at short times because equilibration between plasma and the extracellular space is essentially instantaneous, and sat- urability only occurs during transport from the extracellular to intracellular space. Solid lines represent model simulations. Data points were obtained from one rat at each time. Key: (a) 0.05 mg/kg i.v.; (b) 0.25 mg/kg i.v.; (c) 2.5 mg/kg i.v.; (d) 25 mg/kg i.v. SOURCE: Dedrick et al. (1973).

4 ~ ~ PAU ~ F. MORR ISON ET AL. except for a decade difference in ordinate scale. The first curve (a) also scales with dose by the appropriate factor of 5. This scalability arises, in spite of the presence of nonlinearities in the complete scheme, because over 95% of the volume of MTX distribution is composed of non-transport- saturating, rapidly equilibrating compartments, principally the kidney, liver, plasma, skin, and extracellular volume of the muscle. In addition, the small compartments that do exhibit nonlinear distribution, the intra- cellular spaces of the gut, spleen, and bone marrow, do not rapidly trans- port drug into their cells. Hence, several plasma half-lives in the rat (tin = 0.3 h) can pass after administration of drug in the therapeutic dose range before plasma levels fall to the point where return of the small amount of drug in these deep nonlinear compartments could influence plasma kinetics. On the other hand, if one is interested in assessing the drug delivery to the gut, spleen, and bone marrow, then nonlinear pharmacokinetics prevents dose scaling from applying. These organs exhibit two strong nonlinearities in the first few hours after drug administration: saturable uptake of MTX from plasma, and strong binding of MTX (and its poly- glutamates) to the target enzyme dihydrofolate reductase. This is exhibited in Figure 3, in which total MTX concentrations in rat bone marrow (ex- tracellular MTX + intracellular MTX) are plotted as a function of time for four doses (Dedrick et al., 19731. The bar denotes the concentration of reductase in this tissue. Note that plasma-to-marrow concentration ratios are not constant over the 140 min shown here, and thus that marrow concentration does not scale with dose. The total marrow concentration shown in Figure 3 does scale with dose at short times, but this only reflects the rapid equilibrium attained between plasma and extracellular space (the dominant transport before cell uptake becomes significant). Note also that the marrow curves of the two lower dose levels are flat at long times and lie below the reductase content bar, thereby reflecting strong nonlinear enzyme binding of the drug that enters the cells in the first few minutes of exposure. Because the tight binding prevents drug efflux from occur- ring, the mass of enzyme-bound drug reflects the cumulative result of transport into the cell. At 0.05 and 0.25 mg/kg, this mass scales with dose (0.02 versus 0.10 ~g/ml) and thus transport is linear. However, attempts at using linear transport to extrapolate from 0.25 to 2.5 mg/kg fail, an observation that is consistent with plasma levels at the higher dose exceeding a Michaelis transport constant of about 1 ~M, a value char- acteristic of the range found in a variety of cell lines (Goldman, 1969, 1971; Schilsky et al., 19811. Figure 4 shows the magnitude of this saturable transport effect by dose level as the difference between the dashed and solid lines. The solid line shows rat marrow concentrations when saturation

Methotrexate: Pharmacokinetics and Toxicity 4~9 t00 10 or 1.0 At lo he o - \ 2 S mg~g _ _ ~ _ - - _ - - - - - - - - \L- ~ § 0.1 0.25 0.05 GO' 20 40 60 80 TIME, min 100 120 140 FIGURE 4 Companson of model simulation of linear and saturable transport in bone marrow at several doses. The dashed lines represent the model simulations for linear transport to the intracellular compartment of bone malTow. The solid lines represent model simulations with saturable transport. SOURCE: Dedrick et al. (1973). is operative, while the dashed line, providing a poor fit to the data (not shown), shows the result when linear transport is assumed. Hence, it can be concluded that, for sufficiently short times (e.g., <4 h in the rat), many organ regions are dose scalable, while others, for example, the gut, marrow, and spleen, are not.

420 PAUL F. MORRISON ET AL. DOSE SCHEDULING We next turn our attention to the rather dramatic effects that dose scheduling has on toxic response and to the formal connection between MIX pharmacokinetics and toxic response. Up to this point, the discussion has mainly involved distributional events that occur after bolus dosage. We will now see that the time of inhibition of dihydrofolate reductase is the primary correlate with toxicity. Table 2 shows acute toxic response in terms of the lethal dose for 50% of mice (LDso) exposed to a variety of drug schedules (Zaharko, 19751. The first entry is for a bolus dose of 350 mg/kg, while the next entries correspond to divided doses of decreasing total dose, and the final entry corresponds to a 96-h infusion of a 3-mg/kg total dose. These results immediately show that response does not directly correlate with total dose. Decreasing dose by more than a factor of 100 led to an increase, rather than a substantial decrease, in toxicity. Furthermore, the area under the plasma concentration-time curve, a frequently used metric, does not cor- relate with toxic response. This can be seen in Figure 5 (Zaharko, 19751. The steep curve shows the MIX plasma concentration following the 350- mg/kg bolus dose, while the flat curve shows the concentration following the 3-mg/kg 96-h infusion. The area under the bolus curve is about 2 orders of magnitude greater than that under the infusion curve, yet toxic response is greater with infusion. The principal correlate with response is the length of time that dihy- drofolate reductase and, consequently, DNA synthesis are strongly inhib- ited. If inhibition of DNA synthesis is a good correlate, then one would expect that the onset of toxicity should be observable after the same inhibition time, regardless of dose schedule. For at least two infusion schedules, this has been observed experimentally in mouse small intestine (Zaharko et al., 19771. Figure 6 shows the recovery of DNA synthesis in mice, as measured by deoxyuridine incorporation into DNA, following an infusion of 1 fig of MTX/h for 48 h, a schedule that just barely avoids TABLE 2 Schedule Dependence of Methotrexate Toxicity in Mice - Peak plasma Individual concentration dose (mg/kg) Schedule Total dose (mg/kg) (M) Effect 350 Single dose 350 10-3 LDso 25 Twice daily 50 lO-4 LDSo 3 Every 3 h, 5 times, rest 8 h, and then 24 1o-s >Lids every 3 h, 3 times 0.5 Every 3 h, 20 times 10 lo-6 >LDso 0.8 ~g/h Infusion 96 h 3 10-8 >LDso

Methotrexate: Pharmacokinetics and Toxicity 421 1000 -~ 1 00 10 ~ O BY \ O O O \:, . ~_~ · o . -2 X 1O - 8M 0.001 . 1 . I . 1 . 1 . ~ 0 20 40 60 80 100 Ti me, Or FIGURE 5 Concentration of methotrexate in mouse plasma. Open circles, single dose of 350 mg/kg i.v.; closed circles, constant infusion of 0.8 ~g/h. SOURCE: Zaharko (1975). lethal toxicity. If severe inhibition of greater than 90% is considered, DNA synthesis is inhibited relative to control for 35 h. Figure 7 shows similar data for a 10-fold higher rate of infusion (10 ~g/h) but of shorter duration (17 h). Like the previous schedule, this one is designed to just barely avoid lethal toxicity at the end of the infusion period (Table 2 of Zaharko et al., 1977~. An inhibition time virtually identical to that above was observed, a period of about 30 h. Further indication of DNA synthesis inhibition time as an appropriate measure of toxicity comes from observing the trend in lethality if, at a fixed infusion rate, infusion times are lengthened. As expected, lethality increases. For 10 fig of MTX/h, lethality jumps from 20% following a 24-h infusion to 90% following a 56-h infusion (Zaharko et al., 19771.

200 I ~ 1 o at I o L, ILL I foot fir 50 |I SEM Controk ~ SEM Controk \ 422 PAUL F. MORRISON ET AL. am_ 1 , ~ , i l - b~fus~on _ _ _ 1\ ~ -1 1 1 10 20 30 40 50 ~ 70 80 90 HOURS FIGURE 6 Incorporation of 3H into DNA of small intestine following an MTX infusion of 1 ~g/h for 48 h and an injection of [3H]thymidine (dark circle) or [3H]uridine injection (open circle). Vertical standard error bars indicate the range of the two mice used per point. At least six controls were used for each experiment. The first 10 cm of the small intestine was used from each mouse. SOURCE: Zaharko et al. (1977). Lethality correlation with total dose exists for a fixed infusion rate, but not when extended over a range of infusion rates. To predict toxic response from a given dose and schedule, the phar- macokinetic model of MIX must be coupled to the inhibition of DNA

Methotrexate: Pharmacokinetics and Toxicity 423 I ~ ~ ~ 1 ~ 290 C) ~ 1~ on C) lo he o 0 100 at 50 _ ~ SEW Control 1 /1 ~ _ \Intus~on ~ ~ - \ ~~- _ ' - - 1 610 70 110 90 10 20 30 40 50 HOURS FIGURE 7 Same as described in the legend to Figure 6, except that the MTX infusion was 10 ,ug/h for 17 h. Solid dots without vertical error bars are an identical experiment with MTX (intraperitoneal dose of 25 mg/kg) given at the start of infusion; mean of two mice; the range was not included, but it was similar to those in other experiments. SOURCE: Zaharko et al. (1977). synthesis. In the mid-1970s, this was accomplished very simply by ob- serving that recovery of DNA synthesis occurred when MTX plasma concentrations, as measured by a competitive binding assay, fell below 10-8 M (Chabner and Young, 1973~. Figure 8 shows that recovery in

424 PAUL F. MORRISON ET AL. 175 , 150 125 2 100 a 75 50 25 - Non tumor bearing o- - -a Tumor bearing ( ) Dow : O 1~7 O_ _~>' _~ , ' 1,,,,, , O' O 10 - 8 [MTX] plasma t l I P P (350)/ ,'{50} I , I , , I , , I ,'(350},/ 11° /'i50} (A I I , / I . l (5) 10~9 FIGURE 8 Incorporation of [3H]ur~dine into DNA of mouse bone marrow as a percentage of the pretreatment rate. The indicated doses are in milligrams/kilogram, administered intraper~- toneally. SOURCE: Chabner and Young ( 1973). bone marrow was a strong function of this pseudo-threshold following bolus dosing from 5 to 350 mg/kg. Thus, the pharmacokinetic model outlined earlier only needed to be solved for the length of time that the plasma concentration remained above 10-8 M to infer toxic response. The 10-8 M value was interpreted as the free MTX concentration in equilibrium with just sufficient free reductase (about 5% of the total) to allow re- sumption of thymidylate synthesis. Since the mid-1970s, the interpretation of these observations has become much more complex. The reason for this was the discovery of significant metabolism of MTX. When only low bolus doses were considered, cy- totoxic concentrations of MTX did not exist for a sufficiently long period for significant metabolism to be detected. With the introduction of high- dose protocols, and their long infusion times, in the mid-1970s (Frei et al., 1975; Jaffe et al., 1978), metabolism became apparent, and new assays were developed for the detection of metabolites. Some of these metabolic events, particularly the polyglutamation of MTX (Balinska et al., 1982; Baugh et al., 1973; Jolivet et al., 1982;

Methotrexate: Pharmacokinetics and Toxicity 425 Momson and Allegra, 1987) were shown to create very active drug forms that cleared from the intracellular milieu of cells at rates that were far slower than the clearance of parent drug from the plasma. Hence, a ratio- nale was needed to explain just why 10-8 M plasma concentrations have been observed to correlate so well with recovery of DNA synthesis and, more importantly, to identify the conditions under which this correlation might break down and when another approach to bridging MTX phar- macokinetics to toxicity would be required. Thus, there is no theoretical reason for expecting that when the 10-8 M concentration is reached, DNA synthesis should immediately resume under all protocols. - A full rationale requires more research, but preliminary analysis indi- cates that, under single-bolus conditions (as employed in Figure 8), cells are incapable of producing truly inhibiting quantities of polyglutamates, leaving the parent drug form alone to account for inhibition and recovery of DNA synthesis as observed. Under long-term infusion or closely spaced multiple bolus conditions, however, cells have much more time to produce polyglutamates and attain inhibiting levels of these compounds. Under these circumstances, MTX pharmacokinetic models need to be expanded to include polyglutamation kinetics, and intracellular MTX polyglutamate concentrations, rather than plasma concentrations of parent compound, need to be correlated with levels of DNA synthesis. SUMMARY In summary, we have seen that (1) only about half of the parameters of the (nonmetabolizing) physiological pharmacokinetic model can be obtained from chemical invariants and interspecies scalings; (2) dose scal- ing applies to the large-volume organ regions (e.g., kidney, liver, plasma) for a few hours after injection, but never to the principal organs of toxicity, the sensitive tissues of the marrow and intestinal mucosa; and (3) toxic response is a strong function of dose scheduling, correlating neither with total drug dose nor area under the MTX-plasma concentration curve, but with the time that MTX (and polyglutamate) concentrations remain above inhibiting levels of dihydrofolate reductase. REFERENCES Balinska, M., Z. Nimec, and J. Galivan. 1982. Characteristics of methotrexate polyglu- tamate formation in cultured hepatic cells. Arch. Biochem. 216:466-476. Baugh, C. M., C. L. Krumdieck, and M. G. Nair. 1973. Polygammaglutamyl metabolites of methotrexate. Biochem. Biophys. Res. Commun. 52:27-34. Bischoff, K. B., R. L. Dedrick, D. S. Zaharko, and J. A. Longstreth. 1971. Methotrexate pharmacokinetics. J. Pharm. Sci. 60:1128-1133.

426 PAUL F. MORRISON ET AL. Breithaupt, H., and E. Kuenzlen. 1982. Pharmacokinetics of methotrexate and 7-hydroxy- methotrexate following infusions of high-dose methotrexate. Cancer Treatment Rep. 66: 1733-1741. Chabner, B. A., and R. C. Young. 1973. Threshold methotrexate concentration for in vivo inhibition of DNA synthesis in normal and tumorous target tissues. J. Clin. Invest. 52:1804-1811. Dedrick, R. L., K. B. Bischoff, and D. C. Zaharko. 1970. Interspecies correlation of plasma concentration history of methotrexate (NCS-740) Cancer Treatment Rep. 54:95- 101. Dedrick, R. L., D. S. Zaharko, and R. J. Lutz. 1973. Transport and binding of methotrexate in viva. J. Pharm. Sci. 62:882-890. Frei, E., N. Jaffe, M. H. N. Tattersall, S. Pitman, and L. Parker, 1975. New approach to cancer chemotherapy with methotrexate. N. Engl. J. Med. 292:846-851. Goldman, I. D. 1969. Transport energetics of the folic acid analogue, methotrexate, in L1210 leukemia cells. J. Biol. Chem. 244:3779-3785. Goldman, I. D. 1971. The characteristics of the membrane transport of amethopterin and the naturally occurring folates. Ann. N.Y. Acad. Sci. 186:400-422. Goldman, I. D., and L. H. Matherly. 1986. The cellular pharmacology of methotrexate. Pp. 283-308 in Membrane Transport of Antineoplastic Agents, I. D. Goldman, ed. New York: Pergamon. Jackson, R. C., and K. R. Harrap. 1973. Studies with a mathematical model of folate metabolism. Arch. Biochem. Biophys. 158:827-841. Jackson, R. C., and K. R. Harrap. 1979. Computer models of anticancer drug interaction. Pharmacol. Ther. 4:245-280. Jacobs, S. A., C. J. Derr, and D. G. Johns. 1977. Accumulation of methotrexate diglutamate in human liver during methotrexate therapy. Biochem. Pharmacol. 26:2310-2313. Jaffe, N., E. Frei, H. Watts, and D. Traggis. 1978. High-dose methotrexate in osteogenic sarcoma; a 5-year experience. Cancer Treatment Rep. 62:259-264. Jolivet, J., and B. A. Chabner. 1983. Intracellular pharmacokinetics of methotrexate poly- glutamates in human breast cancer cells. J. Clin. Invest. 72:773-778. Jolivet, J., R. L. Schilsky, B. D. Bailey, J. C. Drake, and B. A. Chabner. 1982. Synthesis, retention, and biological activity of methotrexate polyglutamates in cultured human breast cancer cell. J. Clin. Invest. 70:351-360. Kates, R. E., and T. N. Tozer. 1976. Biliary secretion of methotrexate in rats and its inhibition by probenecid. J. Pharm. Sci. 65: 1348-1352. Kennedy, D. G., R. Clarke, H. W. van den Berg, and R. F. Murphy. 1983. The kinetics of methotrexate polyglutamate formation and efflux in a human breast cancer cell line: The effect of insulin. Biochem. Pharmacol. 32:41-46. Morrison, P. F., and C. J. Allegra. 1987. The kinetics of methotrexate polyglutamation in human breast cancer cells. Arch. Biochem. Biophys. 254:597-610. Peppas, N. A., P. J. Hansen, and P. A. Buri. 1984. A theory of molecular diffusion in the intestinal mucus. Int. J. Pharm. 20:107-118. Schilsky, R. L., B. D. Bailey, and B. A. Chabner. 1981. Characteristics of membrane transport of methotrexate by cultured human breast cancer cells. Biochem. Pharmacol. 30: 1537-1542. Whitehead, V. M., M. M. Perrault, and S. Stelcner. 1975. Tissue-specific synthesis of methotrexate polyglutamates in the rat. Cancer Res. 35:2985-2990. Zaharko, D. S. 1975. The kinetics of drug action. Pp. 69-83 in Pharmacological Basis of Cancer Chemotherapy, 27th Annual Symposium on Fundamental Cancer Research 1974. Baltimore, The Williams & Wilkins Co.

Methotrexate: Pharmacokinetics and Toxicity 427 Zaharko, D. S., and R. L. Dedrick. 1984. Pharmacokinetics of methotrexate in animals and man. Pp. 97-131 in Folate Antagonists as Therapeutic Agents, Vol. 2. Pharma- cology, Experimental and Clinical Therapeutics, F. M. Sirotnak, J. J. Burchall, W. D. Ensminger, and J. A. Montgomery, eds. Orlando, Fla.: Academic Press. Zaharko, D. S., W. P. Fung, and K.-H. Yang. 1977. Relative biochemical aspects of low and high doses of methotrexate in mice. Cancer Res. 37:1602-1607.

Next: VII. Summary: Prospectives and Future Directions »
Drinking Water and Health, Volume 8: Pharmacokinetics in Risk Assessment Get This Book
×
Buy Paperback | $90.00
MyNAP members save 10% online.
Login or Register to save!
Download Free PDF

Pharmacokinetics, the study of the movement of chemicals within the body, is a vital tool in assessing the risk of exposure to environmental chemicals. This book—a collection of papers authored by experts in academia, industry, and government—reviews the progress of the risk-assessment process and discusses the role of pharmacokinetic principles in evaluating risk. In addition, the authors discuss software packages used to analyze data and to build models simulating biological phenomena. A summary chapter provides a view of trends in pharmacokinetic modeling and notes some prospective fields of study.

  1. ×

    Welcome to OpenBook!

    You're looking at OpenBook, NAP.edu's online reading room since 1999. Based on feedback from you, our users, we've made some improvements that make it easier than ever to read thousands of publications on our website.

    Do you want to take a quick tour of the OpenBook's features?

    No Thanks Take a Tour »
  2. ×

    Show this book's table of contents, where you can jump to any chapter by name.

    « Back Next »
  3. ×

    ...or use these buttons to go back to the previous chapter or skip to the next one.

    « Back Next »
  4. ×

    Jump up to the previous page or down to the next one. Also, you can type in a page number and press Enter to go directly to that page in the book.

    « Back Next »
  5. ×

    To search the entire text of this book, type in your search term here and press Enter.

    « Back Next »
  6. ×

    Share a link to this book page on your preferred social network or via email.

    « Back Next »
  7. ×

    View our suggested citation for this chapter.

    « Back Next »
  8. ×

    Ready to take your reading offline? Click here to buy this book in print or download it as a free PDF, if available.

    « Back Next »
Stay Connected!